# AIxEnergy Journal > Independent intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, transmission, and energy policy. Public Ghost content for AI and LLM tooling. This file includes a bounded export of public pages first, then recent public posts. Append `.md` to any post or page URL to get the content in Markdown (for example, `/example-post.md`). ## Pages ### About URL: https://www.aixenergy.io/about/ Last updated: 2026-08-16T17:39:21.000Z Artificial intelligence is often described as a software revolution. Its next phase will be governed just as much by generation, transmission, substations, cooling, water, land, fiber, semiconductors, equipment, permitting, finance, cybersecurity, and public trust. The leading organizations of the AI era will not be defined only by the models they build. They will also be defined by their ability to finance, construct, operate, secure, and govern the systems that make large-scale intelligence possible. AIxEnergy Journal examines this emerging system before its effects are fully visible in load forecasts, interconnection queues, capital plans, utility tariffs, equipment orders, rate cases, investment decisions, reliability events, or local opposition. It serves utilities, data center developers, technology companies, infrastructure providers, investors, regulators, public agencies, researchers, advisers, and executive teams making decisions under uncertainty. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/ChatGPT-Image-Jun-15--2026--09_57_42-PM.png) ## Editorial Purpose AIxEnergy Journal exists because the most consequential questions at the intersection of AI and energy are rarely technological alone. A data center may be commercially attractive but impossible to energize on schedule. A utility may face credible load growth but insufficient evidence to justify major infrastructure commitments. A regulator may support economic development while resisting the transfer of risk to existing customers. An investor may see a large project announcement without knowing whether the developer controls the site, has secured power, ordered equipment, obtained permits, or committed capital. The same problem appears when AI enters the power sector. A forecasting model or operational copilot may perform well in a pilot yet remain unfit for sustained use because the utility lacks reliable data, validation procedures, cybersecurity controls, accountable ownership, or the technical workforce required to maintain it. The decisive issue is often not whether the technology works, but whether the surrounding institution can use it responsibly and repeatedly. The journal’s mission is to clarify these systems before constraints bind, capital is misallocated, public trust deteriorates, or decisions become difficult to reverse. It seeks to reveal the infrastructure beneath the models, the institutions behind the decisions, the incentives shaping behavior, and the consequences forming before consensus catches up. ## Coverage AIxEnergy Journal covers both sides of the emerging AI-energy system: the infrastructure required to support artificial intelligence and the use of AI within energy systems themselves. Its principal subjects include AI-driven electricity demand, data-center development, site selection, power procurement, large-load interconnection, utility tariffs, cost allocation, generation adequacy, transmission and distribution constraints, flexible service, on-site power, energy storage, cooling, water use, equipment supply chains, infrastructure finance, community impacts, permitting, and international competition for digital and energy infrastructure. The journal also examines utility AI strategy, forecasting, planning, operations, asset management, cybersecurity, data architecture, model validation, human oversight, agentic systems, workforce change, and the governance of delegated decisions in critical infrastructure. These subjects are treated as parts of one system. An article about electricity demand is incomplete if it ignores project maturity, equipment lead times, transmission capacity, tariffs, and risk allocation. An article about utility AI is incomplete if it ignores implementation, ownership, validation, cybersecurity, and institutional capability. An article about policy is incomplete if it does not identify who has authority, who must act, and who bears the cost of failure. ## Editorial Method AIxEnergy Journal combines engineering, economics, markets, regulation, capital, institutional analysis, and operating experience. It looks for the constraints and dependencies that determine whether a technology, project, or policy can move from announcement to execution. The journal distinguishes observed facts from forecasts, scenarios, hypotheses, proxy measures, illustrative calculations, expert judgment, and validated conclusions. Announced megawatts are not treated as operating demand. Numerical precision is not mistaken for causal proof. Forecasts are evaluated through their assumptions, time horizons, geographic concentration, and exposure to technological or commercial change. Every article is expected to address the decision beneath the topic. Who controls the relevant asset? Who has authority? Who pays? Who bears the downside? What evidence is missing? Which assumptions drive the conclusion? What must happen next? What would materially weaken or disprove the argument? AIxEnergy Journal favors analysis that identifies structural change, exposes hidden constraints, clarifies risk ownership, distinguishes credible activity from speculation, or gives readers a practical way to act. It avoids material that is primarily promotional, derivative, or designed to generate attention without improving understanding. ## AIxEnergy Journal AIxEnergy Journal is the digital publication at AIxEnergy.io. It publishes original analysis, contributed essays, interviews, research interpretation, international perspectives, visual frameworks, market commentary, and decision tools. The journal is selective rather than comprehensive. It does not attempt to cover every announcement, filing, project, or product release. Its purpose is to identify developments that reveal deeper changes in infrastructure, markets, institutions, and risk. Digital publishing allows the journal to respond quickly while preserving the standards of a serious professional publication. Articles are edited for evidence, clarity, structure, practical relevance, and intellectual independence. When conditions change, material may be updated or corrected to preserve its usefulness. ## AIxEnergy Quarterly AIxEnergy Quarterly is the digital publication of AIxEnergy Journal. Published four times a year, each 50 page issue is organized around a defining theme and presents a curated body of analysis drawn primarily from the preceding quarter. The Quarterly is not an archive of the website. Selected articles may be revised, updated, shortened, expanded, combined, or redesigned for print. Original introductions, interviews, data displays, diagrams, sidebars, scorecards, and decision tools are added to create a coherent publication rather than a collection of unrelated pieces. Each issue is designed as a durable executive reference that readers can retain, annotate, circulate, and use in planning, investment, policy, and strategy. Selected digital articles are included only with the author’s permission and after editorial review. ## Contributors AIxEnergy Journal publishes work from an international network of utility leaders, engineers, researchers, data-center practitioners, technology executives, investors, lawyers, regulators, public officials, and community experts. Contributors are selected for the quality of their reasoning, evidence, research, and operating experience. They are not required to share a common position. The journal welcomes disagreement when it clarifies assumptions, tests prevailing narratives, or reveals legitimate differences among institutions, markets, and regions. Contributions may take the form of original essays, practitioner perspectives, interviews, research adaptations, technical explainers, policy analyses, visual frameworks, or jointly authored articles. The editorial process can begin with a completed draft, an outline, an interview, a presentation, or a set of observations. Unpublished academic research is handled carefully. AIxEnergy Journal will not knowingly disclose novel findings, proprietary frameworks, figures, or distinctive arguments in a manner that could compromise later scholarly publication. Articles related to unpublished research must remain complementary and are published only with the author’s approval. ## Editorial Independence AIxEnergy Journal is independent and non-partisan. Its purpose is analysis, not advocacy, vendor promotion, institutional marketing, or political persuasion. The journal does not sell editorial placement, accept paid articles, publish disguised promotional material, or allow commercial or institutional interests to determine its conclusions. Should sponsorship ever be introduced, it would be clearly identified and separated from independent editorial work. Editorial selection, contributor invitations, publication standards, issue themes, corrections, and final publication decisions rest with Brandon N. Owens, Founder and Editor. This structure is intended to ensure that ideas compete on the quality of their evidence and reasoning rather than the influence of money, status, or affiliation. Contributors remain responsible for their own analysis and conclusions. Publication does not imply endorsement by AIxEnergy Journal, its editor, other contributors, the author’s employer, or any affiliated institution. Authors are expected to disclose relationships that could reasonably affect how readers interpret their work. ## Publication Standards AIxEnergy Journal seeks writing that is factual, clear about uncertainty, technically credible, institutionally grounded, and relevant to real decisions. Sources should be appropriate to the claim, with preference given to primary documents, official filings, public data, technical reports, peer-reviewed research, and direct operating evidence. Writers should state what the evidence establishes, what they infer, what remains unknown, and what they propose. Material forecasts and assumptions should be visible rather than buried. Articles that function mainly as marketing, omit significant conflicts, or overstate uncertain evidence may be revised or declined. The editor may revise submissions for clarity, organization, length, consistency, and style. Substantive changes are reviewed with the author before publication. The journal may correct, update, withdraw, or annotate material when evidence changes or an error is identified. ## Use of Artificial Intelligence AIxEnergy Journal may use artificial intelligence in research, transcription, drafting, editing, analysis, visualization, formatting, and production. AI tools do not hold editorial authority. Human judgment remains responsible for source evaluation, fact-checking, interpretation, causal reasoning, editorial decisions, contributor review, and publication approval. Fluency is not treated as evidence, generated citations are not accepted without verification, and model confidence is not a substitute for validation. ## Founder and Chief Editor Brandon N. Owens is an energy infrastructure executive, author, strategist, and editor whose work spans power systems, markets, technology, innovation, capital, and infrastructure governance. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/Owens_Brandon_Headshot.png) His career has included leadership and analytical roles at the National Renewable Energy Laboratory, General Electric, S&P Global, Cambridge Energy Research Associates, Insight Sourcing Group, and other energy and technology organizations. His work has addressed renewable energy, grid modernization, digital systems, market transformation, infrastructure investment, commercialization, and executive strategy. Owens is the founder and editor of AIxEnergy Journal. He is the author of *The Wind Power Story (2019)* and *The Cognitive Grid (2025)* His work examines how engineering systems, markets, regulation, capital, institutions, and human behavior shape technological change. AIxEnergy Journal is an independent editorial project, separate from Owens’s institutional employment and official responsibilities. ## Contributions and Contact AIxEnergy Journal welcomes proposals from qualified authors working across artificial intelligence, energy, infrastructure, finance, engineering, regulation, public policy, and related fields. Strong proposals present a clear thesis, identify the intended audience, disclose relevant affiliations, and explain why the subject matters now. Selected digital articles may be considered for AIxEnergy Quarterly with the author’s permission. Editorial inquiries, contribution proposals, commercial-platform inquiries, and custom-research requests may be submitted through bowens@aixenergy.io. ## Disclaimer AIxEnergy Journal provides independent analysis and commentary for informational and educational purposes. It does not provide legal, financial, investment, engineering, regulatory, cybersecurity, tax, or accounting advice. The views expressed by contributors are their own and do not necessarily represent those of AIxEnergy Journal, its editor, their employers, affiliated institutions, clients, sponsors, or other contributors. References to companies, projects, technologies, products, policies, or institutions do not constitute endorsement. AIxEnergy Journal seeks to publish accurate and reliable information, but energy, technology, and infrastructure markets change rapidly. Readers remain responsible for obtaining appropriate professional advice and conducting their own analysis before making decisions. For contributor inquiries, strategic collaboration, executive intelligence, or advisory services, contact AIxEnergy Journal [Contact Us ](mailto:bowens@aixenergy.io?subject=AIxEnergy%20Inquiry) [Read our Privacy Policy](https://www.aixenergy.io/privacy-policy/) ### Privacy Policy URL: https://www.aixenergy.io/privacy-policy/ Last updated: 2026-06-06T04:16:32.000Z # AIxEnergy Privacy Policy **Effective date:** June 5, 2026 **Last updated:** June 5, 2026 AIxEnergy is an independent editorial, research, membership, and advisory platform focused on the intersection of artificial intelligence, energy systems, infrastructure, markets, policy, and sustainability. AIxEnergy is operated by **Clean Energy Research LLC d/b/a AIxEnergy**. This Privacy Policy explains how we collect, use, disclose, retain, and protect personal information when you visit **AIxEnergy.io**, subscribe to our newsletters or memberships, contribute content, participate in surveys or events, purchase research or advisory services, communicate with us, or otherwise interact with AIxEnergy. By using AIxEnergy, you acknowledge that you have read and understood this Privacy Policy. Where required by law, we will ask for your consent before collecting or using certain types of information. ## 1\. Who We Are For purposes of this Privacy Policy, “AIxEnergy,” “we,” “us,” or “our” means: **Clean Energy Research LLC d/b/a AIxEnergy** 17717 W. 78th Dr. Arvada, CO 80007 United States Email: [**bowens@aixenergy.io**](mailto:bowens@aixenergy.io) For purposes of the European Union General Data Protection Regulation, the United Kingdom General Data Protection Regulation, and similar privacy laws, Clean Energy Research LLC is the “controller” of personal information collected through AIxEnergy’s website, newsletter, membership, research, editorial, and advisory activities, unless we state otherwise in a written agreement. ## 2\. Scope of This Policy This Privacy Policy applies to personal information we collect through: AIxEnergy.io and related webpages; email newsletters and subscriber communications; paid or free memberships; research briefings, reports, portals, downloadable materials, and webinars; contributor, author, editor, and expert communications; consulting, advisory, market-intelligence, and client engagements; surveys, forms, comments, feedback, and event registrations; and ordinary business communications. This Policy does not apply to third-party websites, platforms, payment processors, social networks, embedded content, or services that we do not control. Those third parties maintain their own privacy policies and practices. ## 3\. Information We Collect We collect personal information in several ways, depending on how you interact with AIxEnergy. ### 3.1 Information You Provide Directly We may collect information that you voluntarily provide, including: Your name, email address, organization, title, role, professional affiliation, billing contact information, subscription preferences, login or membership information, communications with us, comments, survey responses, event registrations, article submissions, contributor biographies, headshots, social media handles, contractual information, and advisory or consulting materials you choose to share. If you purchase a paid subscription, research product, event, or advisory service, payment information may be processed by third-party payment providers such as Stripe, PayPal, or other billing platforms. We do not intentionally store full credit-card numbers on our own systems. ### 3.2 Information Collected Automatically When you visit or use our website or digital services, we and our service providers may automatically collect certain information, including: IP address; browser type; device type; operating system; referring pages; pages viewed; time spent on pages; links clicked; approximate location derived from IP address; date and time of visit; cookie identifiers; newsletter open and click activity; membership login activity; and other diagnostic, security, and usage information. We may collect this information through first-party cookies, server logs, Ghost CMS analytics, email delivery tools, Google Analytics or similar analytics services, security tools, and other standard web technologies. ### 3.3 Information From Third Parties and Public Sources We may receive or collect information from third-party and publicly available sources, including: Professional profiles; company websites; public regulatory filings; public agency records; conference or event pages; social media platforms; publication databases; payment verification services; fraud-prevention tools; referral sources; and business contact databases. We use this information to support editorial research, contributor coordination, audience development, client service, market intelligence, fraud prevention, and ordinary business operations. ### 3.4 Information in Editorial, Research, and Advisory Work AIxEnergy publishes analysis concerning companies, technologies, infrastructure projects, markets, government policy, and public figures acting in professional capacities. Our editorial and research work may include publicly available professional information about individuals, such as names, titles, affiliations, public statements, publications, biographies, and professional contact details. Where we receive confidential client information, non-public business information, or personal information in connection with an advisory engagement, that information is also governed by any applicable written agreement between AIxEnergy and the client. ### 3.5 Sensitive Personal Information AIxEnergy does not seek to collect sensitive personal information such as precise geolocation, government identification numbers, financial account credentials, health information, biometric identifiers, race or ethnicity, religious beliefs, union membership, sexual orientation, or children’s information. Please do not submit sensitive personal information unless it is strictly necessary for a specific engagement and we have agreed in writing to receive it. If we learn that we have collected sensitive personal information that is not necessary for our services, we may delete it. ## 4\. How We Use Information We use personal information for the following purposes: To operate, maintain, and improve AIxEnergy.io, newsletters, memberships, and related services; deliver free and paid content; manage subscriptions and accounts; process payments and invoices; provide research, advisory, consulting, and market-intelligence services; communicate with readers, subscribers, contributors, clients, and prospective clients; review, edit, publish, and promote contributed content; administer events, webinars, briefings, and surveys; personalize content and understand reader interests; analyze site performance and audience trends; protect our systems from fraud, abuse, unauthorized access, and security threats; comply with legal, tax, accounting, contractual, and regulatory obligations; enforce our terms, agreements, and policies; preserve legal claims and business records; and develop new editorial, research, subscription, and advisory offerings. We may also use aggregated, anonymized, or de-identified information for research, analytics, benchmarking, editorial planning, product development, and business strategy. We do not attempt to re-identify information that has been properly de-identified, except as permitted by law for security, compliance, or validation purposes. ## 5\. Legal Bases for Processing Where applicable privacy law requires a legal basis for processing, we rely on one or more of the following grounds: **Contractual necessity**, when processing is needed to provide a subscription, research product, advisory service, event registration, contributor arrangement, or other requested service. **Legitimate interests**, when processing supports our editorial operations, research, audience development, platform security, business communications, fraud prevention, service improvement, and ordinary business administration, provided those interests are not overridden by your rights and interests. **Consent**, when you subscribe to certain marketing communications, accept non-essential cookies where required, participate in optional surveys, or otherwise give us permission to use your information for a specific purpose. **Legal obligation**, when processing is necessary to comply with tax, accounting, corporate, regulatory, court, law-enforcement, or other legal requirements. **Protection of rights and safety**, when processing is necessary to protect AIxEnergy, our users, clients, contributors, systems, or the public from fraud, security threats, abuse, or legal claims. ## 6\. Cookies, Analytics, and Similar Technologies We use cookies, pixels, tags, local storage, server logs, and similar technologies to operate our website, maintain secure sessions, remember preferences, measure site performance, understand readership, improve editorial focus, and support subscription or membership functionality. Some cookies are necessary for the website to function. Others are used for analytics, performance measurement, or marketing-related purposes. Depending on your location and applicable law, we may request consent before placing non-essential cookies. You can control cookies through your browser settings. Blocking cookies may affect login, membership, payment, preference, analytics, or other site functions. We may use third-party services such as Ghost, Google Analytics, Stripe, PayPal, Google Workspace, email delivery platforms, and similar vendors. These providers may collect or process information according to their own privacy policies and contractual obligations. ## 7\. Newsletters, Memberships, and Marketing Communications If you subscribe to AIxEnergy newsletters, memberships, alerts, research updates, event notices, or marketing communications, we may use your contact information and engagement data to send relevant communications. You may unsubscribe from marketing emails at any time by using the unsubscribe link in the email or by contacting us at [**bowens@aixenergy.io**](mailto:bowens@aixenergy.io). We may still send non-marketing messages related to your account, subscription, payment, legal notices, security, contributor relationship, or client engagement. ## 8\. Contributors, Authors, Experts, and Guest Writers If you submit an article, commentary, expert response, interview, biography, image, headshot, or other material for potential publication, we may use the information you provide to evaluate, edit, publish, attribute, promote, archive, and distribute that content. Published contributor information may remain publicly available as part of AIxEnergy’s editorial archive unless we agree otherwise in writing or unless applicable law requires removal. We may retain editorial records, drafts, correspondence, rights documentation, publication history, and related metadata for business, legal, archival, and editorial-integrity purposes. ## 9\. AI Tools and Automated Processing AIxEnergy may use artificial intelligence, machine-learning, automation, transcription, summarization, editing, analytics, research, or productivity tools to support editorial, research, administrative, and advisory workflows. We do not use AI tools to make legally binding decisions about individuals without appropriate human review. We also do not knowingly submit confidential client information, unpublished contributor materials, or sensitive personal information into public AI systems unless authorized, protected by appropriate terms, or otherwise permitted by the relevant agreement and law. Where AI tools are used in client work, the applicable statement of work, engagement letter, or client agreement may provide additional terms. ## 10\. How We Share Information We do not sell personal information in the ordinary meaning of selling names, emails, or profiles for money. We may disclose personal information to the following categories of recipients: **Service providers and processors**, including website hosting providers, Ghost CMS, email delivery platforms, payment processors, analytics providers, cloud storage providers, security vendors, customer-support tools, accounting systems, document-management tools, and other technology providers that help us operate AIxEnergy. **Professional advisers**, including lawyers, accountants, auditors, insurers, bankers, and consultants, where reasonably necessary for business, legal, tax, insurance, compliance, or risk-management purposes. **Clients and collaborators**, where necessary to perform an advisory, research, editorial, sponsorship, event, or consulting engagement, and subject to applicable agreements. **Publication partners, contributors, and editors**, where necessary to review, edit, fact-check, publish, attribute, distribute, or promote editorial content. **Government, regulatory, legal, or law-enforcement authorities**, where required by law, subpoena, court order, legal process, or to protect rights, safety, security, or property. **Successor entities**, in connection with a merger, acquisition, financing, reorganization, sale of assets, transfer of business, or similar transaction, provided the recipient is bound by appropriate confidentiality and privacy obligations. We may also share aggregated, anonymized, or de-identified information that does not reasonably identify you. ## 11\. No Sale or Targeted Advertising Statement AIxEnergy does not currently sell personal information for money. AIxEnergy also does not currently disclose personal information for cross-context behavioral advertising in the manner commonly understood under certain U.S. state privacy laws. If our practices change, we will update this Privacy Policy and provide any legally required opt-out mechanisms. ## 12\. International Data Transfers AIxEnergy is based in the United States, and our service providers may process information in the United States and other countries. If you access AIxEnergy from outside the United States, your information may be transferred to, stored in, or processed in jurisdictions that may not provide the same level of data protection as your home country. Where required, we rely on appropriate safeguards for international transfers, which may include standard contractual clauses, data-processing agreements, adequacy decisions, consent, contractual necessity, or other lawful transfer mechanisms. ## 13\. Data Retention We retain personal information only as long as reasonably necessary for the purposes described in this Privacy Policy, unless a longer retention period is required or permitted by law, contract, tax, accounting, audit, archival, dispute-resolution, or legitimate business needs. As a general guide: Subscriber and membership records are retained while your account is active and for a reasonable period thereafter. Billing, invoice, and tax records may be retained for at least seven years. Newsletter engagement and analytics records may be retained for up to twenty-six months or longer in aggregated or de-identified form. Advisory and client files may be retained for the duration of the engagement and for a reasonable period thereafter to support legal, contractual, audit, professional, and dispute-resolution needs. Editorial, contributor, publication, and archival records may be retained indefinitely where necessary to preserve the integrity, provenance, and historical record of published work. We may delete, anonymize, or de-identify information when it is no longer needed. ## 14\. Security We use reasonable administrative, technical, and organizational safeguards designed to protect personal information against unauthorized access, disclosure, alteration, loss, misuse, or destruction. These safeguards may include HTTPS encryption, access controls, authentication procedures, vendor review, software updates, limited internal access, data minimization, and secure business processes. No website, email system, cloud service, or internet transmission is completely secure. We cannot guarantee absolute security, but we work to maintain safeguards appropriate to the nature of the information we process. ## 15\. Your Privacy Choices and Rights Depending on where you live and the laws that apply, you may have rights to: Request access to personal information we hold about you; request correction of inaccurate information; request deletion of certain information; request portability of certain information; object to or restrict certain processing; withdraw consent where processing is based on consent; opt out of marketing communications; opt out of certain sales, sharing, targeted advertising, or profiling where applicable; and appeal a privacy-rights decision where applicable law provides an appeal right. To exercise a privacy right, contact us at [**bowens@aixenergy.io**](mailto:bowens@aixenergy.io) with the subject line **Privacy Request**. We may need to verify your identity before responding. We will respond within the time required by applicable law. We will not discriminate against you for exercising privacy rights. Some rights are subject to exceptions. For example, we may retain information where necessary for legal compliance, security, fraud prevention, accounting, editorial integrity, contractual performance, dispute resolution, or the establishment, exercise, or defense of legal claims. ## 16\. California Privacy Notice This section applies to California residents to the extent the California Consumer Privacy Act, as amended by the California Privacy Rights Act, applies to AIxEnergy. During the past twelve months, we may have collected the following categories of personal information: Identifiers, such as name, email address, IP address, account identifiers, and billing contact information; customer records, such as subscription, invoice, payment status, and service records; commercial information, such as subscriptions, purchases, memberships, research products, event registrations, and advisory services; internet or network activity, such as site usage, newsletter engagement, browser information, and device data; geolocation information, limited to approximate location derived from IP address; professional or employment-related information, such as title, organization, affiliation, biography, or contributor credentials; audio, visual, or similar information, such as webinar recordings, interview recordings, contributor headshots, or event materials if you choose to provide or participate in them; and inferences, such as content interests, subscription preferences, or audience segments. We collect this information from you, your devices, service providers, payment processors, public sources, professional sources, clients, contributors, and business partners. We use this information for the purposes described in Sections 4, 7, 8, and 9 of this Privacy Policy. We may disclose these categories of information to service providers, professional advisers, clients, collaborators, publication partners, authorities, and successor entities as described in Section 10. AIxEnergy does not knowingly sell personal information for money. AIxEnergy does not knowingly sell or share the personal information of individuals under sixteen years of age. California residents may have the right to know, access, correct, delete, obtain a copy of, and opt out of certain uses or disclosures of personal information, subject to legal exceptions. California residents may also have the right to limit certain uses of sensitive personal information if such information is collected and used in a manner covered by California law. To exercise California privacy rights, contact [**bowens@aixenergy.io**](mailto:bowens@aixenergy.io) with the subject line **California Privacy Request**. ## 17\. Colorado and Other U.S. State Privacy Rights Residents of Colorado, Connecticut, Virginia, Utah, and other states with comprehensive privacy laws may have rights to access, correct, delete, obtain a copy of, or opt out of certain processing of personal information, including certain targeted advertising, sales, or profiling, subject to legal thresholds and exceptions. AIxEnergy does not currently sell personal information or use personal information for targeted advertising in the manner commonly regulated by these state laws. If this changes, we will update this Privacy Policy and provide appropriate opt-out mechanisms. Colorado residents and residents of other applicable states may submit requests by contacting [**bowens@aixenergy.io**](mailto:bowens@aixenergy.io) with the subject line **State Privacy Request**. If we deny your request and your state law provides an appeal right, you may appeal by replying to our decision email with the subject line **Privacy Appeal**. ## 18\. European Economic Area, United Kingdom, and Switzerland Rights If you are located in the European Economic Area, United Kingdom, or Switzerland, you may have rights under applicable data-protection laws, including the right to access, correct, delete, restrict, object to processing, request portability, and withdraw consent. You may also have the right to lodge a complaint with your local data-protection authority. We encourage you to contact us first so we can try to resolve your concern. AIxEnergy does not have an establishment in the European Union or United Kingdom. If our activities require appointment of an EU or UK representative or data-protection officer in the future, we will update this Privacy Policy accordingly. ## 19\. Children’s Privacy AIxEnergy is intended for professionals, researchers, policymakers, investors, infrastructure leaders, and adult readers. It is not directed to children, and we do not knowingly collect personal information from children under sixteen. If you believe a child has provided personal information to AIxEnergy, contact us at [**bowens@aixenergy.io**](mailto:bowens@aixenergy.io), and we will take appropriate steps to delete the information where required. ## 20\. Third-Party Links, Embedded Content, and Social Platforms AIxEnergy may link to or embed content from third-party websites, publications, social media platforms, podcast platforms, video platforms, payment processors, event platforms, analytics providers, or other services. We are not responsible for the privacy practices, security, content, or policies of third-party services. Your interactions with third parties are governed by their terms and privacy policies. ## 21\. Business Clients and Contractual Data Where AIxEnergy provides research, advisory, consulting, market-intelligence, workshop, or enterprise services to a business client, the applicable agreement, statement of work, nondisclosure agreement, or data-processing agreement may supplement or override portions of this Privacy Policy with respect to the client engagement. Clients should not provide personal information, confidential information, regulated data, export-controlled information, sensitive personal information, or third-party proprietary information unless authorized to do so and unless the information is necessary for the engagement. ## 22\. Changes to This Privacy Policy We may update this Privacy Policy from time to time to reflect legal, technical, operational, editorial, or business developments. When we make material changes, we will update the effective date and may provide additional notice through AIxEnergy.io, email, account notices, or other appropriate means. Your continued use of AIxEnergy after an updated Privacy Policy becomes effective means you acknowledge the updated Policy. ## 23\. Contact Us For privacy questions, requests, or concerns, contact: **Clean Energy Research LLC d/b/a AIxEnergy** 17717 W. 78th Dr. Arvada, CO 80007 United States Email: [**bowens@aixenergy.io**](mailto:bowens@aixenergy.io) Because credible intelligence depends on trust, AIxEnergy treats privacy, confidentiality, and editorial integrity as part of the same obligation: protecting the people, institutions, and ideas that make serious analysis possible. ### Report Reviews URL: https://www.aixenergy.io/reviews/ Last updated: 2026-06-30T12:47:49.000Z [AEO 2026 in Review: A Structural Shift in U.S. Energy System DynamicsAEO2026 reframes its “Reference Case” as a control, not a forecast, and reveals a structural shift: electricity demand—driven by data centers and AI—is now the primary system variable, with load shape (continuous vs peak) redefining grid costs and planning![](https://static.ghost.org/v5.0.0/images/link-icon.svg)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/AEO2026_upscayl_3x_upscayl-standard-4x.png)](https://www.aixenergy.io/aeo2026/) [The Grid Meets the MachineA Review of the IEA’s Energy-and-AI Report—Why the Electricity Surge Is a Collision Between Code and Infrastructure![](https://static.ghost.org/v5.0.0/images/link-icon.svg)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/IEA_upscayl_5x_upscayl-standard-4x.png)](https://www.aixenergy.io/the-grid-meets-the-machine/) [The Age of Electricity Meets the Age of Constraint: IEA’s Energy Technology Perspectives 2026 Describes a System in TransitionEnergy is shifting from a single constraint (cost) to a multi-constrained system—power, materials, supply chains, and policy. Clean tech is scaling on economics, but fragility is rising. The transition now hinges not just on deployment, but on aligning constraints across the system.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-175.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/ETP_upscayl_5x_upscayl-standard-4x.png)](https://www.aixenergy.io/the-age-of-electricity-meets-the-age-of-constraint-ieas-energy-technology-perspectives-2026-describes-a-system-in-transition/) [The Shadow Grid Doctrine: The United States Is Building a Second Energy System for Artificial IntelligenceThe White House AI framework accelerates infrastructure at unprecedented speed—but exposes 10 systemic risks, from rising grid fragmentation and cost shifting to reliability, market and governance gaps, as a “Shadow Grid” emerges outside traditional oversight.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-166.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/final.png)](https://www.aixenergy.io/the-shadow-grid-doctrine/) [Energy Innovation at the Edge of Power: A Review of the IEA’s The State of Energy Innovation 2026The IEA’s 2026 report shows energy innovation is now central to competitiveness and security. Public investment drives long-term returns, resilient grids are essential, and institutional capacity—not resources alone—will define leadership in the twenty-first century energy transition.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-159.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/iea_upscayl_4x_upscayl-lite-4x.png)](https://www.aixenergy.io/energyinnovation/) [Extending AI Governance into Grid Governance: A Review of the OECD AI Report from an Infrastructure Systems PerspectiveThe OECD AI report offers a governance framework for advanced AI, but it is platform-centric. As AI embeds into grid operations and hyperscale load, governance must shift from model oversight to physical consequence—linking compute to megawatts, reliability margins, and ratepayer impact.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-154.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/OECD_upscayl_4x_upscayl-standard-4x.png)](https://www.aixenergy.io/extending-ai-governance-into-grid-governance-a-review-of-the-oecd-ai-report-from-an-infrastructure-systems-perspective/) [The IEA’s Electricity 2026 Meets AI: What the World’s Top Power Outlook Says—And What It Doesn’tIEA’s Electricity 2026 isn’t an AI report—but AI runs through it. Data centers drive demand, grids become the bottleneck, queues grow, and reliability risks rise. AI isn’t just adding load—it’s forcing new rules for how the grid is governed.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-153.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/Electricity2026_upscayl_4x_upscayl-standard-4x.png)](https://www.aixenergy.io/electricity2026/) [From Paradox to Progress — A Review of WEF’s Net-Positive AI Energy FrameworkAI has crossed from a marginal energy concern to a system-level driver of electricity demand, infrastructure investment, and grid planning. WEF and Accenture’s net-positive AI framework grounds this shift in lifecycle accounting, demand shaping, and real grid constraints.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-148.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/Cover_upscayl_3x_upscayl-standard-4x.png)](https://www.aixenergy.io/from-paradox-to-progress-a-review-of-wefs-net-positive-ai-energy-framework/) [When Advice Becomes Authority: AI, the Grid, and the Governance of Delegated JudgmentAI is quietly shifting grid operations by moving judgment upstream into decision-support software. The Cognitive Grid argues legitimacy requires permission that is legible, bounded, and auditable at machine speed—implemented via EthosGrid’s runtime governance layer.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-144.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/cover_without_text.jpg)](https://www.aixenergy.io/ctgarticle/) [The Role of Artificial Intelligence in World Energy Employment 2025The International Energy Agency’s (IEA) World Energy Employment 2025 offers a comprehensive assessment of global labor trends across the energy sector. While the report’s core narrative focuses on job growth, skilled-worker shortages, and demographic pressures, it also provides a substantive—though understated—examination of how artificial intelligence (AI)![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-138.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/WEE.png)](https://www.aixenergy.io/the-role-of-artificial-intelligence-in-world-energy-employment-2025/) [The Genesis Mission and the Architecture of National Scientific CapacityThe White House launched the Genesis Mission, directing DOE to build a unified AI platform that integrates federal data, HPC systems, and autonomous research tools to accelerate scientific discovery, with significant implications for compute capacity and energy infrastructure.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-134.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/genesis.png)](https://www.aixenergy.io/the-genesis-mission-and-the-architecture-of-national-scientific-capacity/) [The Age of Electricity Begins: AI Is Rewriting the Global Grid, Says IEAAI is reshaping global electricity demand. The IEA’s WEO 2025 shows data centers doubling power use by 2030 and tripling by 2035, straining grids in the U.S., Europe, and China. Renewables, gas, and nuclear rush to keep pace as AI becomes both a demand shock and an efficiency engine.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-118.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/weo2025_square.png)](https://www.aixenergy.io/the-age-of-electricity-begins-ai-is-rewriting-the-global-grid-says-iea/) [The Age of Compute: How AI and Data Centers Rewired Global Energy InvestmentGlobal energy investment hits $3.3 trillion in 2025, with two-thirds ($2.2 trillion) now flowing to clean energy. The IEA warns that AI and data centers are reshaping demand, potentially consuming 1,000 TWh by 2030 and driving a new “Age of Electricity and Intelligence.”![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-99.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/globalenergyinvestment-1.png)](https://www.aixenergy.io/worldinvestment/) [The Digital Dividend: How Data Centers Are Rewriting the IEA’s Renewables 2025 OutlookBy 2030, renewables will supply 43% of global power—over 4,600 GW, twice 2022’s total. Yet nearly one point of today’s demand growth stems from “AI + AC.” Data centers, once grid burdens, now accelerate renewables—adding 110 TWh, or \~54 GW, of clean capacity by 2030.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-96.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/social_bowenscolo_httpss.mj.runNNhKOFeasY8_--ar_4564_--video_1_c22b9569-cf14-4a8a-ab6b-84c7f6bf8925_0-ezgif.com-optimize.gif)](https://www.aixenergy.io/renewables2025/) [America’s AI Action Plan: An Energy and Infrastructure AppraisalThe America’s AI Action Plan recognizes AI’s massive energy demands and offers tools for permitting, workforce, and diplomacy. Yet it remains a political document, sidestepping climate goals and privileging fossil reliability over renewables—leaving decarbonization unfinished.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-73.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/whitehousecover.png)](https://www.aixenergy.io/americas-ai-action-plan-an-energy-and-infrastructure-appraisal/) [The Machine in the Control Room: Reviewing NERC’s White Paper on AI and ML in Real-Time System OperationsThe white paper is notable, not because it solves the AI challenge but because it acknowledges it. Its insistence on human engagement, explainability, and cybersecurity is both timely and necessary.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-72.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/bowenscolo_high-resolution_photo-realistic_wide_shot_of_a_rea_9f1043d8-0419-4672-a91d-b88efddbbcca_3.png)](https://www.aixenergy.io/the-machine-in-the-control-room-reviewing-nercs-white-paper-on-ai-and-ml-in-real-time-system-operations/) [Reviewing the DOE’s Case for AI-Driven Infrastructure ReformThe U.S. grid may be too weak to power its AI future, warns DOE—AI-driven demand could risk reliability without urgent infrastructure upgrades.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-71.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/doe-1.jpg)](https://www.aixenergy.io/the-grid-behind-the-algorithm-reviewing-the-does-case-for-ai-driven-infrastructure-reform/) [The AI Intelligence Revolution | New Energy RulesAI is reshaping global energy, driving surging electricity demand while unlocking breakthroughs in innovation, resilience, and sustainability. The future of energy and AI is intertwined—and the choices we make now will define both.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-70.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/IEA_AI_Cover-2.jpg)](https://www.aixenergy.io/the-intelligence-revolution-how-ai-is-rewriting-the-rules-of-energy/) ### Books URL: https://www.aixenergy.io/books/ Last updated: 2026-07-25T20:33:56.000Z AIxEnergy Journal publishes books for readers who want to understand the machinery beneath the energy transition: the technologies, institutions, markets, and control systems reshaping modern infrastructure. These works move beyond headlines and technological hype. They trace the long development of energy systems from mechanical machines and industrial networks to software-mediated control, artificial intelligence, and the governance questions now emerging inside the grid. Together, they tell a larger story. Energy infrastructure is no longer defined only by steel, fuel, turbines, and wires. It is increasingly shaped by data, software, markets, and the institutions that determine how these systems are designed, operated, and held accountable. 0:00 /0:34 1× ## The Cognitive Grid ### Artificial Intelligence, Critical Infrastructure, and the Quiet Relocation of Authority into Software Artificial intelligence is moving from the edge of the energy system into its operating core. It now supports load forecasting, asset optimization, anomaly detection, distributed-energy coordination, market analysis, and infrastructure planning. These capabilities may improve speed, efficiency, and situational awareness, but they also raise a more consequential question: when software begins to shape operational judgment, where does authority actually reside? *The Cognitive Grid* examines the gradual relocation of judgment and authority into software. It explores how AI changes the governance of critical infrastructure, why recommendations can become decisions in practice, and how institutions must adapt when machine intelligence begins to influence systems on which society depends continuously. This is a book about electricity, artificial intelligence, and accountability. It is also a book about the next phase of infrastructure governance. The Cognitive Grid Artificial Intelligence, Critical Infrastructure, and the quiet relocation of authority into software. [View on Amazon → ](https://www.amazon.com/dp/B0GDP24DL5?ref=aixenergy.io) ## The Wind Power Story ### A Century of Innovation that Reshaped the Global Energy Landscape Before wind power became one of the world’s most important clean-energy technologies, it was an uncertain experiment. Its rise was shaped by inventors, engineers, utilities, policymakers, entrepreneurs, oil shocks, research laboratories, public subsidies, industrial strategy, and the persistent improvement of machines operating in difficult environments. What began as a scattered set of experiments became a global industry capable of changing power markets, manufacturing systems, and national energy strategies. *The Wind Power Story* traces that century-long transformation. It follows the people, technologies, companies, and policy choices that carried wind energy from the margins of the electricity system into the mainstream of global power generation. The book shows that energy transitions are rarely sudden breaks from the past. They are built over decades through experimentation, failure, institutional learning, and the gradual alignment of technology, markets, capital, and public purpose. The Wind Power Story A Century of Innovation that Reshaped the Global Energy Landscape. [View on Wiley → ](https://www.wiley.com/en-us/The+Wind+Power+Story%3A+A+Century+of+Innovation+that+Reshaped+the+Global+Energy+Landscape-p-9781118794180?ref=aixenergy.io) ## ### The Cognitive Grid Book URL: https://www.aixenergy.io/cognitivegrid/ Last updated: 2026-03-07T00:16:21.000Z ## Purchase Book The Cognitive Grid Artificial Intelligence, Critical Infrastructure, and the quiet relocation of authority into software. [View on Amazon → ](https://www.amazon.com/dp/B0GDP24DL5?ref=aixenergy.io) ## Book Description *The Cognitive Grid* argues that the most important change in modern critical infrastructure is not simply better analytics, but the quiet relocation of operational judgment into automated systems that can recommend—and increasingly trigger—real actions under constraint. The book frames the central risk as governance insufficiency: authority being exercised through software without clear, enforceable limits, attribution, or contestability, especially as decision cycles accelerate beyond the pace of traditional oversight. A recurring theme is the latency gap between machine-speed execution and human-speed governance, where accountability often shows up only after outcomes are already fixed. In response, the book proposes a “constitutional” approach to intelligent infrastructure: governance must become an execution-time property rather than a post-event explanation. It describes an architecture that separates proposing intelligence from authorizing authority, so that capability does not silently become permission. This is operationalized through a small set of enforceable instruments—constraints, permissions, proof thresholds, audit capture, and isolation—intended to make fast automation reviewable, revocable, and accountable in the moment decisions are made. 0:00 /0:34 1× ## Supporting Materials ### *The Cognitive Grid* Readers Guide [Readers\_GuideHow to read The Cognitive Grid: track judgment migration and apply governance carefully.Readers\_Guide.pdf201 KBdownload-circle](https://www.aixenergy.io/content/files/2026/01/Readers%5FGuide.pdf "Download") ### How *The Cognitive Grid* Relates to Other Nonfiction ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/01/Venn_Diagram.png) [The\_Cognitive\_Grid\_and\_the\_Modern\_Nonfiction\_CanonA canon-mapping essay placing The Cognitive Grid within modern governance nonfictionThe\_Cognitive\_Grid\_and\_the\_Modern\_Nonfiction\_Canon.pdf255 KBdownload-circle](https://www.aixenergy.io/content/files/2026/01/The%5FCognitive%5FGrid%5Fand%5Fthe%5FModern%5FNonfiction%5FCanon.pdf "Download") ### How EthosCore Governs ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/01/figure_1-A.jpg) [How\_EthosGrid\_GovernsExplains how EthosGrid enforces execution-time governance without expanding machine authority.How\_EthosGrid\_Governs.pdf290 KBdownload-circle](https://www.aixenergy.io/content/files/2026/01/How%5FEthosGrid%5FGoverns.pdf "Download") ### The EthosGrid Open Standard [Ethos\_Grid\_Open\_StandardA regulatory-ready framework for governing AI authority in critical infrastructure.Ethos\_Grid\_Open\_Standard.pdf213 KBdownload-circle](https://www.aixenergy.io/content/files/2026/01/Ethos%5FGrid%5FOpen%5FStandard.pdf "Download") ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### Executive Intelligence for the AI-Energy Economy URL: https://www.aixenergy.io/executive-intelligence/ Last updated: 2026-07-15T01:39:27.000Z The next phase of AI will be shaped by electricity, interconnection, transmission, cooling, water, land, permitting, capital, regulation, and public trust. These constraints are no longer background conditions. They are becoming the terms on which AI infrastructure can scale. We advise utilities, investors, developers, public agencies, technology companies, and executive teams facing high-stakes decisions about AI infrastructure, data centers, grid capacity, power markets, large-load interconnection, clean energy, and regulatory strategy. AIxEnergy turns fast-moving AI-energy signals into executive-ready analysis, strategic options, and decision support. ## Advisory Services | Service | Best For | Includes | | ------------------------------------------ | -------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------- | | **Full-Day Strategic Planning Session** | Executive teams facing a major AI infrastructure, data center, grid, investment, or policy decision | Full-day workshop, custom briefing materials, facilitated strategy session, executive readout, follow-up recommendations | | **Data Center & Grid Risk Assessment** | Utilities, developers, investors, and public agencies evaluating large-load growth or specific projects | Power availability, interconnection, transmission exposure, regulatory risk, water, permitting, community risk, executive presentation | | **Utility Large-Load Strategy Assessment** | Utilities and public institutions preparing for hyperscale, AI, manufacturing, or other major load growth | Load growth implications, tariff and service options, reliability risks, stakeholder positioning, strategic recommendations | | **AI Infrastructure Investment Diligence** | Investors, developers, technology companies, and strategic buyers | Market analysis, competitive review, risk assessment, commercialization judgment, executive recommendation | | **AI Infrastructure Market Assessment** | Organizations entering or evaluating a market, technology, policy issue, or strategic opportunity | Research, analysis, written assessment, executive briefing | | **Executive AI-Energy Briefing** | Boards, leadership teams, investors, agencies, and industry groups needing a clear view of the AI-energy shift | 90-minute private briefing, tailored presentation, Q&A, follow-up summary memo | | **Strategic Advisory Retainer** | Organizations needing ongoing judgment and interpretation | Two advisory calls per month, priority email access, intelligence updates, executive support | | **Strategic Advisory Partner** | Organizations needing deeper ongoing support | Weekly access, strategy reviews, board and investor preparation, executive materials, ongoing intelligence | ## Representative Engagements ## Who We Help | Audience | Typical Need | | ------------------------------------------ | -------------------------------------------------------------------------------------------------------- | | Utilities and grid operators | Large-load strategy, planning, tariffs, reliability, cost allocation, and regulatory positioning | | Hyperscalers and data center developers | Siting, power procurement, grid risk, water and cooling exposure, permitting, and public acceptance | | Investors and infrastructure funds | Diligence, market sizing, risk assessment, thesis validation, and timing judgment | | Public agencies and policymakers | AI infrastructure policy, economic development, ratepayer protection, reliability, and climate alignment | | Clean energy and grid technology companies | Market entry, positioning, buyer strategy, commercialization, and partnership development | | Executive teams and boards | Briefings, decision memos, risk framing, and strategic interpretation | ## Why AIxEnergy Most AI infrastructure analysis starts with compute demand. That is too narrow. The real constraint is the operating system beneath the AI economy: electricity, interconnection, transmission, water, cooling, permitting, regulation, capital, and public trust. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/06/Brandon_Owens_Landscape_Headshot.jpg) AIxEnergy helps leaders see the full system before constraints become expensive, political, or irreversible. The timing matters. AI infrastructure is moving faster than utility planning cycles, regulatory proceedings, transmission development, procurement rules, and public-sector decision-making. Organizations that wait for certainty will inherit the bottlenecks. Organizations that act early can shape where AI infrastructure goes, who pays for it, which technologies matter, and how the public responds. AIxEnergy was built for this moment. Led by Brandon N. Owens, an energy innovation executive, author, and strategist, AIxEnergy brings together deep experience in energy markets, grid modernization, clean technology commercialization, public policy, sustainability, corporate strategy, and AI infrastructure. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/06/1782760873323-1.jpg) We help utilities, investors, developers, public agencies, technology companies, and executive teams make better decisions at the point where AI growth collides with the physical energy system. The value is not another report. It is earlier visibility, sharper judgment, stronger strategy, and fewer expensive surprises. ## Contact Brandon Owens Founder, AIxEnergy AIxEnergy.io [Request a Strategy Conversation](mailto:bowens@aixenergy.io?subject=AIxEnergy%20Strategy%20Inquiry) [Read our Privacy Policy](https://www.aixenergy.io/privacy-policy/) ### Contributors Network URL: https://www.aixenergy.io/contributors/ Last updated: 2026-06-06T04:04:29.000Z Every major economic transformation creates new institutions. The age of oil produced new energy companies, new financial markets, new geopolitical relationships, and new schools of thought. The age of electrification produced utilities, regulators, engineering disciplines, and entirely new ways of organizing society. The internet era produced platforms, networks, and information ecosystems that reshaped commerce and communication. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/06/ChatGPT-Image-Jun-5--2026--09_55_52-PM_upscayl_4x_upscayl-standard-4x.png) Artificial intelligence is now producing its own infrastructure transition. Data centers are becoming industrial facilities. Electricity is becoming a strategic input to computation. Utilities are confronting load growth not seen in generations. Infrastructure investors are reallocating capital. Governments are reassessing industrial policy. Regulators are confronting questions that did not exist a decade ago. At the same time, artificial intelligence is beginning to enter the operating layer of infrastructure itself, changing how systems are planned, operated, financed, secured, and governed. These developments are too important to be understood through headlines, quarterly earnings calls, technology demonstrations, or policy debates alone. Understanding them requires technical expertise, historical perspective, operational experience, economic insight, and a willingness to connect developments that are typically analyzed in isolation. AIxEnergy was created to provide that forum. The Contributors Network brings together leaders, researchers, practitioners, investors, policymakers, operators, strategists, engineers, economists, and technologists who are helping define the future of artificial intelligence and infrastructure. Our ambition extends beyond publishing articles. We aim to build the leading intellectual community dedicated to understanding the physical economy of artificial intelligence and the systems that make intelligence possible. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/06/bowenscolo_Photorealistic_documentary_photograph_of_distingui_6335aab7-c24e-48f8-ae03-daed9391179b_3.png) # Why Contribute? Most publications focus on reporting developments. AIxEnergy focuses on understanding them. Our readers are not looking for information they can find elsewhere. They are seeking insight into what developments mean, why they matter, how they connect, where risks are accumulating, and where durable opportunities are emerging. Contributors join a growing network examining questions that will shape the next decade: - How will artificial intelligence reshape electricity demand? - How will utilities respond to hyperscale growth? - What new infrastructure models are emerging? - How will private power alter traditional planning assumptions? - What role will nuclear, natural gas, storage, transmission, and flexibility play? - How will AI transform infrastructure operations? - How should intelligent infrastructure be governed? - What are the geopolitical consequences of the AI race? - How will infrastructure shape economic competitiveness? - What institutions will emerge to govern increasingly intelligent systems? These are not narrow technical questions. They are questions about the future structure of economies, institutions, infrastructure systems, and national competitiveness. Contributors help shape that conversation. # Our Perspective AIxEnergy is built around a simple observation: artificial intelligence is no longer only a software story. It is increasingly an infrastructure story. The next phase of the AI era will be shaped not only by advances in models and computation, but also by electricity, cooling, land, fiber, capital, permitting, cybersecurity, reliability, governance, and public trust. The defining strategic question is no longer simply who can build intelligence. It is who can build the systems that make intelligence possible. Understanding that transition requires perspectives from multiple disciplines. No single industry, institution, profession, or academic field can explain it alone. That belief sits at the center of the Contributors Network. # Who We Seek We actively welcome contributions from individuals whose work intersects with the future of intelligence and infrastructure. ## Industry Leaders - Utility executives - Independent power producers - Data center developers - Infrastructure operators - Technology leaders - Energy market participants - Corporate strategists ## Researchers and Academics - Energy economists - Engineers - Artificial intelligence researchers - Public policy scholars - Infrastructure experts - Cybersecurity specialists - National laboratory researchers ## Investors and Financial Professionals - Infrastructure investors - Private equity professionals - Venture investors - Financial analysts - Strategic advisors ## Public Sector Leaders - Regulators - Government officials - Policy advisors - Defense and national security professionals - Public utility commission staff ## Independent Thinkers Some of the most important ideas originate outside traditional institutions. We welcome original thinkers who bring expertise, rigor, curiosity, and a distinctive perspective to the conversation. # What Makes AIxEnergy Different? Most publications specialize in a single domain. AIxEnergy exists at the intersection of several. Our work connects conversations that are typically separated: artificial intelligence, electricity, infrastructure, capital formation, markets, national security, public policy, and governance. This interdisciplinary perspective is increasingly necessary because the most important developments in artificial intelligence are no longer occurring solely inside software companies. They are occurring where technology, infrastructure, energy systems, finance, institutions, and public policy converge. Contributors become part of a community actively building a shared understanding of that transition. # Editorial Standards AIxEnergy is an independent, non-partisan publication dedicated to advancing understanding at the intersection of artificial intelligence, energy, infrastructure, markets, policy, and governance. Our mission is not advocacy. Our mission is understanding. We seek to identify emerging trends, examine competing viewpoints, analyze evidence, challenge assumptions, and explore the infrastructure consequences of artificial intelligence. We believe the decisions being made today will shape energy systems, economies, institutions, and societies for decades to come. Those decisions deserve rigorous analysis grounded in facts, evidence, experience, and reasoned judgment. As a matter of editorial policy, AIxEnergy does not accept advertising, sponsored content, paid placements, native advertising, promotional articles, lead-generation content, affiliate marketing, product endorsements, or company-sponsored editorial content. We do not publish articles designed primarily to market a company, promote a service, advertise a product, influence procurement decisions, support fundraising efforts, or advance commercial interests. We also do not publish partisan political advocacy, campaign commentary, ideological persuasion, or articles whose primary purpose is to support or oppose a political party, elected official, candidate, or political movement. Policy analysis is welcome and encouraged, but it must be grounded in evidence, institutional context, practical implications, and objective assessment rather than political positioning. Contributors may discuss organizations, technologies, projects, policies, markets, and companies when those references are relevant to the analysis. However, the purpose of the article must always be to improve understanding rather than promote an outcome. AIxEnergy values work that is independent, objective, evidence-based, analytical, thoughtful, accessible, forward-looking, and constructive. Contributors are not expected to agree with AIxEnergy’s frameworks, conclusions, or viewpoints. We welcome diverse perspectives and constructive disagreement. What we require is intellectual honesty, transparency of reasoning, respect for evidence, and a commitment to advancing understanding. The purpose of AIxEnergy is not to tell readers what to think. The purpose of AIxEnergy is to help readers think more clearly about the future. # Contributor Relationship & Editorial Governance AIxEnergy is built on a simple principle: the exchange of ideas is more valuable than the exchange of money. Participation in the Contributors Network is a non-financial exchange of value. Contributors are not charged to publish on the platform, and AIxEnergy does not compensate contributors for articles, essays, commentary, or analysis. Similarly, AIxEnergy does not place contributor content behind a paywall, sell individual contributor articles, or charge readers for access to a specific contributor’s work. The value exchanged through the Contributors Network is intellectual rather than financial. Contributors share expertise, experience, analysis, and perspective with a growing community of leaders working at the intersection of artificial intelligence and infrastructure. In return, contributors receive visibility, distribution, editorial support, professional recognition, participation in an interdisciplinary network of experts, and the opportunity to help shape one of the most important conversations emerging in energy, technology, infrastructure, and governance. AIxEnergy is committed to maintaining editorial independence from commercial, political, and institutional influence. The publication does not accept advertising, sponsored content, paid placements, native advertising, promotional articles, product endorsements, affiliate marketing, or company-sponsored editorial content. Editorial decisions are made on the basis of relevance, quality, originality, analytical rigor, and alignment with the mission of advancing understanding. Submission of an article, essay, commentary, or other contribution does not guarantee publication. All editorial, publication, legal, governance, and content decisions remain solely at the discretion of Brandon N. Owens, Founder and Editor of AIxEnergy. This includes decisions regarding article acceptance, editorial revisions, publication timing, placement, contributor eligibility, content standards, legal review, and interpretation of editorial policies. The objective of this structure is straightforward: to preserve intellectual independence, maintain editorial integrity, and create a trusted forum where ideas compete on the strength of their analysis rather than the influence of financial, political, or institutional interests. # Contributor Benefits Contributors receive more than publication. ### Visibility Published work reaches a growing audience of utility executives, infrastructure investors, policymakers, regulators, researchers, technology leaders, and industry practitioners. ### Professional Profile Every contributor receives a dedicated author profile highlighting expertise, affiliations, publications, and areas of focus. ### Intellectual Community Contributors become part of a growing network of experts examining one of the most consequential infrastructure transformations of the twenty-first century. ### AIxEnergy Quarterly If descired, selected articles may be featured in AIxEnergy Quarterly, our flagship executive intelligence journal focused on the AI-electricity transition. ### Long-Term Collaboration Outstanding contributors may be invited to participate in interviews, roundtables, special projects, future publications, editorial initiatives, and collaborative research efforts. ### Lasting Influence The most important infrastructure transitions are ultimately shaped by ideas. Contributors help develop the concepts, frameworks, narratives, and analyses that decision-makers use to understand a rapidly changing world. # Contributor Process We have intentionally designed a lightweight process that respects contributors’ time. ### Step 1: Initial Conversation Prospective contributors should submit a proposed topic, a brief summary of the idea, and relevant background information. ### Step 2: Editorial Discussion The editorial team works with contributors to refine topics, identify strategic implications, and ensure alignment with AIxEnergy’s mission. ### Step 3: Draft Submission Articles typically range from 1,000 to 3,000 words, although longer feature essays are occasionally accepted. Contributors may submit original articles, research summaries, strategic essays, commentary pieces, or adapted versions of existing work where rights permit. ### Step 4: Collaborative Editing Our editorial process emphasizes clarity, structure, accessibility, and strategic relevance while preserving the author’s voice and perspective. ### Step 5: Publication and Distribution Published articles are distributed through the AIxEnergy website and social channels. Selected articles may also be considered for inclusion in future editions of AIxEnergy Quarterly. # Founding Vision The most important conversations of the next decade will not occur exclusively within energy, technology, finance, policy, or national security. They will occur at the intersection of those domains. AIxEnergy was founded on the belief that understanding artificial intelligence requires understanding the infrastructure that makes intelligence possible, and that understanding infrastructure increasingly requires understanding artificial intelligence. The Contributors Network exists to bring together the people working at that intersection. If you are helping shape the future of energy, infrastructure, computation, markets, governance, or technology, we invite you to join the conversation. [Inquire About Collaboration ](mailto:bowens@aixenergy.io?subject=AIxEnergy%20Collaboration%20Inquiry) ### July 2026 Events URL: https://www.aixenergy.io/watchlist/ Last updated: 2026-07-21T02:47:42.000Z July 16 · 10:00 a.m. ET ### FERC July Open Meeting Federal Energy Regulatory Commission Regulation The Commission’s final scheduled open meeting before its August recess may produce signals affecting large-load integration, reliability, and transmission policy. [View event intelligence →](#ferc-open) July 16 · 8:00 a.m. MT ### NERC Large Loads Working Group North American Electric Reliability Corporation Reliability The test is whether large electronic loads begin to face consistent expectations for modeling, telemetry, disturbance behavior, and coordination. [View event intelligence →](#nerc-llwg) Mid-July · Filing window ### FERC Large-Load Resource Adequacy Reports Six jurisdictional RTOs and ISOs Regulation · Reliability The filings will show how organized markets distinguish credible large-load projects from speculative demand before committing customers to long-lived infrastructure. [View event intelligence →](#resource-reports) July 22 ### Alphabet Q2 2026 Financial Results Alphabet Investor Relations Markets Management commentary will test whether announced AI infrastructure spending is translating into construction, contracted power, and energized capacity. [View event intelligence →](#alphabet) July 23 · 9:00 a.m. ET ### FERC Conference on PJM Governance Federal Energy Regulatory Commission Governance The conference will test whether PJM can act faster without weakening reliability, competition, transparency, or consumer protection. [View event intelligence →](#pjm) Event intelligence ## Event Details Each brief identifies why the event matters, who has authority, AIxEnergy’s base case, and the evidence that would change the assessment. Jul 16 10 ET **FERC July Open Meeting** Federal Energy Regulatory Commission · Washington, D.C. and webcast + #### Why it matters This is FERC’s last scheduled open meeting before the August recess. Any electric-market action or commissioner statement touching large loads, transmission access, resource adequacy, or cost allocation could shape the next stage of the federal response to AI demand. #### What to watch - New large-load tariff direction or procedural action - Signals on flexible or non-firm service - Statements linking resource adequacy to cost causation - Transmission planning treatment of uncertain demand #### Who has authority FERC controls federal wholesale-market and transmission tariff decisions. RTOs, ISOs, transmission owners, utilities, and state regulators will determine how many resulting obligations are implemented. #### AIxEnergy base case Incremental direction is more likely than a comprehensive national framework. The larger signal may come through commissioner statements and procedural sequencing rather than a single headline order. #### What would change our view A formal action that accelerates tariff reform, specifies financial assurance for large loads, or establishes a clearer federal template for flexible service and resource adequacy. #### Decision exposure Developers benefit from clearer access rules. Existing customers face greater downside if speculative demand is embedded in long-lived infrastructure plans without enforceable risk-sharing. [Official source ↗](https://www.ferc.gov/news-events/events/july-16-2026-open-meeting-10082025?ref=aixenergy.io) Add to calendar Jul 16 8 MT **NERC Large Loads Working Group Hybrid Meeting** North American Electric Reliability Corporation + #### Why it matters Reliable integration of large electronic loads increasingly depends on dynamic models, disturbance behavior, ride-through, telemetry, and operating coordination rather than nameplate demand alone. #### What to watch - Common model requirements for computational loads - Ride-through and protection expectations - Telemetry and information-sharing obligations - An accelerated standards-development path #### Who has authority NERC and the regional reliability structure can shape standards and guidance. Implementation will depend on standards drafting, regional entities, balancing authorities, transmission operators, utilities, and customers. #### AIxEnergy base case Incremental harmonization is more likely than a single national operating rule. Near-term progress will probably focus on shared terminology, models, data, and study practices. #### What would change our view A concrete accelerated standards path with defined performance obligations for large electronic loads and a clear implementation schedule. #### Engineering signal Data center electrical design is moving closer to grid operations as synchronized computing loads introduce power-system behavior that static megawatt forecasts cannot represent. [Official source ↗](https://www.nerc.com/events/07-16-26-lltf-meeting?ref=aixenergy.io) Add to calendar Mid-Jul Window **FERC Large-Load Resource Adequacy Information Reports** Six jurisdictional RTOs and ISOs · Filing deadline window + #### Why it matters The central question is not whether regional forecasts show demand growth. It is whether organized markets can identify which proposed projects are likely to become operating load before committing customers to durable infrastructure. #### What to watch - Commercial-readiness tests and project milestones - Financial security and withdrawal consequences - Probability assumptions applied to proposed load - Expected board actions and possible tariff filings #### Who has authority FERC, the jurisdictional RTOs and ISOs, and participating transmission owners shape the federal response. State regulators retain important authority over retail rates and many cost-allocation outcomes. #### AIxEnergy base case Methods will differ materially across regions. Aggregate gigawatt totals will attract attention, but the more consequential evidence will sit in supporting schedules and project-screening assumptions. #### What would change our view Convergence around enforceable milestones, meaningful deposits, explicit attrition factors, and standardized treatment of duplicate or speculative requests. #### Customer exposure Weak screening allows strategic queue positions to influence generation and transmission plans. Stronger screening can protect existing customers while giving credible developers clearer pathways. [FERC source hub ↗](https://www.ferc.gov/?ref=aixenergy.io) Add deadline window Jul 22 **Alphabet Q2 2026 Financial Results Conference Call** Alphabet Investor Relations · Online + #### Why it matters Alphabet is one of the world’s largest buyers of AI infrastructure. Its results and management commentary provide a test of whether announced compute ambition is translating into physical deployment, contracted power, and utilized capacity. #### What to watch - Capital expenditure guidance and construction timing - Data center utilization and regional concentration - Power procurement and financing structures - Language separating committed demand from option value #### Who has authority Alphabet controls its capital program and procurement strategy. Utilities, generators, grid operators, regulators, equipment suppliers, and local authorities control many of the physical pathways to energization. #### AIxEnergy base case Capital spending will remain elevated, but deliverable electricity capacity will continue to lag announced compute plans in constrained regions. #### What would change our view Detailed disclosure connecting capital commitments to contracted power, firm energization dates, utilization, and a visibly diversified regional deployment strategy. #### Assumption at risk Announced compute investment and energized capacity do not move on the same schedule. Treating them as equivalent overstates near-term load and understates execution risk. [Investor relations ↗](https://abc.xyz/investor/?ref=aixenergy.io) Add to calendar Jul 23 9 ET **FERC Technical Conference on PJM Governance and Stakeholder Reforms** Docket AD26-7-000 · Washington, D.C. and webcast + #### Why it matters PJM governance is becoming an infrastructure variable. The region faces rising demand, resource adequacy pressure, contested stakeholder processes, and growing costs from institutional delay. #### What to watch - Changes to board authority and unilateral action - A stronger or more formal role for states - Section 205 filing rights and document architecture - Fast-path or time-bound stakeholder procedures - Voting thresholds, transparency, and minority positions #### Who has authority FERC can influence or require tariff changes. PJM’s board, members, transmission owners, states, and stakeholder institutions control much of the regional decision process and its practical legitimacy. #### AIxEnergy base case The conference will expose broad agreement that PJM must act faster but less agreement over which actors should gain authority and which procedural protections should be reduced. #### What would change our view Specific proposals that reallocate decision rights, establish firm timelines, or give states a defined role without creating another consultative layer. #### Risk allocation Faster governance can reduce infrastructure delay. Speed without disciplined reliability, competition, and cost tests can shift risk to consumers or incumbent market participants. [Official source and agenda ↗](https://www.ferc.gov/news-events/events/PJM-Governance-Stakeholder-Reforms?ref=aixenergy.io) Add to calendar Download all July events ### Definitions URL: https://www.aixenergy.io/dictionary/ Last updated: 2026-07-21T02:48:07.000Z ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/ChatGPT-Image-Jul-12--2026--01_54_57-PM-1.png) ## Contents Section-level contents. The alphabetical index that follows lists all 300 entries. | **No.** | **Section** | | ------- | ------------------------------------------------------------------- | | | [Using This Guide](#using-this-guide) | | | [How the 20 Domains Fit Together](#how-the-20-domains-fit-together) | | | [Alphabetical Index](#all-300-definitions) | | 01 | [**AI Infrastructure**](#ai-infrastructure) | | 02 | [**Artificial Intelligence**](#artificial-intelligence) | | 03 | [**Data Centers**](#data-centers) | | 04 | [**Electric Grid**](#electric-grid) | | 05 | [**Electric Utilities**](#electric-utilities) | | 06 | [**Power Markets**](#power-markets) | | 07 | [**Generation**](#generation) | | 08 | [**Transmission**](#transmission) | | 09 | [**Distribution**](#distribution) | | 10 | [**Interconnection**](#interconnection) | | 11 | [**Reliability**](#reliability) | | 12 | [**Large Loads**](#large-loads) | | 13 | [**Semiconductors**](#semiconductors) | | 14 | [**Cooling and Water**](#cooling-and-water) | | 15 | [**Energy Storage**](#energy-storage) | | 16 | [**Regulation and Policy**](#regulation-and-policy) | | 17 | [**AI Governance**](#ai-governance) | | 18 | [**Climate and Sustainability**](#climate-and-sustainability) | | 19 | [**Supply Chains and Finance**](#supply-chains-and-finance) | | 20 | [**AIxEnergy Concepts**](#aixenergy-concepts) | ALPHABETICAL INDEX ## All 300 Definitions Terms are listed alphabetically. The domain label distinguishes repeated terms and provides context. | **Term** | **Domain** | | ----------------------------------------------------- | -------------------------- | | **#** | | | [24/7 clean energy](#entry%5FKB257) | Climate and Sustainability | | **A** | | | [Additionality](#entry%5FKB263) | Climate and Sustainability | | [Adequacy vs security](#entry%5FKB164) | Reliability | | [Advanced nodes](#entry%5FKB188) | Semiconductors | | [AI accelerators](#entry%5FKB182) | Semiconductors | | [AI agents](#entry%5FKB023) | Artificial Intelligence | | [AI auditability](#entry%5FKB248) | AI Governance | | [AI benchmarking](#entry%5FKB028) | Artificial Intelligence | | [AI campuses](#entry%5FKB014) | AI Infrastructure | | [AI compute infrastructure](#entry%5FKB001) | AI Infrastructure | | [AI demand drivers](#entry%5FKB029) | Artificial Intelligence | | [AI hardware competition](#entry%5FKB191) | Semiconductors | | [AI incident response](#entry%5FKB253) | AI Governance | | [AI infrastructure economics](#entry%5FKB004) | AI Infrastructure | | [AI infrastructure strategy](#entry%5FKB015) | AI Infrastructure | | [AI load corridors](#entry%5FKB010) | AI Infrastructure | | [AI Load Corridors](#entry%5FKB297) | AIxEnergy Concepts | | [AI operations](#entry%5FKB030) | Artificial Intelligence | | [AI safety](#entry%5FKB027) | Artificial Intelligence | | [AI workloads](#entry%5FKB022) | Artificial Intelligence | | [AI-ready facilities](#entry%5FKB034) | Data Centers | | [Air cooling](#entry%5FKB196) | Cooling and Water | | [Algorithmic accountability](#entry%5FKB254) | AI Governance | | [Ancillary services](#entry%5FKB079) | Power Markets | | [ASICs](#entry%5FKB192) | Semiconductors | | **B** | | | [Backup batteries](#entry%5FKB215) | Energy Storage | | [Backup power](#entry%5FKB038) | Data Centers | | [Battery safety](#entry%5FKB221) | Energy Storage | | [Battery storage](#entry%5FKB211) | Energy Storage | | [Behind-the-meter assets](#entry%5FKB132) | Distribution | | [Bulk power system](#entry%5FKB046) | Electric Grid | | **C** | | | [Campus development](#entry%5FKB039) | Data Centers | | [Capacity markets](#entry%5FKB077) | Power Markets | | [Capital discipline](#entry%5FKB285) | Supply Chains and Finance | | [Capital expenditure](#entry%5FKB275) | Supply Chains and Finance | | [Carbon accounting](#entry%5FKB256) | Climate and Sustainability | | [Chip packaging](#entry%5FKB184) | Semiconductors | | [Chip supply chains](#entry%5FKB189) | Semiconductors | | [Circuit hosting capacity](#entry%5FKB129) | Distribution | | [Circular economy](#entry%5FKB266) | Climate and Sustainability | | [Clean firm power](#entry%5FKB099) | Generation | | [Clean power procurement](#entry%5FKB262) | Climate and Sustainability | | [Climate resilience](#entry%5FKB270) | Climate and Sustainability | | [Climate risk](#entry%5FKB210) | Cooling and Water | | [Closed-loop systems](#entry%5FKB203) | Cooling and Water | | [Cloud infrastructure](#entry%5FKB007) | AI Infrastructure | | [Co-located generation](#entry%5FKB097) | Generation | | [Co-located storage](#entry%5FKB219) | Energy Storage | | [Cognitive Grid](#entry%5FKB287) | AIxEnergy Concepts | | [Coincident peak](#entry%5FKB177) | Large Loads | | [Collateral requirements](#entry%5FKB170) | Large Loads | | [Colocation](#entry%5FKB032) | Data Centers | | [Combined-cycle plants](#entry%5FKB101) | Generation | | [Commitment Discipline](#entry%5FKB290) | AIxEnergy Concepts | | [Community impacts](#entry%5FKB268) | Climate and Sustainability | | [Community-scale data centers](#entry%5FKB045) | Data Centers | | [Competitive transmission](#entry%5FKB108) | Transmission | | [Compute as strategic resource](#entry%5FKB006) | AI Infrastructure | | [Compute density](#entry%5FKB195) | Semiconductors | | [Congestion pricing](#entry%5FKB080) | Power Markets | | [Construction labor](#entry%5FKB274) | Supply Chains and Finance | | [Contingency analysis](#entry%5FKB161) | Reliability | | [Cooling energy penalty](#entry%5FKB209) | Cooling and Water | | [Cooling retrofits](#entry%5FKB206) | Cooling and Water | | [Cooling systems](#entry%5FKB037) | Data Centers | | [Cooperatives](#entry%5FKB063) | Electric Utilities | | [Cost allocation policy](#entry%5FKB231) | Regulation and Policy | | [Cost causation](#entry%5FKB178) | Large Loads | | [Cost recovery](#entry%5FKB066) | Electric Utilities | | [Credible Deliverability](#entry%5FKB296) | AIxEnergy Concepts | | [Credit support](#entry%5FKB284) | Supply Chains and Finance | | [Critical infrastructure AI](#entry%5FKB251) | AI Governance | | [Critical load](#entry%5FKB162) | Reliability | | [Critical minerals](#entry%5FKB273) | Supply Chains and Finance | | [Critical path management](#entry%5FKB120) | Transmission | | [Curtailable service](#entry%5FKB173) | Large Loads | | [Curtailment terms](#entry%5FKB147) | Interconnection | | [Customer classes](#entry%5FKB068) | Electric Utilities | | [Customer obligations](#entry%5FKB179) | Large Loads | | [Customer protection](#entry%5FKB075) | Electric Utilities | | [Cybersecurity](#entry%5FKB245) | AI Governance | | **D** | | | [Data center economics](#entry%5FKB042) | Data Centers | | [Data center financing](#entry%5FKB281) | Supply Chains and Finance | | [Data center load forecasts](#entry%5FKB167) | Large Loads | | [Data center permitting](#entry%5FKB041) | Data Centers | | [Data center policy](#entry%5FKB238) | Regulation and Policy | | [Data governance](#entry%5FKB250) | AI Governance | | [Demand charge management](#entry%5FKB218) | Energy Storage | | [Demand response](#entry%5FKB084) | Power Markets | | [Developer credibility](#entry%5FKB180) | Large Loads | | [Diesel backup](#entry%5FKB095) | Generation | | [Digital infrastructure supply chain](#entry%5FKB012) | AI Infrastructure | | [Direct-to-chip cooling](#entry%5FKB199) | Cooling and Water | | [Dispatchable Flexibility](#entry%5FKB291) | AIxEnergy Concepts | | [Distributed energy resources](#entry%5FKB128) | Distribution | | [Distribution automation](#entry%5FKB123) | Distribution | | [Distribution grid](#entry%5FKB048) | Electric Grid | | [Distribution planning](#entry%5FKB121) | Distribution | | [Distribution reliability](#entry%5FKB130) | Distribution | | [Dynamic line ratings](#entry%5FKB114) | Transmission | | [Dynamic load models](#entry%5FKB160) | Reliability | | **E** | | | [Economic development policy](#entry%5FKB235) | Regulation and Policy | | [Edge AI infrastructure](#entry%5FKB008) | AI Infrastructure | | [Electricity Customer Bill of Rights](#entry%5FKB293) | AIxEnergy Concepts | | [Embodied carbon](#entry%5FKB267) | Climate and Sustainability | | [Emergency operations](#entry%5FKB156) | Reliability | | [Emissions impact](#entry%5FKB259) | Climate and Sustainability | | [Energization timelines](#entry%5FKB145) | Interconnection | | [Energy markets](#entry%5FKB078) | Power Markets | | [Enterprise AI adoption](#entry%5FKB026) | Artificial Intelligence | | [Enterprise data centers](#entry%5FKB033) | Data Centers | | [Environmental justice](#entry%5FKB269) | Climate and Sustainability | | [Environmental permitting](#entry%5FKB233) | Regulation and Policy | | [Equipment lead times](#entry%5FKB279) | Supply Chains and Finance | | [Evaporative cooling](#entry%5FKB202) | Cooling and Water | | [Explainability](#entry%5FKB249) | AI Governance | | [Export controls](#entry%5FKB190) | Semiconductors | | **F** | | | [Facilities studies](#entry%5FKB141) | Interconnection | | [Fast-track interconnection](#entry%5FKB142) | Interconnection | | [Federal-state coordination](#entry%5FKB236) | Regulation and Policy | | [Feeder capacity](#entry%5FKB122) | Distribution | | [FERC jurisdiction](#entry%5FKB226) | Regulation and Policy | | [Firm power](#entry%5FKB098) | Generation | | [Flexible interconnection](#entry%5FKB127) | Distribution | | [Flexible loads](#entry%5FKB171) | Large Loads | | [Foundation models](#entry%5FKB018) | Artificial Intelligence | | [Frequency response](#entry%5FKB155) | Reliability | | [Frequency stability](#entry%5FKB057) | Electric Grid | | [Fuel availability](#entry%5FKB103) | Generation | | **G** | | | [Generation interconnection](#entry%5FKB104) | Generation | | [Generation siting](#entry%5FKB105) | Generation | | [Generative AI](#entry%5FKB017) | Artificial Intelligence | | [Generator interconnection](#entry%5FKB137) | Interconnection | | [GPU clusters](#entry%5FKB003) | AI Infrastructure | | [GPUs](#entry%5FKB181) | Semiconductors | | [Grid as institution](#entry%5FKB298) | AIxEnergy Concepts | | [Grid congestion](#entry%5FKB051) | Electric Grid | | [Grid constraints](#entry%5FKB050) | Electric Grid | | [Grid emissions](#entry%5FKB260) | Climate and Sustainability | | [Grid modernization](#entry%5FKB052) | Electric Grid | | [Grid operations](#entry%5FKB055) | Electric Grid | | [Grid planning](#entry%5FKB054) | Electric Grid | | [Grid reliability](#entry%5FKB053) | Electric Grid | | [Grid resilience](#entry%5FKB060) | Electric Grid | | [Grid services](#entry%5FKB217) | Energy Storage | | [Grid visibility](#entry%5FKB059) | Electric Grid | | [Grid-edge operations](#entry%5FKB072) | Electric Utilities | | [Grid-enhancing technologies](#entry%5FKB115) | Transmission | | **H** | | | [HBM memory](#entry%5FKB183) | Semiconductors | | [Hedging](#entry%5FKB087) | Power Markets | | [High load factor customers](#entry%5FKB176) | Large Loads | | [High-voltage lines](#entry%5FKB110) | Transmission | | [Human oversight](#entry%5FKB247) | AI Governance | | [Hybrid resources](#entry%5FKB225) | Energy Storage | | [Hydropower](#entry%5FKB094) | Generation | | [Hyperscale data centers](#entry%5FKB031) | Data Centers | | **I** | | | [Immersion cooling](#entry%5FKB198) | Cooling and Water | | [Inference](#entry%5FKB019) | Artificial Intelligence | | [Infrastructure insurance](#entry%5FKB278) | Supply Chains and Finance | | [Infrastructure Intelligence](#entry%5FKB289) | AIxEnergy Concepts | | [Infrastructure readiness](#entry%5FKB009) | AI Infrastructure | | [Infrastructure Readiness](#entry%5FKB292) | AIxEnergy Concepts | | [Integrated resource planning](#entry%5FKB065) | Electric Utilities | | [Interconnection risk](#entry%5FKB148) | Interconnection | | [Inverter-based resources](#entry%5FKB058) | Electric Grid | | [Investor-owned utilities](#entry%5FKB061) | Electric Utilities | | [ISO 42001](#entry%5FKB243) | AI Governance | | **L** | | | [Large-load interconnection](#entry%5FKB136) | Interconnection | | [Large-load policy](#entry%5FKB230) | Regulation and Policy | | [Large-load service](#entry%5FKB069) | Electric Utilities | | [Large-load tariffs](#entry%5FKB166) | Large Loads | | [Latency and geography](#entry%5FKB013) | AI Infrastructure | | [Liquid cooling](#entry%5FKB197) | Cooling and Water | | [Load forecasting](#entry%5FKB126) | Distribution | | [Load management](#entry%5FKB134) | Distribution | | [Load ramping](#entry%5FKB175) | Large Loads | | [Load shape](#entry%5FKB174) | Large Loads | | [Load shedding](#entry%5FKB157) | Reliability | | [Load verification](#entry%5FKB144) | Interconnection | | [Local permitting](#entry%5FKB135) | Distribution | | [Location strategy](#entry%5FKB005) | AI Infrastructure | | [Locational marginal pricing](#entry%5FKB081) | Power Markets | | [Long-duration storage](#entry%5FKB212) | Energy Storage | | **M** | | | [Machine learning](#entry%5FKB016) | Artificial Intelligence | | [Market monitoring](#entry%5FKB088) | Power Markets | | [Market power](#entry%5FKB083) | Power Markets | | [Megawatt Economy](#entry%5FKB294) | AIxEnergy Concepts | | [Microgrids](#entry%5FKB133) | Distribution | | [Milestone deposits](#entry%5FKB139) | Interconnection | | [Minimum demand charges](#entry%5FKB169) | Large Loads | | [Model efficiency](#entry%5FKB024) | Artificial Intelligence | | [Model risk management](#entry%5FKB244) | AI Governance | | [Model scaling](#entry%5FKB021) | Artificial Intelligence | | [Modular data centers](#entry%5FKB044) | Data Centers | | [Moore's Law](#entry%5FKB187) | Semiconductors | | [Moratoria](#entry%5FKB234) | Regulation and Policy | | **N** | | | [Natural gas generation](#entry%5FKB091) | Generation | | [NERC oversight](#entry%5FKB228) | Regulation and Policy | | [NERC reliability standards](#entry%5FKB151) | Reliability | | [Network upgrades](#entry%5FKB143) | Interconnection | | [Networking chips](#entry%5FKB193) | Semiconductors | | [NIST AI RMF](#entry%5FKB242) | AI Governance | | [Non-firm service](#entry%5FKB172) | Large Loads | | [Nuclear power](#entry%5FKB092) | Generation | | **O** | | | [On-site generation](#entry%5FKB096) | Generation | | [Open models](#entry%5FKB025) | Artificial Intelligence | | [Operational AI](#entry%5FKB246) | AI Governance | | [Operational proof](#entry%5FKB299) | AIxEnergy Concepts | | [Operational reliability](#entry%5FKB153) | Reliability | | **P** | | | [Peaking resources](#entry%5FKB100) | Generation | | [Permitting risk](#entry%5FKB280) | Supply Chains and Finance | | [Phantom Data Centers](#entry%5FKB288) | AIxEnergy Concepts | | [Phantom data centers](#entry%5FKB168) | Large Loads | | [Phased energization](#entry%5FKB150) | Interconnection | | [Planning reserve margins](#entry%5FKB159) | Reliability | | [Power density](#entry%5FKB011) | AI Infrastructure | | [Power efficiency](#entry%5FKB186) | Semiconductors | | [Power procurement risk](#entry%5FKB277) | Supply Chains and Finance | | [Power purchase agreements](#entry%5FKB086) | Power Markets | | [Power usage effectiveness](#entry%5FKB036) | Data Centers | | [Power-to-intelligence conversion](#entry%5FKB300) | AIxEnergy Concepts | | [Price volatility](#entry%5FKB089) | Power Markets | | [Project finance](#entry%5FKB276) | Supply Chains and Finance | | [Public power](#entry%5FKB062) | Electric Utilities | | **Q** | | | [Queue reform](#entry%5FKB138) | Interconnection | | **R** | | | [Rack density](#entry%5FKB035) | Data Centers | | [Rate design](#entry%5FKB067) | Electric Utilities | | [Ratepayer protection](#entry%5FKB232) | Regulation and Policy | | [Recycled water](#entry%5FKB208) | Cooling and Water | | [Regional planning](#entry%5FKB117) | Transmission | | [Regulatory compact](#entry%5FKB240) | Regulation and Policy | | [Reliability accountability](#entry%5FKB165) | Reliability | | [Reliability entities](#entry%5FKB158) | Reliability | | [Reliability upgrades](#entry%5FKB118) | Transmission | | [Reliability-must-run resources](#entry%5FKB163) | Reliability | | [Renewable matching](#entry%5FKB258) | Climate and Sustainability | | [Renewables](#entry%5FKB093) | Generation | | [Resilience value](#entry%5FKB223) | Energy Storage | | [Resource adequacy](#entry%5FKB082) | Power Markets | | [Resource adequacy](#entry%5FKB152) | Reliability | | [Retail choice](#entry%5FKB085) | Power Markets | | [Right-of-way](#entry%5FKB113) | Transmission | | [RTO and ISO tariffs](#entry%5FKB229) | Regulation and Policy | | **S** | | | [Scarcity pricing](#entry%5FKB090) | Power Markets | | [Scope 2 emissions](#entry%5FKB261) | Climate and Sustainability | | [Semiconductor fabs](#entry%5FKB185) | Semiconductors | | [Service agreements](#entry%5FKB146) | Interconnection | | [Service transformers](#entry%5FKB125) | Distribution | | [Shadow Grid](#entry%5FKB286) | AIxEnergy Concepts | | [Site selection](#entry%5FKB040) | Data Centers | | [Small modular reactors](#entry%5FKB102) | Generation | | [Smart meters](#entry%5FKB131) | Distribution | | [Speculative load](#entry%5FKB149) | Interconnection | | [Speed to Power](#entry%5FKB295) | AIxEnergy Concepts | | [State public utility commissions](#entry%5FKB227) | Regulation and Policy | | [Storage dispatch](#entry%5FKB224) | Energy Storage | | [Storage duration](#entry%5FKB216) | Energy Storage | | [Storage economics](#entry%5FKB222) | Energy Storage | | [Storage interconnection](#entry%5FKB220) | Energy Storage | | [Stranded costs](#entry%5FKB283) | Supply Chains and Finance | | [Studies](#entry%5FKB140) | Interconnection | | [Substations](#entry%5FKB049) | Electric Grid | | [Sustainability claims](#entry%5FKB264) | Climate and Sustainability | | **T** | | | [Tariff design](#entry%5FKB239) | Regulation and Policy | | [Thermal design power](#entry%5FKB194) | Semiconductors | | [Thermal management](#entry%5FKB205) | Cooling and Water | | [Thermal storage](#entry%5FKB213) | Energy Storage | | [Training](#entry%5FKB020) | Artificial Intelligence | | [Training vs inference](#entry%5FKB002) | AI Infrastructure | | [Transformer lead times](#entry%5FKB112) | Transmission | | [Transformer supply chains](#entry%5FKB271) | Supply Chains and Finance | | [Transmission bottlenecks](#entry%5FKB111) | Transmission | | [Transmission cost allocation](#entry%5FKB109) | Transmission | | [Transmission grid](#entry%5FKB047) | Electric Grid | | [Transmission interconnection](#entry%5FKB107) | Transmission | | [Transmission planning](#entry%5FKB106) | Transmission | | [Transmission queues](#entry%5FKB119) | Transmission | | [Transmission reform](#entry%5FKB237) | Regulation and Policy | | [Transmission tariffs](#entry%5FKB116) | Transmission | | [Trustworthy AI](#entry%5FKB255) | AI Governance | | [Turbine supply chains](#entry%5FKB272) | Supply Chains and Finance | | **U** | | | [UPS systems](#entry%5FKB214) | Energy Storage | | [Uptime and resiliency](#entry%5FKB043) | Data Centers | | [Utility AI adoption](#entry%5FKB071) | Electric Utilities | | [Utility AI governance](#entry%5FKB241) | AI Governance | | [Utility balance sheets](#entry%5FKB282) | Supply Chains and Finance | | [Utility governance](#entry%5FKB074) | Electric Utilities | | [Utility planning](#entry%5FKB064) | Electric Utilities | | [Utility procurement](#entry%5FKB073) | Electric Utilities | | [Utility risk management](#entry%5FKB070) | Electric Utilities | | **V** | | | [Vendor governance](#entry%5FKB252) | AI Governance | | [Voltage management](#entry%5FKB124) | Distribution | | [Voltage ride-through](#entry%5FKB154) | Reliability | | [Voltage stability](#entry%5FKB056) | Electric Grid | | **W** | | | [Waste heat reuse](#entry%5FKB204) | Cooling and Water | | [Water permitting](#entry%5FKB207) | Cooling and Water | | [Water stress](#entry%5FKB201) | Cooling and Water | | [Water sustainability](#entry%5FKB265) | Climate and Sustainability | | [Water usage effectiveness](#entry%5FKB200) | Cooling and Water | | [Wholesale electricity markets](#entry%5FKB076) | Power Markets | SECTION 01 | 15 DEFINITIONS ## AI Infrastructure How compute systems, power density, geography, economics, and delivery constraints shape the physical foundation of artificial intelligence. **Terms in this section:** [AI compute infrastructure](#entry%5FKB001) · [Training vs inference](#entry%5FKB002) · [GPU clusters](#entry%5FKB003) · [AI infrastructure economics](#entry%5FKB004) · [Location strategy](#entry%5FKB005) · [Compute as strategic resource](#entry%5FKB006) · [Cloud infrastructure](#entry%5FKB007) · [Edge AI infrastructure](#entry%5FKB008) · [Infrastructure readiness](#entry%5FKB009) · [AI load corridors](#entry%5FKB010) · [Power density](#entry%5FKB011) · [Digital infrastructure supply chain](#entry%5FKB012) · [Latency and geography](#entry%5FKB013) · [AI campuses](#entry%5FKB014) · [AI infrastructure strategy](#entry%5FKB015) ### AI compute infrastructure What is AI compute infrastructure? AI compute infrastructure is the integrated system of processors, memory, storage, networking, software, data-center facilities, power delivery, and cooling used to train and run artificial-intelligence models. It includes both the digital equipment that performs computation and the physical systems that keep that equipment connected, powered, cooled, secure, and available. **Related definitions:** [Training vs inference](#entry%5FKB002) · [Networking chips](#entry%5FKB193) · [Data center financing](#entry%5FKB281) · [AI-ready facilities](#entry%5FKB034) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-001 ### Training vs inference Why does Training vs inference matter for AI infrastructure? Training is the process of adjusting a model’s parameters using large datasets and substantial computation; inference is the use of a trained model to produce predictions, classifications, or generated outputs. Training often creates large, concentrated computing campaigns, while inference may be more geographically distributed, latency-sensitive, and variable with user demand. **Related definitions:** [AI compute infrastructure](#entry%5FKB001) · [GPU clusters](#entry%5FKB003) · [Colocation](#entry%5FKB032) · [Power procurement risk](#entry%5FKB277) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-002 ### GPU clusters How does GPU clusters affect electricity demand? A GPU cluster is a group of graphics processing units or similar accelerators linked by high-speed networks so they can operate as one computing system. Electricity demand reflects the number and type of accelerators, utilization, networking and memory loads, cooling requirements, and the efficiency losses between the grid connection and the chips. **Related definitions:** [Training vs inference](#entry%5FKB002) · [AI infrastructure economics](#entry%5FKB004) · [AI accelerators](#entry%5FKB182) · [Modular data centers](#entry%5FKB044) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-003 ### AI infrastructure economics What determines whether AI infrastructure economics can scale? AI infrastructure economics describe the costs, revenues, utilization, financing, and risk allocation associated with building and operating compute capacity. Scale depends on whether expensive chips, power systems, land, cooling, networks, and long-lived facilities can remain sufficiently utilized while electricity prices, model economics, technology cycles, and customer demand continue to change. **Related definitions:** [GPU clusters](#entry%5FKB003) · [Location strategy](#entry%5FKB005) · [Data center economics](#entry%5FKB042) · [Power procurement risk](#entry%5FKB277) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-004 ### Location strategy How should decision-makers evaluate Location strategy? Location strategy is the process of choosing where to place compute facilities based on power availability, interconnection timing, network connectivity, land, water, climate, taxes, permitting, labor, security, and proximity to users. The cheapest site is not necessarily the best site if electricity cannot be delivered reliably or the facility cannot be completed on schedule. **Related definitions:** [AI infrastructure economics](#entry%5FKB004) · [Compute as strategic resource](#entry%5FKB006) · [Site selection](#entry%5FKB040) · [Compute density](#entry%5FKB195) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-005 ### Compute as strategic resource What is Compute as strategic resource? Compute becomes a strategic resource when access to processing capacity materially affects economic competitiveness, scientific capability, military power, industrial development, or control over important digital services. Like energy, transport, and communications infrastructure, compute then becomes a question of capacity, resilience, supply concentration, investment, and public policy rather than merely an information-technology purchase. **Related definitions:** [Location strategy](#entry%5FKB005) · [Cloud infrastructure](#entry%5FKB007) · [Export controls](#entry%5FKB190) · [Grid resilience](#entry%5FKB060) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-006 ### Cloud infrastructure Why does Cloud infrastructure matter for AI infrastructure? Cloud infrastructure is the shared computing, storage, networking, and software environment delivered as an on-demand service from distributed data centers. For AI, cloud platforms aggregate expensive accelerators, provide development tools, and allow customers to scale without owning facilities, while concentrating significant electricity demand in the regions where cloud capacity is built. **Related definitions:** [Compute as strategic resource](#entry%5FKB006) · [Edge AI infrastructure](#entry%5FKB008) · [Colocation](#entry%5FKB032) · [Power efficiency](#entry%5FKB186) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-007 ### Edge AI infrastructure How does Edge AI infrastructure affect electricity demand? Edge AI infrastructure places computing resources close to devices, industrial systems, or end users rather than relying entirely on distant cloud data centers. It can reduce latency and network traffic, improve local resilience, and support privacy-sensitive applications, but it also spreads smaller power and cooling requirements across many sites that may have limited electrical capacity. **Related definitions:** [Cloud infrastructure](#entry%5FKB007) · [Infrastructure readiness](#entry%5FKB009)· [Modular data centers](#entry%5FKB044) · [Power efficiency](#entry%5FKB186) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-008 ### Infrastructure readiness What determines whether Infrastructure readiness can scale? Infrastructure readiness is the degree to which a location, institution, or project can convert an announced need into dependable operating capacity. It includes available grid capacity, interconnection studies, equipment, permits, land, water, communications, financing, construction capability, operating agreements, and clear responsibility for delivery rather than merely a favorable resource map. **Related definitions:** [Edge AI infrastructure](#entry%5FKB008) · [AI load corridors](#entry%5FKB010) · [Data center economics](#entry%5FKB042) · [Grid resilience](#entry%5FKB060) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-009 ### AI load corridors How should decision-makers evaluate AI load corridors? AI load corridors are geographic concentrations where data-center demand, transmission access, fiber networks, generation, land, incentives, and development pipelines reinforce one another. A corridor can accelerate investment, but it can also concentrate reliability, cost, water, permitting, and community risks that are less visible when projects are evaluated one site at a time. **Related definitions:** [Infrastructure readiness](#entry%5FKB009) · [Power density](#entry%5FKB011) · [Campus development](#entry%5FKB039) · [Grid congestion](#entry%5FKB051) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-010 ### Power density What is Power density? Power density is the amount of electrical load concentrated within a rack, room, building, or campus, commonly expressed in kilowatts per rack or watts per unit of floor area. Higher density changes equipment layout, cooling technology, power distribution, redundancy design, fire protection, maintenance practices, and the size and speed of the required utility connection. **Related definitions:** [AI load corridors](#entry%5FKB010) · [Digital infrastructure supply chain](#entry%5FKB012) · [Rack density](#entry%5FKB035) · [Compute density](#entry%5FKB195) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-011 ### Digital infrastructure supply chain Why does Digital infrastructure supply chain matter for AI infrastructure? The digital infrastructure supply chain includes the chips, servers, networking equipment, transformers, switchgear, generators, batteries, cooling systems, cables, construction materials, software, logistics, and skilled labor needed to deliver compute capacity. A project can have sufficient capital and customer demand yet still be delayed by a single constrained component or specialized workforce. **Related definitions:** [Power density](#entry%5FKB011) · [Latency and geography](#entry%5FKB013) · [Export controls](#entry%5FKB190) · [Data center economics](#entry%5FKB042) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-012 ### Latency and geography How does Latency and geography affect electricity demand? Latency is the time required for data to travel through a system and for a response to return. Geography affects latency because physical distance, network routing, congestion, and interconnection points add delay; applications such as real-time control, trading, gaming, and interactive AI may therefore value proximity more than workloads that can run asynchronously. **Related definitions:** [Digital infrastructure supply chain](#entry%5FKB012) · [AI campuses](#entry%5FKB014)· [Power procurement risk](#entry%5FKB277) · [ASICs](#entry%5FKB192) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-013 ### AI campuses What determines whether AI campuses can scale? An AI campus is a large, coordinated development containing multiple data-center buildings and shared electrical, cooling, communications, security, and support infrastructure. Campuses can scale in phases, but their ultimate load may be far larger than the initial building, so utilities and communities must evaluate the full development plan rather than only the first energization request. **Related definitions:** [Latency and geography](#entry%5FKB013) · [AI infrastructure strategy](#entry%5FKB015) · [Campus development](#entry%5FKB039) · [Stranded costs](#entry%5FKB283) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-014 ### AI infrastructure strategy How should decision-makers evaluate AI infrastructure strategy? AI infrastructure strategy is a coordinated plan for securing compute, power, sites, networks, capital, suppliers, and operating capabilities over time. A credible strategy connects business demand with physical delivery, distinguishes committed projects from options, and accounts for regulatory, reliability, technology, cost, and concentration risks across the full infrastructure portfolio. **Related definitions:** [AI campuses](#entry%5FKB014) · [Grid reliability](#entry%5FKB053) · [Compute density](#entry%5FKB195) · [Bulk power system](#entry%5FKB046) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-015 SECTION 02 | 15 DEFINITIONS ## Artificial Intelligence The workloads, models, efficiency trends, and operating patterns that determine how artificial intelligence uses infrastructure and electricity. **Terms in this section:** [Machine learning](#entry%5FKB016) · [Generative AI](#entry%5FKB017) · [Foundation models](#entry%5FKB018) · [Inference](#entry%5FKB019) · [Training](#entry%5FKB020)· [Model scaling](#entry%5FKB021) · [AI workloads](#entry%5FKB022) · [AI agents](#entry%5FKB023) · [Model efficiency](#entry%5FKB024) · [Open models](#entry%5FKB025) · [Enterprise AI adoption](#entry%5FKB026) · [AI safety](#entry%5FKB027) · [AI benchmarking](#entry%5FKB028) · [AI demand drivers](#entry%5FKB029) · [AI operations](#entry%5FKB030) ### Machine learning What is Machine learning? Machine learning is a class of computational methods that identifies patterns in data and uses them to make predictions, classifications, recommendations, or decisions without relying only on explicitly programmed rules. Models learn from examples, but their performance depends on data quality, objective design, evaluation, operating context, and continued monitoring after deployment. **Related definitions:** [Generative AI](#entry%5FKB017) · [Model risk management](#entry%5FKB244) · [AI accelerators](#entry%5FKB182) · [ASICs](#entry%5FKB192) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-016 ### Generative AI How does Generative AI affect energy demand? Generative AI uses learned statistical patterns to create new text, images, audio, video, software, or other content. Its energy impact comes from model training, repeated inference, data movement, storage, and supporting infrastructure; total demand depends as much on adoption, utilization, model design, and hardware efficiency as on the capability of any single model. **Related definitions:** [Machine learning](#entry%5FKB016) · [Foundation models](#entry%5FKB018) · [AI accelerators](#entry%5FKB182) · [Training vs inference](#entry%5FKB002) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-017 ### Foundation models Why does Foundation models matter for utilities? A foundation model is a large, broadly trained model that can be adapted to many downstream tasks through prompting, fine-tuning, retrieval, or additional tools. Foundation models matter to utilities because they can support analysis and automation across many functions, while also introducing shared dependencies, opaque failure modes, data-governance requirements, and potentially concentrated computing demand. **Related definitions:** [Generative AI](#entry%5FKB017) · [Inference](#entry%5FKB019) · [Edge AI infrastructure](#entry%5FKB008) · [Vendor governance](#entry%5FKB252) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-018 ### Inference What infrastructure does Inference require? Inference is the execution of a trained model to produce an output from new input data. It requires processors, memory, storage, networking, software, and power, but its infrastructure needs vary widely: a small model may run on a device, while a high-volume service may require globally distributed accelerator clusters operating continuously. **Related definitions:** [Foundation models](#entry%5FKB018) · [Training](#entry%5FKB020) · [Training vs inference](#entry%5FKB002) · [Power efficiency](#entry%5FKB186) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-019 ### Training How could Training change grid operations? Training is the computational process used to fit a model to data by repeatedly updating its internal parameters. Large training runs can create high, sustained electrical loads and demand reliable access to accelerators and networks, but they may also be schedulable across time or location if data, commercial deadlines, and system architecture permit. **Related definitions:** [Inference](#entry%5FKB019) · [Model scaling](#entry%5FKB021) · [Training vs inference](#entry%5FKB002) · [Grid-edge operations](#entry%5FKB072) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-020 ### Model scaling What is Model scaling? Model scaling is the practice of increasing parameters, training data, computation, context, or other resources to improve model capability. Scaling can raise electricity and infrastructure requirements, but the relationship is not fixed because algorithmic improvements, specialized hardware, model compression, utilization, and changing workload mixes can alter the energy required for a given level of performance. **Related definitions:** [Training](#entry%5FKB020) · [AI workloads](#entry%5FKB022) · [Power efficiency](#entry%5FKB186) · [Training vs inference](#entry%5FKB002) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-021 ### AI workloads How does AI workloads affect energy demand? AI workloads are the specific computing tasks used to train, fine-tune, evaluate, and operate AI systems. Their energy profiles differ according to model architecture, batch size, precision, latency requirements, utilization, data movement, and whether work is continuous or episodic; treating all AI demand as one uniform load can therefore mislead infrastructure planners. **Related definitions:** [Model scaling](#entry%5FKB021) · [AI agents](#entry%5FKB023) · [GPU clusters](#entry%5FKB003) · [AI accelerators](#entry%5FKB182) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-022 ### AI agents Why does AI agents matter for utilities? AI agents are systems that combine models with memory, tools, planning, and the ability to take multiple actions toward a goal. For utilities, agents may coordinate analysis or workflows, but they increase governance needs because errors can propagate through connected systems, repeated tool use can raise compute demand, and autonomous actions require clear limits and human accountability. **Related definitions:** [AI workloads](#entry%5FKB022) · [Model efficiency](#entry%5FKB024) · [Vendor governance](#entry%5FKB252) · [AI compute infrastructure](#entry%5FKB001) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-023 ### Model efficiency What infrastructure does Model efficiency require? Model efficiency is the amount of useful performance obtained from a given quantity of computation, memory, energy, time, or cost. Efficiency can improve through algorithms, quantization, sparsity, better hardware, caching, routing, and software optimization, although lower unit cost may increase total usage if cheaper AI services stimulate much greater demand. **Related definitions:** [AI agents](#entry%5FKB023) · [Open models](#entry%5FKB025) · [Moore's Law](#entry%5FKB187) · [Latency and geography](#entry%5FKB013) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-024 ### Open models How could Open models change grid operations? Open models are AI models whose weights, code, documentation, or related components are made available under terms that permit varying degrees of inspection and reuse. They can broaden access and enable local deployment, but openness does not eliminate requirements for security, licensing, data governance, evaluation, computing capacity, or accountability for downstream use. **Related definitions:** [Model efficiency](#entry%5FKB024) · [Enterprise AI adoption](#entry%5FKB026) · [Vendor governance](#entry%5FKB252) · [Edge AI infrastructure](#entry%5FKB008) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-025 ### Enterprise AI adoption What is Enterprise AI adoption? Enterprise AI adoption is the process by which an organization moves AI from experimentation into repeatable business and operating use. It requires more than model access: organizations need approved use cases, data controls, integration, evaluation, cybersecurity, workforce capability, procurement discipline, monitoring, and ownership of the outcomes produced by AI-assisted decisions. **Related definitions:** [Open models](#entry%5FKB025) · [AI safety](#entry%5FKB027) · [Utility AI adoption](#entry%5FKB071) · [Utility AI governance](#entry%5FKB241) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-026 ### AI safety How does AI safety affect energy demand? AI safety is the discipline of preventing AI systems from causing unacceptable harm through error, misuse, manipulation, loss of control, insecure design, or interaction with people and institutions. Safety practices include testing, threat modeling, access controls, human oversight, incident response, monitoring, and designing systems so failures are contained rather than amplified. **Related definitions:** [Enterprise AI adoption](#entry%5FKB026) · [AI benchmarking](#entry%5FKB028) · [Cybersecurity](#entry%5FKB245) · [Utility AI adoption](#entry%5FKB071) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-027 ### AI benchmarking Why does AI benchmarking matter for utilities? AI benchmarking is the structured comparison of models or systems using defined datasets, tasks, metrics, and operating conditions. Benchmarks can reveal relative performance, cost, speed, robustness, or energy use, but they can also distort decisions when they do not reflect real utility workflows, changing data, rare events, or the consequences of failure. **Related definitions:** [AI safety](#entry%5FKB027) · [AI demand drivers](#entry%5FKB029) · [Investor-owned utilities](#entry%5FKB061) · [HBM memory](#entry%5FKB183) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-028 ### AI demand drivers What infrastructure does AI demand drivers require? AI demand drivers are the factors that determine how much AI computation is used, including model capability, application adoption, user volume, inference frequency, software design, pricing, regulation, automation, and the economic value of outputs. Electricity demand cannot be forecast from chip shipments alone because utilization and the services built on that hardware are equally important. **Related definitions:** [AI benchmarking](#entry%5FKB028) · [AI operations](#entry%5FKB030) · [ASICs](#entry%5FKB192) · [Training vs inference](#entry%5FKB002) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-029 ### AI operations How could AI operations change grid operations? AI operations, often called AI operations management or MLOps in narrower contexts, are the processes used to deploy, monitor, update, secure, and retire AI systems in production. In grid environments, these practices must support version control, traceability, performance monitoring, human intervention, cybersecurity, and safe fallback when data or models behave unexpectedly. **Related definitions:** [AI demand drivers](#entry%5FKB029) · [Model risk management](#entry%5FKB244) · [Utility AI adoption](#entry%5FKB071) · [Vendor governance](#entry%5FKB252) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-030 SECTION 03 | 15 DEFINITIONS ## Data Centers Facility types, design choices, operating requirements, economics, and development constraints across the data-center market. **Terms in this section:** [Hyperscale data centers](#entry%5FKB031) · [Colocation](#entry%5FKB032) · [Enterprise data centers](#entry%5FKB033) · [AI-ready facilities](#entry%5FKB034) · [Rack density](#entry%5FKB035) · [Power usage effectiveness](#entry%5FKB036) · [Cooling systems](#entry%5FKB037) · [Backup power](#entry%5FKB038) · [Campus development](#entry%5FKB039) · [Site selection](#entry%5FKB040) · [Data center permitting](#entry%5FKB041) · [Data center economics](#entry%5FKB042) · [Uptime and resiliency](#entry%5FKB043) · [Modular data centers](#entry%5FKB044) · [Community-scale data centers](#entry%5FKB045) ### Hyperscale data centers What is Hyperscale data centers? A hyperscale data center is a very large facility or campus designed to support massive, standardized computing operations that can expand through repeatable modules. Hyperscale operators typically integrate servers, networks, software, power, cooling, and procurement at exceptional scale, creating electricity needs that can influence utility planning and regional infrastructure investment. **Related definitions:** [Colocation](#entry%5FKB032) · [Air cooling](#entry%5FKB196) · [Cost causation](#entry%5FKB178) · [Large-load tariffs](#entry%5FKB166) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-031 ### Colocation Why does Colocation matter for AI data centers? Colocation is a data-center business model in which customers lease space, power, cooling, and connectivity within a facility operated by another company. Colocation can aggregate many customers or host dedicated hyperscale deployments, so the facility owner may control the physical infrastructure while tenants determine a significant share of the computing load. **Related definitions:** [Hyperscale data centers](#entry%5FKB031) · [Enterprise data centers](#entry%5FKB033) · [Cloud infrastructure](#entry%5FKB007) · [High load factor customers](#entry%5FKB176) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-032 ### Enterprise data centers How does Enterprise data centers affect grid interconnection? An enterprise data center is a facility primarily built to serve the internal computing needs of a company, government, university, or other organization. It is often smaller and less standardized than hyperscale infrastructure, but its grid impact depends on the site’s total load, redundancy, growth plans, operating hours, and use of cloud or colocation services. **Related definitions:** [Colocation](#entry%5FKB032) · [AI-ready facilities](#entry%5FKB034) · [Load shape](#entry%5FKB174) · [Large-load interconnection](#entry%5FKB136) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-033 ### AI-ready facilities What are the main constraints on AI-ready facilities? An AI-ready facility is a data center designed or upgraded for dense accelerator-based computing, high-speed networking, substantial electrical distribution, and advanced cooling. Readiness requires more than available floor space: the building must support the heat, weight, power quality, redundancy, communications, commissioning, and operating practices associated with modern AI systems. **Related definitions:** [Enterprise data centers](#entry%5FKB033) · [Rack density](#entry%5FKB035) · [Power density](#entry%5FKB011) · [Cooling retrofits](#entry%5FKB206) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-034 ### Rack density How should utilities evaluate Rack density? Rack density is the electrical load and heat output concentrated in a server rack, usually measured in kilowatts. Higher rack density can reduce the floor area needed for a given amount of compute, but it increases demands on busways, cabling, cooling, structural design, fire protection, maintenance, and the facility’s ability to remove heat reliably. **Related definitions:** [AI-ready facilities](#entry%5FKB034) · [Power usage effectiveness](#entry%5FKB036) · [Power density](#entry%5FKB011) · [Liquid cooling](#entry%5FKB197) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-035 ### Power usage effectiveness What is Power usage effectiveness? Power usage effectiveness, or PUE, is the ratio of total data-center energy use to the energy consumed by information-technology equipment. A PUE of 1.0 would mean all energy reaches computing equipment, while higher values reflect cooling, power conversion, lighting, and other overhead; PUE does not measure compute productivity, water use, or carbon performance. **Related definitions:** [Rack density](#entry%5FKB035) · [Cooling systems](#entry%5FKB037) · [Water usage effectiveness](#entry%5FKB200) · [AI infrastructure economics](#entry%5FKB004) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-036 ### Cooling systems Why does Cooling systems matter for AI data centers? Cooling systems remove heat from servers and supporting equipment so components remain within safe operating limits. Data centers may use air cooling, chilled water, evaporative systems, direct-to-chip liquid cooling, immersion, or combinations of these approaches, with different implications for electricity, water, maintenance, density, climate suitability, and capital cost. **Related definitions:** [Power usage effectiveness](#entry%5FKB036) · [Backup power](#entry%5FKB038) · [Liquid cooling](#entry%5FKB197) · [Power density](#entry%5FKB011) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-037 ### Backup power How does Backup power affect grid interconnection? Backup power is the collection of batteries, generators, fuel systems, controls, and switching equipment used to maintain operations when grid service is interrupted or power quality falls outside acceptable limits. Backup systems support resilience, but their emissions, fuel supply, testing, interconnection, operating permissions, and possible grid-service roles require explicit planning. **Related definitions:** [Cooling systems](#entry%5FKB037) · [Campus development](#entry%5FKB039) · [Cooling retrofits](#entry%5FKB206) · [Customer obligations](#entry%5FKB179) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-038 ### Campus development What are the main constraints on Campus development? Campus development is the phased construction of multiple data-center buildings and shared infrastructure on a coordinated site. It can reduce unit costs and speed later expansion, but planning must account for the full expected electrical load, transmission and distribution upgrades, water, roads, permits, financing, and the risk that later phases do not materialize. **Related definitions:** [Backup power](#entry%5FKB038) · [Site selection](#entry%5FKB040) · [AI campuses](#entry%5FKB014) · [Generator interconnection](#entry%5FKB137) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-039 ### Site selection How should utilities evaluate Site selection? Data-center site selection evaluates whether a location can support the project’s power, connectivity, land, water, cooling, workforce, tax, security, permitting, and schedule requirements. A technically attractive site can still fail if grid capacity is not deliverable, community acceptance is weak, equipment lead times are long, or the commercial load commitment is uncertain. **Related definitions:** [Campus development](#entry%5FKB039) · [Data center permitting](#entry%5FKB041) · [Location strategy](#entry%5FKB005) · [Developer credibility](#entry%5FKB180) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-040 ### Data center permitting What is Data center permitting? Data-center permitting is the set of land-use, building, environmental, utility, water, air-quality, and operating approvals required to construct and run a facility. Requirements vary by jurisdiction and design, and the critical path may involve local zoning, utility facilities, backup generation, transmission upgrades, water rights, wetlands, noise, or community review. **Related definitions:** [Site selection](#entry%5FKB040) · [Data center economics](#entry%5FKB042) · [Water permitting](#entry%5FKB207) · [AI load corridors](#entry%5FKB010) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-041 ### Data center economics Why does Data center economics matter for AI data centers? Data-center economics describe how revenue, utilization, electricity, equipment, land, financing, construction, cooling, network access, and replacement cycles determine project value. Because chips can become obsolete faster than buildings and grid assets, developers must align short technology cycles with long-lived infrastructure and credible customer commitments. **Related definitions:** [Data center permitting](#entry%5FKB041) · [Uptime and resiliency](#entry%5FKB043)· [AI infrastructure economics](#entry%5FKB004) · [Load verification](#entry%5FKB144) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-042 ### Uptime and resiliency How does Uptime and resiliency affect grid interconnection? Uptime is the share of time a service remains available, while resiliency is the ability to withstand disruption and recover without unacceptable loss. Data-center designs use redundancy, backup power, diverse networks, spare capacity, and operating procedures to meet service expectations, but those choices can increase requested grid capacity and reduce apparent flexibility. **Related definitions:** [Data center economics](#entry%5FKB042) · [Modular data centers](#entry%5FKB044) · [High load factor customers](#entry%5FKB176) · [Infrastructure readiness](#entry%5FKB009) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-043 ### Modular data centers What are the main constraints on Modular data centers? A modular data center uses prefabricated or repeatable building, electrical, cooling, or computing units that can be deployed in stages. Modularity can shorten construction and match capacity to demand, but it does not remove constraints involving utility interconnection, site permits, supply chains, commissioning, network connectivity, or the long-term efficiency of many incremental additions. **Related definitions:** [Uptime and resiliency](#entry%5FKB043) · [Community-scale data centers](#entry%5FKB045) · [Edge AI infrastructure](#entry%5FKB008) · [Fast-track interconnection](#entry%5FKB142) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-044 ### Community-scale data centers How should utilities evaluate Community-scale data centers? A community-scale data center is a mid-sized facility intended to serve regional cloud, enterprise, edge, or colocation demand rather than a global hyperscale campus. The term has no single universal megawatt threshold, so its meaning should be defined by context, but it generally implies meaningful local grid impact with a smaller footprint and development scale than hyperscale projects. **Related definitions:** [Modular data centers](#entry%5FKB044) · [Edge AI infrastructure](#entry%5FKB008) · [Coincident peak](#entry%5FKB177) · [Water stress](#entry%5FKB201) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-045 SECTION 04 | 15 DEFINITIONS ## Electric Grid The bulk and local systems that generate, move, balance, and deliver electricity to increasingly concentrated digital loads. **Terms in this section:** [Bulk power system](#entry%5FKB046) · [Transmission grid](#entry%5FKB047) · [Distribution grid](#entry%5FKB048) · [Substations](#entry%5FKB049) · [Grid constraints](#entry%5FKB050) · [Grid congestion](#entry%5FKB051) · [Grid modernization](#entry%5FKB052) · [Grid reliability](#entry%5FKB053) · [Grid planning](#entry%5FKB054) · [Grid operations](#entry%5FKB055) · [Voltage stability](#entry%5FKB056) · [Frequency stability](#entry%5FKB057) · [Inverter-based resources](#entry%5FKB058) · [Grid visibility](#entry%5FKB059) · [Grid resilience](#entry%5FKB060) ### Bulk power system What is Bulk power system? The bulk power system is the interconnected generation and high-voltage transmission network used to move electricity across large areas and maintain system reliability. In North America, the regulatory definition is specific and does not include every local distribution facility, but large data-center demand can still affect bulk planning through generation needs, transmission flows, and reliability studies. **Related definitions:** [Transmission grid](#entry%5FKB047) · [Reliability entities](#entry%5FKB158) · [Transmission planning](#entry%5FKB106) · [Distribution planning](#entry%5FKB121) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-046 ### Transmission grid Why does Transmission grid matter for AI infrastructure? The transmission grid is the high-voltage network that carries large quantities of electricity between generators, regions, substations, and major customers. For AI infrastructure, transmission determines whether power is physically deliverable, not merely available in aggregate, and upgrades can require long planning, permitting, equipment, and construction timelines. **Related definitions:** [Bulk power system](#entry%5FKB046) · [Distribution grid](#entry%5FKB048) · [Distribution planning](#entry%5FKB121) · [Transmission interconnection](#entry%5FKB107) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-047 ### Distribution grid How can AI data centers affect Distribution grid? The distribution grid is the lower-voltage network of substations, feeders, transformers, protection systems, and local lines that delivers electricity to most customers. Large data centers can require new substations or dedicated feeders and may affect voltage, fault current, protection, equipment loading, power quality, and the timing of upgrades for surrounding customers. **Related definitions:** [Transmission grid](#entry%5FKB047) · [Substations](#entry%5FKB049) · [Distribution planning](#entry%5FKB121) · [Transmission bottlenecks](#entry%5FKB111) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-048 ### Substations What are the operational risks around Substations? A substation changes voltage levels, switches power flows, protects equipment, and connects transmission, distribution, generation, and customer facilities. Concentrated data-center demand can make a substation a critical point of failure, requiring careful attention to transformer capacity, redundancy, protection, physical security, spare equipment, maintenance, and future expansion. **Related definitions:** [Distribution grid](#entry%5FKB048) · [Grid constraints](#entry%5FKB050) · [Distribution reliability](#entry%5FKB130) · [Operational reliability](#entry%5FKB153) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-049 ### Grid constraints How should planners manage Grid constraints? Grid constraints are physical or operating limits that prevent electricity from moving or being used as desired. They can arise from line ratings, transformer capacity, voltage, stability, protection, generation availability, fuel supply, contingency criteria, or local distribution equipment; a region can have ample annual energy yet lack capacity at the required place and time. **Related definitions:** [Substations](#entry%5FKB049) · [Grid congestion](#entry%5FKB051) · [Transmission bottlenecks](#entry%5FKB111) · [Distribution planning](#entry%5FKB121) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-050 ### Grid congestion What is Grid congestion? Grid congestion occurs when desired power flows exceed the secure capability of transmission or other network elements. Operators then redispatch generation, curtail transactions, or use other controls, which can raise local prices and indicate a need for transmission, generation, storage, demand flexibility, or changes in where new load connects. **Related definitions:** [Grid constraints](#entry%5FKB050) · [Grid modernization](#entry%5FKB052) · [Grid-edge operations](#entry%5FKB072) · [Distributed energy resources](#entry%5FKB128) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-051 ### Grid modernization Why does Grid modernization matter for AI infrastructure? Grid modernization is the improvement of power-system equipment, data, communications, controls, planning, and operating practices. It can include advanced sensors, automation, distributed-energy integration, cybersecurity, flexible demand, digital substations, analytics, and new market designs, but modernization creates value only when technology is integrated with clear operating responsibilities and investment priorities. **Related definitions:** [Grid congestion](#entry%5FKB051) · [Grid reliability](#entry%5FKB053) · [Smart meters](#entry%5FKB131) · [Grid-enhancing technologies](#entry%5FKB115) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-052 ### Grid reliability How can AI data centers affect Grid reliability? Grid reliability is the ability of the electric system to supply demand within accepted standards under expected conditions and credible disturbances. It includes resource adequacy as well as the real-time security of voltage, frequency, equipment, and power flows; rapid large-load growth can affect both planning margins and operating behavior. **Related definitions:** [Grid modernization](#entry%5FKB052) · [Grid planning](#entry%5FKB054) · [Operational reliability](#entry%5FKB153) · [Grid-edge operations](#entry%5FKB072) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-053 ### Grid planning What are the operational risks around Grid planning? Grid planning is the forward-looking process used to identify generation, transmission, distribution, and operating needs under expected demand, resource, policy, and contingency conditions. Good planning tests multiple futures, identifies common needs, distinguishes uncertainty from commitment, and assigns costs and responsibilities before reliability problems become urgent. **Related definitions:** [Grid reliability](#entry%5FKB053) · [Grid operations](#entry%5FKB055) · [Integrated resource planning](#entry%5FKB065) · [Planning reserve margins](#entry%5FKB159) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-054 ### Grid operations How should planners manage Grid operations? Grid operations are the real-time and near-term actions used to balance supply and demand, maintain frequency and voltage, manage transmission limits, schedule outages, and respond to disturbances. Operators rely on telemetry, forecasts, models, procedures, communications, reserves, and authority to direct resources and, when necessary, curtail load. **Related definitions:** [Grid planning](#entry%5FKB054) · [Voltage stability](#entry%5FKB056) · [Grid-edge operations](#entry%5FKB072) · [Frequency response](#entry%5FKB155) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-055 ### Voltage stability What is Voltage stability? Voltage stability is the ability of a power system to maintain acceptable voltage after changes in load, generation, or network conditions. Large concentrated loads can increase reactive-power needs and sensitivity to faults or equipment outages, making local network strength, controls, compensation, dynamic models, and ride-through behavior important. **Related definitions:** [Grid operations](#entry%5FKB055) · [Frequency stability](#entry%5FKB057) · [Operational reliability](#entry%5FKB153) · [Voltage management](#entry%5FKB124) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-056 ### Frequency stability Why does Frequency stability matter for AI infrastructure? Frequency stability is the ability of the power system to keep electrical frequency within acceptable limits after an imbalance between generation and demand. Large loads affect the size and speed of disturbances; abrupt disconnection or restoration of data-center demand can therefore matter to reserve needs, controls, protection settings, and system models. **Related definitions:** [Voltage stability](#entry%5FKB056) · [Inverter-based resources](#entry%5FKB058) · [Operational reliability](#entry%5FKB153) · [Transmission bottlenecks](#entry%5FKB111) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-057 ### Inverter-based resources How can AI data centers affect Inverter-based resources? Inverter-based resources connect to the grid through power-electronic converters rather than directly coupled rotating machines. Solar, batteries, wind, and some loads use inverters, whose controls shape voltage, frequency response, fault behavior, and grid strength; growing AI load can change where these resources are needed and how system stability is assessed. **Related definitions:** [Frequency stability](#entry%5FKB057) · [Grid visibility](#entry%5FKB059) · [Frequency response](#entry%5FKB155) · [Behind-the-meter assets](#entry%5FKB132) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-058 ### Grid visibility What are the operational risks around Grid visibility? Grid visibility is the ability to observe relevant system conditions through meters, sensors, telemetry, models, communications, and data quality. Visibility allows operators and planners to understand load, power flows, equipment status, disturbances, and emerging risk; without it, flexibility and reliability claims cannot be verified in time to support decisions. **Related definitions:** [Inverter-based resources](#entry%5FKB058) · [Grid resilience](#entry%5FKB060) · [Operational reliability](#entry%5FKB153) · [Grid-edge operations](#entry%5FKB072) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-059 ### Grid resilience How should planners manage Grid resilience? Grid resilience is the ability to prepare for, withstand, adapt to, and recover from disruptive events such as extreme weather, equipment failures, cyberattacks, fuel disruptions, or supply-chain shortages. Resilience includes physical hardening, redundancy, restoration capability, communications, spare equipment, flexible operations, and institutional coordination rather than a single technology. **Related definitions:** [Grid visibility](#entry%5FKB059) · [Distribution reliability](#entry%5FKB130) · [Grid-edge operations](#entry%5FKB072) · [Distribution planning](#entry%5FKB121) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-060 SECTION 05 | 15 DEFINITIONS ## Electric Utilities How investor-owned utilities, public power, cooperatives, and regulators plan for, serve, and govern major new electricity customers. **Terms in this section:** [Investor-owned utilities](#entry%5FKB061) · [Public power](#entry%5FKB062) · [Cooperatives](#entry%5FKB063) · [Utility planning](#entry%5FKB064) · [Integrated resource planning](#entry%5FKB065) · [Cost recovery](#entry%5FKB066) · [Rate design](#entry%5FKB067) · [Customer classes](#entry%5FKB068) · [Large-load service](#entry%5FKB069) · [Utility risk management](#entry%5FKB070) · [Utility AI adoption](#entry%5FKB071) · [Grid-edge operations](#entry%5FKB072) · [Utility procurement](#entry%5FKB073) · [Utility governance](#entry%5FKB074) · [Customer protection](#entry%5FKB075) ### Investor-owned utilities What is Investor-owned utilities? An investor-owned utility is a privately owned company that provides regulated electricity service and earns an authorized return on approved investments, subject to state and sometimes federal oversight. Its incentives and obligations are shaped by rate regulation, service territories, cost recovery, performance requirements, capital markets, and duties to serve customers reliably and fairly. **Related definitions:** [Public power](#entry%5FKB062) · [Regulatory compact](#entry%5FKB240) · [Retail choice](#entry%5FKB085) · [Vendor governance](#entry%5FKB252) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-061 ### Public power Why does Public power matter for AI load growth? Public power utilities are owned by municipalities, states, public authorities, or other governmental entities and serve customers without private shareholders. Their governance, financing, tax status, resource choices, and economic-development roles can differ from investor-owned utilities, affecting how they evaluate large AI loads, allocate risk, and make infrastructure commitments. **Related definitions:** [Investor-owned utilities](#entry%5FKB061) · [Cooperatives](#entry%5FKB063) · [State public utility commissions](#entry%5FKB227) · [Wholesale electricity markets](#entry%5FKB076) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-062 ### Cooperatives How should utilities approach Cooperatives? Electric cooperatives are member-owned utilities, often serving rural or less densely populated areas. They may purchase power from generation-and-transmission cooperatives or wholesale suppliers and can face distinctive capital, geographic, and load-concentration challenges when a single data center is large relative to the cooperative’s existing system. **Related definitions:** [Public power](#entry%5FKB062) · [Utility planning](#entry%5FKB064) · [State public utility commissions](#entry%5FKB227) · [Resource adequacy](#entry%5FKB082) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-063 ### Utility planning What risks does Utility planning create for customers? Utility planning is the process of forecasting customer needs and selecting investments, contracts, programs, and operating changes to provide safe, reliable, and affordable service. When large-load forecasts are uncertain, planning must test the consequences of both underbuilding and overbuilding and protect existing customers from unsupported commitments. **Related definitions:** [Cooperatives](#entry%5FKB063) · [Integrated resource planning](#entry%5FKB065) · [Customer obligations](#entry%5FKB179)· [State public utility commissions](#entry%5FKB227) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-064 ### Integrated resource planning How can regulators evaluate Integrated resource planning? Integrated resource planning is a structured process for comparing demand-side and supply-side options over a long horizon under multiple assumptions. Regulators use it to examine load forecasts, resource adequacy, costs, risks, emissions, transmission needs, and customer impacts, although the exact legal requirements and methods vary by jurisdiction. **Related definitions:** [Utility planning](#entry%5FKB064) · [Cost recovery](#entry%5FKB066) · [Energy markets](#entry%5FKB078) · [State public utility commissions](#entry%5FKB227) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-065 ### Cost recovery What is Cost recovery? Cost recovery is the process through which a utility collects approved expenses and investment returns from customers, usually through rates or specific charges. For large loads, the central question is whether facilities and contracts are paid for by the customer causing the need, shared with other customers, or exposed to stranded-cost risk if the project changes. **Related definitions:** [Integrated resource planning](#entry%5FKB065) · [Rate design](#entry%5FKB067) · [Cost causation](#entry%5FKB178) · [Cost allocation policy](#entry%5FKB231) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-066 ### Rate design Why does Rate design matter for AI load growth? Rate design is the structure of prices and charges used to recover utility costs and influence customer behavior. Large-load rates may include demand charges, energy charges, minimum bills, contract terms, deposits, contribution requirements, curtailment options, and exit provisions intended to reflect cost causation and protect other customers. **Related definitions:** [Cost recovery](#entry%5FKB066) · [Customer classes](#entry%5FKB068) · [Large-load tariffs](#entry%5FKB166) · [Tariff design](#entry%5FKB239) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-067 ### Customer classes How should utilities approach Customer classes? Customer classes group utility customers with similar service characteristics for pricing and regulatory purposes, such as residential, commercial, industrial, or large-load categories. Classification affects which costs are shared, how demand is measured, what service obligations apply, and whether a new type of customer requires a distinct tariff. **Related definitions:** [Rate design](#entry%5FKB067) · [Large-load service](#entry%5FKB069) · [Tariff design](#entry%5FKB239) · [Minimum demand charges](#entry%5FKB169) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-068 ### Large-load service What risks does Large-load service create for customers? Large-load service is the set of utility facilities, contracts, rates, studies, operating terms, and customer obligations used to serve a major new electricity demand. Because a single project can require generation and network investments comparable to those of many existing customers, service terms must address schedule, credit, load ramp, curtailment, and stranded-cost exposure. **Related definitions:** [Customer classes](#entry%5FKB068) · [Utility risk management](#entry%5FKB070) · [Large-load tariffs](#entry%5FKB166) · [Tariff design](#entry%5FKB239) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-069 ### Utility risk management How can regulators evaluate Utility risk management? Utility risk management is the identification, measurement, allocation, mitigation, and monitoring of risks that could affect reliability, affordability, safety, compliance, or financial performance. Large AI loads introduce forecast, counterparty, construction, technology, concentration, regulatory, and operational risks that should be managed through both contracts and system planning. **Related definitions:** [Large-load service](#entry%5FKB069) · [Utility AI adoption](#entry%5FKB071) · [Ratepayer protection](#entry%5FKB232) · [Price volatility](#entry%5FKB089) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-070 ### Utility AI adoption What is Utility AI adoption? Utility AI adoption is the controlled use of artificial intelligence in functions such as forecasting, asset management, customer service, engineering, planning, cybersecurity, and operations. Adoption requires validated use cases, governed data, technical integration, human oversight, monitoring, incident response, and evidence that the system improves decisions without introducing unacceptable risk. **Related definitions:** [Utility risk management](#entry%5FKB070) · [Grid-edge operations](#entry%5FKB072) · [Utility AI governance](#entry%5FKB241) · [NERC oversight](#entry%5FKB228) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-071 ### Grid-edge operations Why does Grid-edge operations matter for AI load growth? Grid-edge operations manage devices and conditions near customers and the distribution system, including meters, distributed energy resources, flexible loads, storage, electric vehicles, and local controls. AI load growth can make the grid edge more consequential by increasing the value of precise forecasts, automated coordination, demand flexibility, and real-time visibility. **Related definitions:** [Utility AI adoption](#entry%5FKB071) · [Utility procurement](#entry%5FKB073) · [Energy markets](#entry%5FKB078) · [Load ramping](#entry%5FKB175) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-072 ### Utility procurement How should utilities approach Utility procurement? Utility procurement is the process of acquiring equipment, construction, software, services, energy, and capacity under rules intended to balance cost, competition, quality, delivery, and accountability. Rapid infrastructure growth puts pressure on procurement because long-lead equipment and specialized expertise may need earlier commitments than traditional project cycles allow. **Related definitions:** [Grid-edge operations](#entry%5FKB072) · [Utility governance](#entry%5FKB074) · [Transmission reform](#entry%5FKB237) · [Market power](#entry%5FKB083) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-073 ### Utility governance What risks does Utility governance create for customers? Utility governance is the system of boards, executives, regulators, policies, controls, incentives, and accountability through which utility decisions are made. Strong governance clarifies who can commit capital, approve risk, accept forecasts, manage conflicts, oversee AI, and protect customers when economic-development pressure and infrastructure uncertainty collide. **Related definitions:** [Utility procurement](#entry%5FKB073) · [Customer protection](#entry%5FKB075) · [Utility AI governance](#entry%5FKB241) · [Data center policy](#entry%5FKB238) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-074 ### Customer protection How can regulators evaluate Customer protection? Customer protection is the set of regulatory, contractual, financial, and service safeguards designed to prevent unreasonable costs, discrimination, service degradation, or unfair risk transfer. For large loads, protection includes transparent rates, cost-causation principles, financial assurance, performance obligations, complaint processes, and clear treatment of stranded assets. **Related definitions:** [Utility governance](#entry%5FKB074) · [Ratepayer protection](#entry%5FKB232) · [Cost causation](#entry%5FKB178) · [Tariff design](#entry%5FKB239) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-075 SECTION 06 | 15 DEFINITIONS ## Power Markets How large AI-related loads affect energy prices, capacity, congestion, ancillary services, resource adequacy, and market risk. **Terms in this section:** [Wholesale electricity markets](#entry%5FKB076) · [Capacity markets](#entry%5FKB077) · [Energy markets](#entry%5FKB078) · [Ancillary services](#entry%5FKB079) · [Congestion pricing](#entry%5FKB080) · [Locational marginal pricing](#entry%5FKB081) · [Resource adequacy](#entry%5FKB082) · [Market power](#entry%5FKB083) · [Demand response](#entry%5FKB084) · [Retail choice](#entry%5FKB085) · [Power purchase agreements](#entry%5FKB086) · [Hedging](#entry%5FKB087) · [Market monitoring](#entry%5FKB088) · [Price volatility](#entry%5FKB089) · [Scarcity pricing](#entry%5FKB090) ### Wholesale electricity markets What is Wholesale electricity markets? Wholesale electricity markets are organized arrangements through which generators, suppliers, utilities, traders, and large customers buy and sell energy, capacity, and reliability services. In the United States, regional market designs differ, but prices generally reflect system conditions, transmission constraints, resource offers, operating rules, and the need to balance supply and demand. **Related definitions:** [Capacity markets](#entry%5FKB077) · [Planning reserve margins](#entry%5FKB159) · [Peaking resources](#entry%5FKB100) · [Reliability-must-run resources](#entry%5FKB163) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-076 ### Capacity markets How can AI data centers affect Capacity markets? Capacity markets compensate resources for being available to meet future peak demand and reliability requirements, rather than paying only for energy produced. Large data-center forecasts can increase procurement needs and prices, but speculative or overstated demand can also cause consumers to pay for capacity that is not ultimately needed. **Related definitions:** [Wholesale electricity markets](#entry%5FKB076) · [Energy markets](#entry%5FKB078) · [High load factor customers](#entry%5FKB176) · [Resource adequacy](#entry%5FKB152) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-077 ### Energy markets Why does Energy markets matter for electricity prices? Energy markets schedule and price electricity over day-ahead and real-time intervals. Prices vary with fuel costs, weather, outages, demand, transmission constraints, and resource availability; large, high-load-factor data centers can increase local energy consumption and change which generators operate, especially in constrained regions. **Related definitions:** [Capacity markets](#entry%5FKB077) · [Ancillary services](#entry%5FKB079) · [High load factor customers](#entry%5FKB176) · [Integrated resource planning](#entry%5FKB065) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-078 ### Ancillary services What risks does Ancillary services create for market participants? Ancillary services are reliability products used to maintain frequency, reserves, voltage support, and other operating needs beyond the delivery of energy. Market participants face performance, telemetry, availability, settlement, and penalty risks, and new large loads may affect both the quantity of services required and opportunities for flexible demand to provide them. **Related definitions:** [Energy markets](#entry%5FKB078) · [Congestion pricing](#entry%5FKB080) · [Frequency response](#entry%5FKB155) · [Utility planning](#entry%5FKB064) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-079 ### Congestion pricing How should investors interpret Congestion pricing? Congestion pricing reflects the cost of serving load when transmission constraints prevent the least-cost generation from reaching a location. Persistent price differences can signal local scarcity or transmission limitations, but investors must distinguish temporary operating conditions from durable structural constraints and account for changes in generation, load, transmission, and market rules. **Related definitions:** [Ancillary services](#entry%5FKB079) · [Locational marginal pricing](#entry%5FKB081) · [Co-located generation](#entry%5FKB097) · [Reliability-must-run resources](#entry%5FKB163) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-080 ### Locational marginal pricing What is Locational marginal pricing? Locational marginal pricing is a method used in several organized U.S. markets to calculate the incremental cost of supplying one additional unit of electricity at a specific location. The price generally reflects energy, transmission congestion, and electrical losses, making it a useful signal of where network constraints and generation patterns affect the cost of serving load. **Related definitions:** [Congestion pricing](#entry%5FKB080) · [Resource adequacy](#entry%5FKB082) · [Co-located generation](#entry%5FKB097) · [Large-load service](#entry%5FKB069) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-081 ### Resource adequacy How can AI data centers affect Resource adequacy? Resource adequacy is the ability of the power system to have enough dependable capacity and energy to meet expected demand under defined reliability criteria. Large data-center demand affects adequacy through its magnitude, load shape, timing, flexibility, and forecast credibility, as well as through the generation and transmission needed to serve it. **Related definitions:** [Locational marginal pricing](#entry%5FKB081) · [Market power](#entry%5FKB083) · [Resource adequacy](#entry%5FKB152) · [High load factor customers](#entry%5FKB176) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-082 ### Market power Why does Market power matter for electricity prices? Market power is the ability of a seller or buyer to influence price or market outcomes because competition is limited or alternatives are constrained. Rapid load growth, local transmission limits, concentrated generation ownership, and scarce capacity can increase market-power concerns, requiring monitoring, mitigation rules, and careful interpretation of prices. **Related definitions:** [Resource adequacy](#entry%5FKB082) · [Demand response](#entry%5FKB084) · [Reliability-must-run resources](#entry%5FKB163) · [Utility procurement](#entry%5FKB073) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-083 ### Demand response What risks does Demand response create for market participants? Demand response is a change in electricity use in response to prices, incentives, or operating instructions. Its value depends on measurable baselines, telemetry, response speed, duration, notice, rebound effects, and customer performance; unverified or unavailable flexibility can create reliability and settlement risks for market participants. **Related definitions:** [Market power](#entry%5FKB083) · [Retail choice](#entry%5FKB085) · [Flexible loads](#entry%5FKB171) · [Frequency response](#entry%5FKB155) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-084 ### Retail choice How should investors interpret Retail choice? Retail choice allows eligible customers in some jurisdictions to select a competitive electricity supplier while the local utility continues to deliver power. Investors must separate the retail energy contract from regulated delivery, capacity, transmission, reliability, interconnection, and infrastructure obligations that may remain with utilities or system operators. **Related definitions:** [Demand response](#entry%5FKB084) · [Power purchase agreements](#entry%5FKB086) · [Nuclear power](#entry%5FKB092) · [Reliability entities](#entry%5FKB158) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-085 ### Power purchase agreements What is Power purchase agreements? A power purchase agreement is a contract under which a buyer agrees to purchase electricity, environmental attributes, capacity, or related products from a generator over specified terms. Physical, financial, sleeved, and virtual structures allocate price, volume, basis, credit, delivery, curtailment, and development risks differently and do not automatically guarantee local deliverability. **Related definitions:** [Retail choice](#entry%5FKB085) · [Hedging](#entry%5FKB087) · [Curtailable service](#entry%5FKB173) · [Contingency analysis](#entry%5FKB161) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-086 ### Hedging How can AI data centers affect Hedging? Hedging uses contracts or financial instruments to reduce exposure to uncertain electricity, fuel, capacity, congestion, or basis prices. Data-center growth can change the size and location of exposures, but hedges introduce counterparty, collateral, liquidity, volume, shape, and mismatch risks when actual consumption differs from contracted assumptions. **Related definitions:** [Power purchase agreements](#entry%5FKB086) · [Market monitoring](#entry%5FKB088) · [Utility risk management](#entry%5FKB070) · [Minimum demand charges](#entry%5FKB169) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-087 ### Market monitoring Why does Market monitoring matter for electricity prices? Market monitoring is the independent or regulatory oversight of electricity-market behavior, prices, offers, rules, and structural conditions. Monitors identify manipulation, market-power concerns, inefficient design, and unintended incentives, providing essential context when rapid load growth and infrastructure constraints produce unusual price outcomes. **Related definitions:** [Hedging](#entry%5FKB087) · [Price volatility](#entry%5FKB089) · [Rate design](#entry%5FKB067) · [Operational reliability](#entry%5FKB153) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-088 ### Price volatility What risks does Price volatility create for market participants? Price volatility is the degree to which electricity or related market prices change over time. It can arise from weather, fuel costs, outages, transmission constraints, variable generation, forecast errors, or scarcity; market participants manage it through contracts, operational flexibility, diversification, reserves, and financial hedges. **Related definitions:** [Market monitoring](#entry%5FKB088) · [Scarcity pricing](#entry%5FKB090) · [Utility risk management](#entry%5FKB070) · [Resource adequacy](#entry%5FKB152) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-089 ### Scarcity pricing How should investors interpret Scarcity pricing? Scarcity pricing raises energy or reserve prices when the system approaches shortage conditions, signaling the value of reliability and encouraging availability or demand reduction. Investors should examine the trigger rules, caps, frequency, settlement exposure, resource performance, and whether scarcity events reflect genuine system need or recurring structural weakness. **Related definitions:** [Price volatility](#entry%5FKB089) · [Non-firm service](#entry%5FKB172) · [Peaking resources](#entry%5FKB100) · [Frequency response](#entry%5FKB155) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-090 SECTION 07 | 15 DEFINITIONS ## Generation The power-supply options that can serve AI infrastructure, including gas, nuclear, renewables, onsite generation, and clean firm resources. **Terms in this section:** [Natural gas generation](#entry%5FKB091) · [Nuclear power](#entry%5FKB092) · [Renewables](#entry%5FKB093) · [Hydropower](#entry%5FKB094) · [Diesel backup](#entry%5FKB095) · [On-site generation](#entry%5FKB096) · [Co-located generation](#entry%5FKB097) · [Firm power](#entry%5FKB098) · [Clean firm power](#entry%5FKB099) · [Peaking resources](#entry%5FKB100) · [Combined-cycle plants](#entry%5FKB101) · [Small modular reactors](#entry%5FKB102) · [Fuel availability](#entry%5FKB103) · [Generation interconnection](#entry%5FKB104) · [Generation siting](#entry%5FKB105) ### Natural gas generation What is Natural gas generation? Natural gas generation produces electricity by burning natural gas in combustion turbines, reciprocating engines, or combined-cycle plants. It can provide dispatchable output and relatively fast development compared with some alternatives, but projects depend on turbine availability, pipeline capacity, fuel price, air permits, emissions policy, water, interconnection, and the ability to operate during stressed conditions. **Related definitions:** [Nuclear power](#entry%5FKB092) · [Additionality](#entry%5FKB263) · [Co-located storage](#entry%5FKB219) · [Market power](#entry%5FKB083) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-091 ### Nuclear power Can Nuclear power support AI data centers? Nuclear power can support data centers by providing large quantities of low-carbon, high-capacity-factor electricity. Existing plants may offer dependable supply, while new reactors face substantial licensing, financing, construction, fuel-cycle, workforce, and schedule requirements; a contractual association with nuclear generation does not by itself remove grid-delivery and reliability obligations. **Related definitions:** [Natural gas generation](#entry%5FKB091) · [Renewables](#entry%5FKB093) · [Retail choice](#entry%5FKB085) · [Grid-enhancing technologies](#entry%5FKB115) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-092 ### Renewables What are the limits of Renewables? Renewable resources such as wind and solar can provide low-operating-cost, low-carbon electricity, but their output varies with weather and time. Their ability to serve data centers depends on portfolio diversity, transmission, storage, flexible demand, firming resources, market purchases, and whether the objective is annual energy matching, hourly matching, cost control, or physical reliability. **Related definitions:** [Nuclear power](#entry%5FKB092) · [Hydropower](#entry%5FKB094) · [24/7 clean energy](#entry%5FKB257) · [Hybrid resources](#entry%5FKB225) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-093 ### Hydropower How does Hydropower affect reliability? Hydropower generates electricity from flowing or stored water and can provide energy, capacity, reserves, inertia, and flexible operation depending on the facility. Reliability value varies with hydrology, environmental constraints, reservoir rules, competing water uses, transmission access, equipment condition, and the operational limits of each plant. **Related definitions:** [Renewables](#entry%5FKB093) · [Diesel backup](#entry%5FKB095) · [Grid services](#entry%5FKB217) · [Wholesale electricity markets](#entry%5FKB076) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-094 ### Diesel backup What should developers consider before relying on Diesel backup? Diesel backup generators can provide dependable onsite power during outages, but developers must consider fuel storage and delivery, air emissions, noise, maintenance, testing, startup performance, permitting, extreme-weather resilience, and limits on operating hours. Backup generation should not be treated as unlimited substitute capacity without explicit operating and regulatory analysis. **Related definitions:** [Hydropower](#entry%5FKB094) · [On-site generation](#entry%5FKB096) · [Storage interconnection](#entry%5FKB220) · [Emissions impact](#entry%5FKB259) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-095 ### On-site generation What is On-site generation? Onsite generation is electricity produced at or near the customer’s facility, using resources such as gas engines, turbines, fuel cells, solar, wind, or combined heat and power. It can reduce grid purchases or improve resilience, but interconnection, fuel, emissions, economics, maintenance, islanding, and utility standby obligations remain important. **Related definitions:** [Diesel backup](#entry%5FKB095) · [Co-located generation](#entry%5FKB097) · [Demand charge management](#entry%5FKB218) · [Retail choice](#entry%5FKB085) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-096 ### Co-located generation Can Co-located generation support AI data centers? Co-located generation is a generating resource located near or electrically associated with a major load. It may reduce some network use or create contracting advantages, but the actual grid impact depends on electrical configuration, dispatch, outages, surplus exports, backup service, transmission rights, reliability responsibilities, and applicable market and tariff rules. **Related definitions:** [On-site generation](#entry%5FKB096) · [Firm power](#entry%5FKB098) · [Co-located storage](#entry%5FKB219) · [Transmission planning](#entry%5FKB106) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-097 ### Firm power What are the limits of Firm power? Firm power is electricity service backed by resources, contracts, network capability, and operating arrangements intended to remain available through defined conditions. Firmness is not absolute: it depends on the governing tariff or contract, fuel and equipment availability, transmission security, contingency assumptions, and any permitted interruption or force-majeure provisions. **Related definitions:** [Co-located generation](#entry%5FKB097) · [Clean firm power](#entry%5FKB099) · [Resilience value](#entry%5FKB223) · [Clean power procurement](#entry%5FKB262) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-098 ### Clean firm power How does Clean firm power affect reliability? Clean firm power refers to low-carbon electricity resources capable of supplying dependable output when needed, rather than only when weather conditions permit. The category can include nuclear, geothermal, hydropower, fossil generation with effective carbon capture, or other technologies, but cost, scalability, fuel, siting, maturity, and environmental impacts differ substantially. **Related definitions:** [Firm power](#entry%5FKB098) · [Peaking resources](#entry%5FKB100) · [Community impacts](#entry%5FKB268) · [Capacity markets](#entry%5FKB077) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-099 ### Peaking resources What should developers consider before relying on Peaking resources? Peaking resources are generators or demand-side resources intended to operate during periods of high demand, scarcity, or system stress. They may run relatively few hours but provide important capacity and reserves; developers must assess fuel, emissions, startup speed, performance during extreme conditions, market revenue, and whether the resource remains economic under changing policy. **Related definitions:** [Clean firm power](#entry%5FKB099) · [Combined-cycle plants](#entry%5FKB101) · [Hybrid resources](#entry%5FKB225) · [Wholesale electricity markets](#entry%5FKB076) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-100 ### Combined-cycle plants What is Combined-cycle plants? A combined-cycle power plant uses both a gas turbine and a steam turbine, capturing hot exhaust from the gas turbine to produce additional electricity. This improves efficiency relative to a simple-cycle plant, but combined-cycle facilities generally require more complex equipment, cooling, water, construction, and startup arrangements. **Related definitions:** [Peaking resources](#entry%5FKB100) · [Small modular reactors](#entry%5FKB102) · [Thermal storage](#entry%5FKB213) · [Emissions impact](#entry%5FKB259) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-101 ### Small modular reactors Can Small modular reactors support AI data centers? Small modular reactors are nuclear reactors designed for lower unit output and greater use of standardized or factory-produced components than traditional large plants. They could provide firm low-carbon electricity, but commercial deployment depends on licensing, design maturity, manufacturing, fuel, financing, waste management, site preparation, workforce, and demonstrated construction performance. **Related definitions:** [Combined-cycle plants](#entry%5FKB101) · [Fuel availability](#entry%5FKB103) · [Grid services](#entry%5FKB217) · [Transformer lead times](#entry%5FKB112) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-102 ### Fuel availability What are the limits of Fuel availability? Fuel availability is the ability to obtain and deliver the energy input required for generation when it is needed. Reliability depends not only on nominal supply but also on pipelines, storage, transportation, contracts, weather, competing demand, inventories, infrastructure failures, and the fuel-assurance arrangements of individual plants. **Related definitions:** [Small modular reactors](#entry%5FKB102) · [Generation interconnection](#entry%5FKB104) · [Capacity markets](#entry%5FKB077) · [Co-located storage](#entry%5FKB219) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-103 ### Generation interconnection How does Generation interconnection affect reliability? Generation interconnection is the technical and contractual process for connecting a new power plant or storage resource to the grid. Studies determine network upgrades, protection, stability, deliverability, cost responsibility, and operating requirements; delays can affect data-center supply strategies even when the customer and generator are commercially linked. **Related definitions:** [Fuel availability](#entry%5FKB103) · [Generation siting](#entry%5FKB105) · [Transmission interconnection](#entry%5FKB107) · [Storage interconnection](#entry%5FKB220) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-104 ### Generation siting What should developers consider before relying on Generation siting? Generation siting is the selection and approval of a location for a power project based on fuel, transmission, land, water, environmental impacts, community acceptance, construction, safety, and economics. A site near a data center may appear attractive but still face network constraints, permitting delays, fuel limitations, or incompatibility with regional planning. **Related definitions:** [Generation interconnection](#entry%5FKB104) · [Community impacts](#entry%5FKB268) · [Competitive transmission](#entry%5FKB108) · [Congestion pricing](#entry%5FKB080) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-105 SECTION 08 | 15 DEFINITIONS ## Transmission The high-voltage networks, planning processes, equipment constraints, and cost-allocation rules that determine speed to power. **Terms in this section:** [Transmission planning](#entry%5FKB106) · [Transmission interconnection](#entry%5FKB107) · [Competitive transmission](#entry%5FKB108) · [Transmission cost allocation](#entry%5FKB109) · [High-voltage lines](#entry%5FKB110) · [Transmission bottlenecks](#entry%5FKB111) · [Transformer lead times](#entry%5FKB112) · [Right-of-way](#entry%5FKB113) · [Dynamic line ratings](#entry%5FKB114) · [Grid-enhancing technologies](#entry%5FKB115) · [Transmission tariffs](#entry%5FKB116) · [Regional planning](#entry%5FKB117) · [Reliability upgrades](#entry%5FKB118) · [Transmission queues](#entry%5FKB119) · [Critical path management](#entry%5FKB120) ### Transmission planning What is Transmission planning? Transmission planning identifies additions or changes needed in the high-voltage network to maintain reliability, reduce congestion, connect resources, serve load, and support policy goals. It uses forecasts, power-flow and stability studies, contingency analysis, scenarios, benefit measures, and cost-allocation rules over time horizons that can extend well beyond individual projects. **Related definitions:** [Transmission interconnection](#entry%5FKB107) · [Bulk power system](#entry%5FKB046) · [Network upgrades](#entry%5FKB143) · [Grid reliability](#entry%5FKB053) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-106 ### Transmission interconnection Why does Transmission interconnection matter for speed to power? Transmission interconnection is the process of connecting a customer, generator, or network facility to the high-voltage grid under applicable technical studies, tariffs, agreements, and operating requirements. It matters for speed to power because major upgrades, transformers, rights-of-way, and regional approvals often take longer than the data-center buildings they serve. **Related definitions:** [Transmission planning](#entry%5FKB106) · [Competitive transmission](#entry%5FKB108) · [Large-load interconnection](#entry%5FKB136) · [Transmission grid](#entry%5FKB047) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-107 ### Competitive transmission How can AI load growth affect Competitive transmission? Competitive transmission refers to transmission projects selected or developed through competitive processes rather than assigned solely to incumbent utilities. Competition may affect cost, innovation, financing, and schedule, but successful projects still require regional need determination, siting authority, land rights, interconnection, construction capability, and long-term operational accountability. **Related definitions:** [Transmission interconnection](#entry%5FKB107) · [Transmission cost allocation](#entry%5FKB109) · [Generator interconnection](#entry%5FKB137) · [Dynamic load models](#entry%5FKB160) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-108 ### Transmission cost allocation What bottlenecks limit Transmission cost allocation? Transmission cost allocation determines which customers, utilities, regions, or market participants pay for transmission investments. Methods may reflect reliability need, usage, economic benefits, public policy, negotiated arrangements, or direct assignment; disputes arise because costs are visible immediately while benefits are distributed across locations and decades. **Related definitions:** [Competitive transmission](#entry%5FKB108) · [High-voltage lines](#entry%5FKB110) · [Network upgrades](#entry%5FKB143) · [Stranded costs](#entry%5FKB283) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-109 ### High-voltage lines How should regulators evaluate High-voltage lines? High-voltage lines carry large amounts of electricity over long distances at voltages that reduce losses and conductor requirements. Regulators evaluate their need, routing, cost, reliability benefits, environmental effects, land impacts, alternatives, and allocation, while recognizing that permitting and construction can take many years. **Related definitions:** [Transmission cost allocation](#entry%5FKB109) · [Transmission bottlenecks](#entry%5FKB111) · [Network upgrades](#entry%5FKB143) · [Transmission grid](#entry%5FKB047) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-110 ### Transmission bottlenecks What is Transmission bottlenecks? Transmission bottlenecks are network limitations that restrict the movement of electricity between areas. They can result from line ratings, voltage or stability limits, transformer capacity, contingency requirements, or delayed expansion, causing congestion, local price differences, generation curtailment, and reduced ability to serve new large loads. **Related definitions:** [High-voltage lines](#entry%5FKB110) · [Transformer lead times](#entry%5FKB112) · [Grid constraints](#entry%5FKB050) · [Service agreements](#entry%5FKB146) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-111 ### Transformer lead times Why does Transformer lead times matter for speed to power? Transformer lead time is the period required to specify, manufacture, test, transport, install, and commission a transformer. Large power transformers are specialized, capital-intensive, and logistically difficult, so limited manufacturing capacity, material shortages, design changes, factory slots, transportation, and field construction can make them a critical schedule constraint. **Related definitions:** [Transmission bottlenecks](#entry%5FKB111) · [Right-of-way](#entry%5FKB113) · [Equipment lead times](#entry%5FKB279) · [Substations](#entry%5FKB049) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-112 ### Right-of-way How can AI load growth affect Right-of-way? A transmission right-of-way is the land corridor used for lines, towers, access, vegetation management, and maintenance. Securing it may require easements, public approvals, environmental review, negotiation, or eminent-domain authority; route opposition and fragmented ownership can make land control a major determinant of transmission schedule. **Related definitions:** [Transformer lead times](#entry%5FKB112) · [Dynamic line ratings](#entry%5FKB114) · [Load verification](#entry%5FKB144) · [Grid operations](#entry%5FKB055) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-113 ### Dynamic line ratings What bottlenecks limit Dynamic line ratings? Dynamic line ratings adjust a transmission line’s allowable loading using real-time or forecast conditions such as temperature, wind, solar heating, and conductor sag. They can reveal additional capacity when conditions are favorable, but require sensors, communications, forecasts, operating procedures, and confidence that ratings remain secure under changing conditions. **Related definitions:** [Right-of-way](#entry%5FKB113) · [Grid-enhancing technologies](#entry%5FKB115) · [Grid constraints](#entry%5FKB050) · [Adequacy vs security](#entry%5FKB164) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-114 ### Grid-enhancing technologies How should regulators evaluate Grid-enhancing technologies? Grid-enhancing technologies are hardware and software that increase the capacity, controllability, or visibility of the existing transmission system. Examples include dynamic line ratings, advanced power-flow control, topology optimization, and improved sensors; they can complement new lines but do not eliminate every thermal, voltage, stability, or reliability constraint. **Related definitions:** [Dynamic line ratings](#entry%5FKB114) · [Transmission tariffs](#entry%5FKB116) · [Bulk power system](#entry%5FKB046) · [Reliability entities](#entry%5FKB158) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-115 ### Transmission tariffs What is Transmission tariffs? Transmission tariffs are the filed rules, rates, procedures, and service terms governing access to transmission systems. They address services, planning, interconnection, cost allocation, scheduling, congestion, losses, credit, and settlements, and are subject to federal or other jurisdictional oversight depending on the system. **Related definitions:** [Grid-enhancing technologies](#entry%5FKB115) · [Regional planning](#entry%5FKB117) · [Service agreements](#entry%5FKB146) · [Utility balance sheets](#entry%5FKB282) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-116 ### Regional planning Why does Regional planning matter for speed to power? Regional transmission planning evaluates network needs and solutions across multiple utilities or states rather than treating each service territory independently. It can identify shared reliability, economic, and policy benefits, but requires common assumptions, transparent models, benefit measures, governance, cost allocation, and coordination across jurisdictions. **Related definitions:** [Transmission tariffs](#entry%5FKB116) · [Reliability upgrades](#entry%5FKB118) · [Network upgrades](#entry%5FKB143) · [Bulk power system](#entry%5FKB046) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-117 ### Reliability upgrades How can AI load growth affect Reliability upgrades? Reliability upgrades are transmission or related facilities required to keep the system within applicable reliability criteria after a new generator, load, or network change. The label indicates the purpose of the upgrade, but cost responsibility, timing, scope, alternatives, and whether broader regional benefits exist remain important questions. **Related definitions:** [Regional planning](#entry%5FKB117) · [Transmission queues](#entry%5FKB119) · [Network upgrades](#entry%5FKB143) · [Reliability-must-run resources](#entry%5FKB163) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-118 ### Transmission queues What bottlenecks limit Transmission queues? A transmission queue is the ordered or managed pipeline of requests seeking interconnection or transmission service. Queues help coordinate studies, but they can become congested when speculative projects, repeated restudies, shared upgrades, and limited staff obscure which proposals are commercially ready and which network needs should be planned more broadly. **Related definitions:** [Reliability upgrades](#entry%5FKB118) · [Critical path management](#entry%5FKB120) · [Queue reform](#entry%5FKB138) · [Bulk power system](#entry%5FKB046) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-119 ### Critical path management How should regulators evaluate Critical path management? Critical path management identifies the sequence of tasks that determines the earliest possible project completion date. For transmission and large-load service, the critical path may include studies, land, permits, equipment, financing, outage windows, construction, testing, and customer milestones, and it can shift as risks materialize. **Related definitions:** [Transmission queues](#entry%5FKB119) · [Generator interconnection](#entry%5FKB137) · [Load shedding](#entry%5FKB157) · [Transmission grid](#entry%5FKB047) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-120 SECTION 09 | 15 DEFINITIONS ## Distribution The substations, feeders, service equipment, load forecasts, and local operating systems that connect facilities below the transmission grid. **Terms in this section:** [Distribution planning](#entry%5FKB121) · [Feeder capacity](#entry%5FKB122) · [Distribution automation](#entry%5FKB123) · [Voltage management](#entry%5FKB124) · [Service transformers](#entry%5FKB125) · [Load forecasting](#entry%5FKB126) · [Flexible interconnection](#entry%5FKB127) · [Distributed energy resources](#entry%5FKB128) · [Circuit hosting capacity](#entry%5FKB129) · [Distribution reliability](#entry%5FKB130) · [Smart meters](#entry%5FKB131) · [Behind-the-meter assets](#entry%5FKB132) · [Microgrids](#entry%5FKB133) · [Load management](#entry%5FKB134) · [Local permitting](#entry%5FKB135) ### Distribution planning What is Distribution planning? Distribution planning forecasts local electricity needs and identifies substations, feeders, voltage-control equipment, protection, automation, and operating changes required to serve customers reliably. It increasingly incorporates distributed resources, flexible demand, electrification, extreme weather, data quality, and the possibility that a single large customer can transform a local network. **Related definitions:** [Feeder capacity](#entry%5FKB122) · [Distribution grid](#entry%5FKB048) · [Utility planning](#entry%5FKB064) · [Bulk power system](#entry%5FKB046) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-121 ### Feeder capacity How does Feeder capacity affect AI data center service? Feeder capacity is the amount of load or generation a distribution circuit can accommodate while meeting thermal, voltage, protection, and reliability requirements. Available capacity varies by location, time, contingency, and planned upgrades, so a simple nameplate rating does not establish whether a data center can be served. **Related definitions:** [Distribution planning](#entry%5FKB121) · [Distribution automation](#entry%5FKB123) · [High load factor customers](#entry%5FKB176) · [Grid constraints](#entry%5FKB050) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-122 ### Distribution automation Why does Distribution automation matter below the transmission system? Distribution automation uses sensors, communications, switches, controls, and software to monitor and operate local networks. It can isolate faults, restore service, manage voltage, coordinate distributed resources, and improve visibility, but automation requires cybersecurity, communications resilience, validated logic, maintenance, and clear operator authority. **Related definitions:** [Feeder capacity](#entry%5FKB122) · [Voltage management](#entry%5FKB124) · [Grid modernization](#entry%5FKB052) · [Grid-edge operations](#entry%5FKB072) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-123 ### Voltage management What data are needed to evaluate Voltage management? Voltage management keeps customer and equipment voltage within acceptable ranges as loads, generation, and network conditions change. Utilities use transformer taps, regulators, capacitors, inverters, network reconfiguration, and operating procedures; large or rapidly changing loads can require more detailed models and faster local control. **Related definitions:** [Distribution automation](#entry%5FKB123) · [Service transformers](#entry%5FKB125) · [Voltage stability](#entry%5FKB056) · [Developer credibility](#entry%5FKB180) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-124 ### Service transformers How can utilities manage Service transformers? Service transformers reduce distribution voltage to the level required by a customer or facility. Their capacity, cooling, redundancy, fault characteristics, losses, location, and replacement availability matter for data centers, particularly when load grows in phases or specialized high-capacity units have long manufacturing timelines. **Related definitions:** [Voltage management](#entry%5FKB124) · [Load forecasting](#entry%5FKB126) · [Grid constraints](#entry%5FKB050) · [Curtailable service](#entry%5FKB173) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-125 ### Load forecasting What is Load forecasting? Load forecasting estimates future electricity demand by location, time, customer type, and operating condition. Forecasts support generation, transmission, distribution, rate, and procurement decisions, but large AI projects require separate attention to commercial readiness, phased ramp, duplication across requests, workload uncertainty, and the consequences of forecast error. **Related definitions:** [Service transformers](#entry%5FKB125) · [Flexible interconnection](#entry%5FKB127) · [Data center load forecasts](#entry%5FKB167) · [Large-load service](#entry%5FKB069) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-126 ### Flexible interconnection How does Flexible interconnection affect AI data center service? Flexible interconnection allows a customer or resource to connect under defined operating limits rather than waiting for all upgrades needed for unrestricted service. It may use curtailment, export limits, schedules, controls, or staged capacity, but requires enforceable agreements, telemetry, dispatch authority, performance monitoring, and clear rules for later firm service. **Related definitions:** [Load forecasting](#entry%5FKB126) · [Distributed energy resources](#entry%5FKB128) · [Non-firm service](#entry%5FKB172) · [Hybrid resources](#entry%5FKB225) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-127 ### Distributed energy resources Why does Distributed energy resources matter below the transmission system? Distributed energy resources are smaller generation, storage, flexible-load, and control assets connected to distribution systems or located behind customer meters. They can supply local needs or provide grid services, but their value depends on location, availability, interconnection, aggregation, controls, market access, and coordination with utility operations. **Related definitions:** [Flexible interconnection](#entry%5FKB127) · [Circuit hosting capacity](#entry%5FKB129) · [Grid-edge operations](#entry%5FKB072) · [Grid services](#entry%5FKB217) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-128 ### Circuit hosting capacity What data are needed to evaluate Circuit hosting capacity? Circuit hosting capacity is an estimate of how much additional load or generation a distribution circuit can accommodate without violating defined limits or requiring major upgrades. Results depend on models, data, operating assumptions, contingencies, resource behavior, and the specific location, so hosting-capacity maps are screening tools rather than service commitments. **Related definitions:** [Distributed energy resources](#entry%5FKB128) · [Distribution reliability](#entry%5FKB130) · [Storage interconnection](#entry%5FKB220) · [Data center load forecasts](#entry%5FKB167) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-129 ### Distribution reliability How can utilities manage Distribution reliability? Distribution reliability is the ability of local networks to deliver electricity with acceptable frequency and duration of interruptions and acceptable power quality. It depends on equipment condition, vegetation, weather, redundancy, protection, automation, maintenance, workforce, spare parts, and the concentration of critical customers. **Related definitions:** [Circuit hosting capacity](#entry%5FKB129) · [Smart meters](#entry%5FKB131) · [Substations](#entry%5FKB049) · [UPS systems](#entry%5FKB214) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-130 ### Smart meters What is Smart meters? A smart meter is a digital electricity meter capable of recording usage at shorter intervals and communicating data remotely. Smart meters support billing, outage detection, demand programs, distributed-resource integration, and analysis, but their value depends on communications, data governance, cybersecurity, customer protections, and useful operational integration. **Related definitions:** [Distribution reliability](#entry%5FKB130) · [Behind-the-meter assets](#entry%5FKB132) · [Grid modernization](#entry%5FKB052) · [Resilience value](#entry%5FKB223) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-131 ### Behind-the-meter assets How does Behind-the-meter assets affect AI data center service? Behind-the-meter assets are equipment located on the customer side of the utility meter, including generators, batteries, solar, UPS systems, controls, and flexible loads. They can change net grid demand and resilience, but utilities need visibility into operating modes, export, fault contribution, islanding, and performance during system events. **Related definitions:** [Smart meters](#entry%5FKB131) · [Microgrids](#entry%5FKB133) · [UPS systems](#entry%5FKB214) · [Flexible loads](#entry%5FKB171) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-132 ### Microgrids Why does Microgrids matter below the transmission system? A microgrid is a group of interconnected loads and energy resources that can operate as part of the wider grid and, when designed for it, separate and operate as an electrical island. Microgrids can improve resilience, but require controls, protection, generation or storage, fuel, interconnection agreements, and well-defined transitions between grid-connected and islanded operation. **Related definitions:** [Behind-the-meter assets](#entry%5FKB132) · [Load management](#entry%5FKB134) · [Co-located storage](#entry%5FKB219) · [Frequency stability](#entry%5FKB057) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-133 ### Load management What data are needed to evaluate Load management? Load management is the deliberate control, scheduling, reduction, or shifting of electricity use to meet customer or grid objectives. For data centers, potential tools include workload scheduling, cooling optimization, battery dispatch, redundancy management, and service curtailment, but performance must be measurable and compatible with computing-service obligations. **Related definitions:** [Microgrids](#entry%5FKB133) · [Local permitting](#entry%5FKB135) · [Flexible loads](#entry%5FKB171) · [Large-load service](#entry%5FKB069) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-134 ### Local permitting How can utilities manage Local permitting? Local permitting covers municipal or county approvals for land use, construction, fire safety, noise, traffic, water, backup generation, and other site impacts. Utilities can manage the interface by coordinating schedules and requirements early, while avoiding assumptions that a utility service commitment substitutes for local approval. **Related definitions:** [Load management](#entry%5FKB134) · [Investor-owned utilities](#entry%5FKB061) · [Co-located storage](#entry%5FKB219) · [Utility risk management](#entry%5FKB070) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-135 SECTION 10 | 15 DEFINITIONS ## Interconnection The studies, milestones, deposits, agreements, upgrades, and operating terms that move a project from proposal to energized service. **Terms in this section:** [Large-load interconnection](#entry%5FKB136) · [Generator interconnection](#entry%5FKB137) · [Queue reform](#entry%5FKB138) · [Milestone deposits](#entry%5FKB139) · [Studies](#entry%5FKB140) · [Facilities studies](#entry%5FKB141) · [Fast-track interconnection](#entry%5FKB142) · [Network upgrades](#entry%5FKB143) · [Load verification](#entry%5FKB144) · [Energization timelines](#entry%5FKB145) · [Service agreements](#entry%5FKB146) · [Curtailment terms](#entry%5FKB147) · [Interconnection risk](#entry%5FKB148) · [Speculative load](#entry%5FKB149) · [Phased energization](#entry%5FKB150) ### Large-load interconnection What is Large-load interconnection? Large-load interconnection is the process used to evaluate and connect a major new electricity customer to the grid. It includes load verification, technical studies, facility design, upgrade identification, cost responsibility, credit requirements, service agreements, operating obligations, construction milestones, and a schedule for phased or full energization. **Related definitions:** [Generator interconnection](#entry%5FKB137) · [Transmission interconnection](#entry%5FKB107) · [Customer obligations](#entry%5FKB179) · [Large-load policy](#entry%5FKB230) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-136 ### Generator interconnection Why does Generator interconnection delay data center projects? Generator interconnection can delay data-center projects when a customer’s supply plan depends on new generation that must complete separate studies, upgrades, permits, and construction. The load and generator may be commercially linked but electrically subject to different queues, tariffs, milestones, and system conditions. **Related definitions:** [Large-load interconnection](#entry%5FKB136) · [Queue reform](#entry%5FKB138) · [Critical path management](#entry%5FKB120) · [Campus development](#entry%5FKB039) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-137 ### Queue reform How should utilities evaluate Queue reform? Queue reform changes the rules and methods used to process interconnection requests so studies focus on credible projects and shared network needs. Common tools include readiness requirements, deposits, cluster studies, standardized data, withdrawal penalties, transparent milestones, study automation, and movement from project-by-project analysis toward proactive planning. **Related definitions:** [Generator interconnection](#entry%5FKB137) · [Milestone deposits](#entry%5FKB139) · [Transmission queues](#entry%5FKB119) · [Large-load policy](#entry%5FKB230) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-138 ### Milestone deposits What evidence makes Milestone deposits credible? A milestone deposit is a financial payment required at a defined stage to demonstrate project seriousness or secure study and construction resources. It is credible when sized to real costs and risks, tied to clear milestones, subject to transparent refund and forfeiture rules, and supported by enforceable customer obligations. **Related definitions:** [Queue reform](#entry%5FKB138) · [Studies](#entry%5FKB140) · [Customer obligations](#entry%5FKB179) · [Ratepayer protection](#entry%5FKB232) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-139 ### Studies How can policy improve Studies? Interconnection studies are engineering analyses used to determine whether and how a proposed load or resource can connect safely and reliably. Policy can improve them through standardized data, realistic models, coordinated study assumptions, transparent timelines, cluster methods, independent review, sufficient staffing, and rules that limit speculative or duplicative requests. **Related definitions:** [Milestone deposits](#entry%5FKB139) · [Facilities studies](#entry%5FKB141) · [Transmission planning](#entry%5FKB106) · [Data center load forecasts](#entry%5FKB167) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-140 ### Facilities studies What is Facilities studies? A facilities study is a detailed interconnection analysis that specifies the equipment, design, cost estimate, construction responsibilities, and schedule required to connect a project after broader system impacts have been assessed. It translates an approved interconnection concept into a more concrete buildable scope. **Related definitions:** [Studies](#entry%5FKB140) · [Fast-track interconnection](#entry%5FKB142) · [Reliability upgrades](#entry%5FKB118) · [Cost causation](#entry%5FKB178) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-141 ### Fast-track interconnection Why does Fast-track interconnection delay data center projects? Fast-track interconnection is an expedited process for projects that meet defined technical and size criteria or can connect without major upgrades. It can shorten timelines for low-impact projects, but calling a process fast-track does not remove engineering, equipment, permitting, data, protection, or construction constraints. **Related definitions:** [Facilities studies](#entry%5FKB141) · [Network upgrades](#entry%5FKB143) · [Modular data centers](#entry%5FKB044) · [Transmission interconnection](#entry%5FKB107) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-142 ### Network upgrades How should utilities evaluate Network upgrades? Network upgrades are changes to shared transmission or distribution facilities needed to accommodate a new project while maintaining reliability. Utilities evaluate necessity, alternatives, cost allocation, construction timing, broader benefits, and whether the upgrade should be directly assigned, planned regionally, or replaced by operating limits or staged service. **Related definitions:** [Fast-track interconnection](#entry%5FKB142) · [Load verification](#entry%5FKB144) · [Reliability upgrades](#entry%5FKB118) · [Cost causation](#entry%5FKB178) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-143 ### Load verification What evidence makes Load verification credible? Load verification is the evidence-based process used to determine whether a requested electricity demand is commercially credible, technically defined, and likely to materialize on the proposed schedule. Evidence can include site control, permits, financing, customer contracts, equipment orders, development milestones, corporate approvals, and financial assurance. **Related definitions:** [Network upgrades](#entry%5FKB143) · [Energization timelines](#entry%5FKB145) · [Developer credibility](#entry%5FKB180) · [Critical path management](#entry%5FKB120) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-144 ### Energization timelines How can policy improve Energization timelines? Energization timelines describe when a project can receive interim and full electrical service. Policy can improve them by standardizing study stages, publishing dependencies, coordinating permits and equipment, requiring customer milestones, enabling phased service where secure, and making schedule assumptions and responsibilities transparent. **Related definitions:** [Load verification](#entry%5FKB144) · [Service agreements](#entry%5FKB146) · [Critical path management](#entry%5FKB120) · [Developer credibility](#entry%5FKB180) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-145 ### Service agreements What is Service agreements? A service agreement is the contract between a utility or system provider and a customer defining the terms of electric service. For large loads, it may address capacity, rates, minimum payments, construction, deposits, credit, operating limits, curtailment, metering, milestones, default, termination, and responsibility for stranded facilities. **Related definitions:** [Energization timelines](#entry%5FKB145) · [Curtailment terms](#entry%5FKB147) · [Customer obligations](#entry%5FKB179) · [Critical path management](#entry%5FKB120) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-146 ### Curtailment terms Why does Curtailment terms delay data center projects? Curtailment terms define when, how much, and for how long a customer’s electricity service may be reduced. They can delay projects when operational limits are unclear, customer systems cannot tolerate interruption, compensation is unresolved, or utilities lack telemetry and controls needed to rely on the arrangement. **Related definitions:** [Service agreements](#entry%5FKB146) · [Interconnection risk](#entry%5FKB148) · [Customer obligations](#entry%5FKB179) · [Competitive transmission](#entry%5FKB108) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-147 ### Interconnection risk How should utilities evaluate Interconnection risk? Interconnection risk is the possibility that studies, upgrades, permits, costs, schedules, system conditions, or customer behavior prevent a project from connecting as expected. Utilities evaluate it through scenarios, milestone discipline, technical models, commercial evidence, contingency planning, contractual protections, and explicit allocation of uncertainty. **Related definitions:** [Curtailment terms](#entry%5FKB147) · [Speculative load](#entry%5FKB149) · [Developer credibility](#entry%5FKB180) · [Campus development](#entry%5FKB039) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-148 ### Speculative load What evidence makes Speculative load credible? Speculative load is a requested or forecast electricity demand that lacks sufficient evidence of commercial commitment, site readiness, financing, equipment, customers, or schedule. Speculative requests can distort planning and reserve scarce study or infrastructure capacity, although early-stage uncertainty must be distinguished from deliberate duplication or misrepresentation. **Related definitions:** [Interconnection risk](#entry%5FKB148) · [Phased energization](#entry%5FKB150) · [Developer credibility](#entry%5FKB180) · [Site selection](#entry%5FKB040) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-149 ### Phased energization How can policy improve Phased energization? Phased energization provides electrical service in defined increments as facilities, upgrades, and customer demand become ready. It can reduce stranded-cost and schedule risk, but each phase needs technical limits, milestone evidence, payment obligations, operating terms, and a clear relationship to the project’s ultimate requested load. **Related definitions:** [Speculative load](#entry%5FKB149) · [Cost allocation policy](#entry%5FKB231) · [Customer obligations](#entry%5FKB179)· [Campus development](#entry%5FKB039) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-150 SECTION 11 | 15 DEFINITIONS ## Reliability The planning and operating standards that keep the power system adequate, stable, secure, and accountable as large loads grow. **Terms in this section:** [NERC reliability standards](#entry%5FKB151) · [Resource adequacy](#entry%5FKB152)· [Operational reliability](#entry%5FKB153) · [Voltage ride-through](#entry%5FKB154) · [Frequency response](#entry%5FKB155) · [Emergency operations](#entry%5FKB156) · [Load shedding](#entry%5FKB157) · [Reliability entities](#entry%5FKB158) · [Planning reserve margins](#entry%5FKB159) · [Dynamic load models](#entry%5FKB160) · [Contingency analysis](#entry%5FKB161) · [Critical load](#entry%5FKB162) · [Reliability-must-run resources](#entry%5FKB163) · [Adequacy vs security](#entry%5FKB164) · [Reliability accountability](#entry%5FKB165) ### NERC reliability standards What is NERC reliability standards? NERC reliability standards are enforceable requirements approved through the North American reliability framework for planning and operating the bulk power system. They cover subjects such as planning, operations, protection, cybersecurity, modeling, emergency preparedness, and resource performance; applicability depends on registered functions and the facilities involved. **Related definitions:** [Resource adequacy](#entry%5FKB152) · [NERC oversight](#entry%5FKB228) · [Bulk power system](#entry%5FKB046) · [Grid reliability](#entry%5FKB053) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-151 ### Resource adequacy Why does Resource adequacy matter for large AI loads? Resource adequacy is the ability to maintain enough dependable resources and deliverable capacity to meet forecast demand and reserves over a planning horizon. Large AI loads matter because their size, concentration, ramp, flexibility, and forecast credibility can materially change capacity needs and the timing of resource procurement. **Related definitions:** [NERC reliability standards](#entry%5FKB151) · [Operational reliability](#entry%5FKB153) · [Resource adequacy](#entry%5FKB082) · [Frequency stability](#entry%5FKB057) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-152 ### Operational reliability How can data centers affect Operational reliability? Operational reliability is the ability to keep the power system within secure limits in real time and through credible disturbances. Data centers can affect it through concentrated demand, voltage and reactive-power needs, abrupt load changes, protection behavior, backup-system transitions, limited observability, or large simultaneous responses to external events. **Related definitions:** [Resource adequacy](#entry%5FKB152) · [Voltage ride-through](#entry%5FKB154) · [Grid reliability](#entry%5FKB053) · [Resource adequacy](#entry%5FKB082) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-153 ### Voltage ride-through What evidence is needed to evaluate Voltage ride-through? Voltage ride-through is the capability of equipment to remain connected and behave predictably during short voltage disturbances rather than disconnecting immediately. Evidence may include equipment specifications, test results, validated dynamic models, protection settings, commissioning records, and demonstrated performance under conditions relevant to the grid. **Related definitions:** [Operational reliability](#entry%5FKB153) · [Frequency response](#entry%5FKB155) · [Voltage stability](#entry%5FKB056) · [Developer credibility](#entry%5FKB180) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-154 ### Frequency response How should reliability institutions govern Frequency response? Frequency response is the change in generation or demand that helps arrest and correct system frequency after an imbalance. Reliability institutions govern it through performance requirements, reserves, controls, telemetry, testing, models, and operating procedures, including expectations for large resources and loads whose sudden actions could be consequential. **Related definitions:** [Voltage ride-through](#entry%5FKB154) · [Emergency operations](#entry%5FKB156) · [Frequency stability](#entry%5FKB057) · [Ancillary services](#entry%5FKB079) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-155 ### Emergency operations What is Emergency operations? Emergency operations are predefined actions used when normal system conditions deteriorate or reliability is threatened. They can include committing reserves, reconfiguring the network, importing power, requesting conservation, curtailing transactions, interrupting load, deploying restoration plans, and coordinating communications across responsible entities. **Related definitions:** [Frequency response](#entry%5FKB155) · [Load shedding](#entry%5FKB157) · [Bulk power system](#entry%5FKB046) · [NERC oversight](#entry%5FKB228) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-156 ### Load shedding Why does Load shedding matter for large AI loads? Load shedding is the intentional disconnection or reduction of customer demand to prevent wider system collapse or equipment damage. Large loads matter because their size can make them valuable emergency tools or significant contingencies, but shedding plans must consider selectivity, critical functions, restoration sequence, customer agreements, and operational control. **Related definitions:** [Emergency operations](#entry%5FKB156) · [Reliability entities](#entry%5FKB158) · [Frequency stability](#entry%5FKB057) · [High load factor customers](#entry%5FKB176) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-157 ### Reliability entities How can data centers affect Reliability entities? Reliability entities are organizations registered or assigned to perform defined planning and operating functions for the interconnected power system, such as balancing authorities, reliability coordinators, transmission operators, planning coordinators, and resource operators. A large data center may not automatically be one, but its scale can create obligations and information needs similar to reliability-relevant facilities. **Related definitions:** [Load shedding](#entry%5FKB157) · [Planning reserve margins](#entry%5FKB159) · [Bulk power system](#entry%5FKB046) · [Resource adequacy](#entry%5FKB082) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-158 ### Planning reserve margins What evidence is needed to evaluate Planning reserve margins? A planning reserve margin is the amount by which expected resource capacity exceeds forecast peak demand, expressed as a percentage or quantity. Evidence supporting the margin includes load uncertainty, resource accreditation, outage risk, weather, transmission limits, operating reserves, demand response, and the reliability criterion used in the region. **Related definitions:** [Reliability entities](#entry%5FKB158) · [Dynamic load models](#entry%5FKB160) · [Resource adequacy](#entry%5FKB082) · [Grid reliability](#entry%5FKB053) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-159 ### Dynamic load models How should reliability institutions govern Dynamic load models? A dynamic load model represents how electricity demand changes during and after disturbances rather than treating load as a fixed number. For data centers, credible models may need to reflect power electronics, UPS systems, motors, cooling, protection, control logic, backup transitions, and restoration behavior. **Related definitions:** [Planning reserve margins](#entry%5FKB159) · [Contingency analysis](#entry%5FKB161) · [Frequency stability](#entry%5FKB057) · [Data center load forecasts](#entry%5FKB167) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-160 ### Contingency analysis What is Contingency analysis? Contingency analysis evaluates whether the power system remains within acceptable limits after specified outages or disturbances, such as the loss of a line, transformer, generator, or major load. It is central to planning and operations because reliability depends on performance after failures, not only under normal conditions. **Related definitions:** [Dynamic load models](#entry%5FKB160) · [Critical load](#entry%5FKB162) · [Grid reliability](#entry%5FKB053) · [FERC jurisdiction](#entry%5FKB226) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-161 ### Critical load Why does Critical load matter for large AI loads? Critical load is electricity demand whose interruption would cause unacceptable safety, security, economic, operational, or public-service consequences. The designation should identify the specific functions requiring continuity, the duration and quality of required service, backup capability, restoration priority, and who bears the cost of enhanced resilience. **Related definitions:** [Contingency analysis](#entry%5FKB161) · [Reliability-must-run resources](#entry%5FKB163) · [Curtailable service](#entry%5FKB173) · [State public utility commissions](#entry%5FKB227) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-162 ### Reliability-must-run resources How can data centers affect Reliability-must-run resources? Reliability-must-run resources are generators retained or dispatched because their operation is needed to maintain local reliability when market revenues alone would not keep them available. Data-center growth can change the need for such resources by increasing local demand before transmission or replacement generation is completed. **Related definitions:** [Critical load](#entry%5FKB162) · [Adequacy vs security](#entry%5FKB164) · [Bulk power system](#entry%5FKB046) · [Market power](#entry%5FKB083) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-163 ### Adequacy vs security What evidence is needed to evaluate Adequacy vs security? Adequacy concerns whether sufficient resources and network capability exist to meet demand over time; security concerns whether the system can withstand disturbances and remain within operating limits. A system may be adequate in annual energy terms yet insecure during a specific contingency, or secure today but inadequate for future peak demand. **Related definitions:** [Reliability-must-run resources](#entry%5FKB163) · [Reliability accountability](#entry%5FKB165) · [Capacity markets](#entry%5FKB077) · [Grid reliability](#entry%5FKB053) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-164 ### Reliability accountability How should reliability institutions govern Reliability accountability? Reliability accountability is the clear assignment of obligations, authority, information, performance standards, and consequences among planners, operators, utilities, customers, and regulators. It prevents critical requirements from being treated as voluntary attributes and ensures that models, telemetry, flexibility, and emergency actions are verifiable. **Related definitions:** [Adequacy vs security](#entry%5FKB164) · [Large-load policy](#entry%5FKB230) · [Retail choice](#entry%5FKB085) · [Grid operations](#entry%5FKB055) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-165 SECTION 12 | 15 DEFINITIONS ## Large Loads The tariffs, forecasts, financial assurances, service classes, flexibility obligations, and customer protections associated with major electricity demand. **Terms in this section:** [Large-load tariffs](#entry%5FKB166) · [Data center load forecasts](#entry%5FKB167) · [Phantom data centers](#entry%5FKB168) · [Minimum demand charges](#entry%5FKB169) · [Collateral requirements](#entry%5FKB170) · [Flexible loads](#entry%5FKB171) · [Non-firm service](#entry%5FKB172) · [Curtailable service](#entry%5FKB173) · [Load shape](#entry%5FKB174) · [Load ramping](#entry%5FKB175) · [High load factor customers](#entry%5FKB176) · [Coincident peak](#entry%5FKB177) · [Cost causation](#entry%5FKB178) · [Customer obligations](#entry%5FKB179) · [Developer credibility](#entry%5FKB180) ### Large-load tariffs What is Large-load tariffs? A large-load tariff is a utility or transmission rate schedule designed for customers whose demand is large enough to create unusual planning, construction, financial, or reliability requirements. It may include deposits, minimum bills, contract demand, contribution requirements, credit support, ramp schedules, curtailment terms, exit charges, and customer-specific facilities. **Related definitions:** [Data center load forecasts](#entry%5FKB167) · [Rate design](#entry%5FKB067) · [Service agreements](#entry%5FKB146) · [Reliability entities](#entry%5FKB158) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-166 ### Data center load forecasts Why does Data center load forecasts matter for data center development? A data-center load forecast estimates the magnitude, timing, location, and operating pattern of electricity demand from proposed and existing facilities. Credible forecasts distinguish announced capacity from committed load, remove duplicate requests, model phased ramp and utilization, and disclose uncertainty rather than treating every interconnection inquiry as eventual consumption. **Related definitions:** [Large-load tariffs](#entry%5FKB166) · [Phantom data centers](#entry%5FKB168) · [Dynamic load models](#entry%5FKB160) · [Speculative load](#entry%5FKB149) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-167 ### Phantom data centers How should utilities treat Phantom data centers? Phantom data centers are proposed or reported projects that appear in development pipelines, forecasts, queues, or public claims without enough evidence that they will be financed, built, equipped, and energized as described. The concept focuses attention on commitment quality, not on dismissing legitimate early-stage development. **Related definitions:** [Data center load forecasts](#entry%5FKB167) · [Minimum demand charges](#entry%5FKB169) · [Speculative load](#entry%5FKB149) · [Public power](#entry%5FKB062) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-168 ### Minimum demand charges What risks does Minimum demand charges create for ratepayers? A minimum demand charge requires a customer to pay for at least a specified level of demand or contracted capacity even when actual usage is lower. It protects other customers when utilities build facilities for a large load that ramps slowly, underperforms, relocates, or is cancelled. **Related definitions:** [Phantom data centers](#entry%5FKB168) · [Collateral requirements](#entry%5FKB170) · [Rate design](#entry%5FKB067) · [Service agreements](#entry%5FKB146) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-169 ### Collateral requirements How can regulators make Collateral requirements enforceable? Collateral requirements obligate a customer or developer to provide cash, a letter of credit, guarantee, bond, or other financial support for utility exposure. They are enforceable when the amount, eligible instruments, credit standards, replenishment, draw conditions, duration, transfer, and release rules are clearly documented. **Related definitions:** [Minimum demand charges](#entry%5FKB169) · [Flexible loads](#entry%5FKB171) · [NERC reliability standards](#entry%5FKB151) · [Milestone deposits](#entry%5FKB139) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-170 ### Flexible loads What is Flexible loads? Flexible loads are electricity users capable of changing consumption in response to prices, incentives, schedules, or operating instructions. Flexibility may involve shifting computation, reducing cooling load, using batteries, delaying noncritical work, or curtailing service, but it is valuable only when performance is measurable, available, and compatible with customer obligations. **Related definitions:** [Collateral requirements](#entry%5FKB170) · [Non-firm service](#entry%5FKB172) · [Phased energization](#entry%5FKB150) · [Load shedding](#entry%5FKB157) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-171 ### Non-firm service Why does Non-firm service matter for data center development? Non-firm service provides electricity subject to interruption or reduction under defined conditions, often in exchange for faster connection or lower cost than fully firm service. Its value depends on the curtailment trigger, notice, duration, frequency, restoration, compensation, telemetry, enforcement, and whether the customer can actually respond. **Related definitions:** [Flexible loads](#entry%5FKB171) · [Curtailable service](#entry%5FKB173) · [Curtailment terms](#entry%5FKB147) · [Critical load](#entry%5FKB162) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-172 ### Curtailable service How should utilities treat Curtailable service? Curtailable service is a tariff or contract under which a customer agrees to reduce load when directed or when specified system conditions occur. Utilities should evaluate the customer’s technical capability, business tolerance, communications, controls, testing, baseline, response time, duration, rebound, and consequences for nonperformance. **Related definitions:** [Non-firm service](#entry%5FKB172) · [Load shape](#entry%5FKB174) · [Customer classes](#entry%5FKB068) · [Large-load interconnection](#entry%5FKB136) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-173 ### Load shape What risks does Load shape create for ratepayers? Load shape is the pattern of electricity demand across hours, days, seasons, and operating conditions. A high annual energy total can create very different infrastructure and market needs depending on peak timing, ramp rates, utilization, weather sensitivity, redundancy, maintenance, and the ability to shift work. **Related definitions:** [Curtailable service](#entry%5FKB173) · [Load ramping](#entry%5FKB175) · [Large-load service](#entry%5FKB069) · [Enterprise data centers](#entry%5FKB033) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-174 ### Load ramping How can regulators make Load ramping enforceable? Load ramping is the rate at which customer demand increases or decreases over time, both during facility buildout and within operations. Enforceable ramp rules specify capacity milestones, notice, metering, maximum changes, testing, consequences for deviation, and how utility construction aligns with demonstrated demand. **Related definitions:** [Load shape](#entry%5FKB174) · [High load factor customers](#entry%5FKB176) · [Service agreements](#entry%5FKB146) · [Utility planning](#entry%5FKB064) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-175 ### High load factor customers What is High load factor customers? A high-load-factor customer uses a large share of its peak demand consistently over time. Data centers often seek high utilization, which can improve use of fixed infrastructure but also creates sustained energy, fuel, cooling, and generation requirements and reduces the amount of unused capacity available during system stress. **Related definitions:** [Load ramping](#entry%5FKB175) · [Coincident peak](#entry%5FKB177) · [Customer classes](#entry%5FKB068) · [Planning reserve margins](#entry%5FKB159) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-176 ### Coincident peak Why does Coincident peak matter for data center development? Coincident peak is the portion of a customer’s demand that occurs at the same time as the relevant system, network, or class peak. It matters because many generation, transmission, and distribution costs are driven by collective peak conditions rather than by each customer’s separate maximum demand. **Related definitions:** [High load factor customers](#entry%5FKB176) · [Cost causation](#entry%5FKB178) · [Large-load service](#entry%5FKB069) · [Generator interconnection](#entry%5FKB137) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-177 ### Cost causation How should utilities treat Cost causation? Cost causation is the regulatory principle that costs should be assigned to the customers or activities that cause them, while recognizing shared benefits and practical rate design. For large loads, it guides treatment of dedicated facilities, network upgrades, capacity procurement, planning costs, and stranded investment risk. **Related definitions:** [Coincident peak](#entry%5FKB177) · [Customer obligations](#entry%5FKB179) · [Cost recovery](#entry%5FKB066) · [Network upgrades](#entry%5FKB143) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-178 ### Customer obligations What risks does Customer obligations create for ratepayers? Customer obligations are the contractual, financial, technical, and operating responsibilities a large-load customer accepts as a condition of service. They may include accurate data, deposits, minimum payments, milestones, equipment standards, telemetry, curtailment, power-quality limits, notice of changes, and responsibility for customer-caused costs. **Related definitions:** [Cost causation](#entry%5FKB178) · [Developer credibility](#entry%5FKB180) · [Service agreements](#entry%5FKB146) · [Rate design](#entry%5FKB067) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-179 ### Developer credibility How can regulators make Developer credibility enforceable? Developer credibility is the demonstrated capacity and commitment to deliver a proposed project. Regulators and utilities can evaluate it through experience, financial strength, site control, customer contracts, permits, equipment orders, financing, corporate approvals, schedule evidence, transparent ownership, and willingness to provide meaningful financial assurance. **Related definitions:** [Customer obligations](#entry%5FKB179) · [Load verification](#entry%5FKB144) · [Site selection](#entry%5FKB040) · [Interconnection risk](#entry%5FKB148) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-180 SECTION 13 | 15 DEFINITIONS ## Semiconductors The chips, memory, packaging, supply chains, efficiency trends, and compute-density changes driving AI infrastructure requirements. **Terms in this section:** [GPUs](#entry%5FKB181) · [AI accelerators](#entry%5FKB182) · [HBM memory](#entry%5FKB183) · [Chip packaging](#entry%5FKB184) · [Semiconductor fabs](#entry%5FKB185) · [Power efficiency](#entry%5FKB186) · [Moore's Law](#entry%5FKB187) · [Advanced nodes](#entry%5FKB188) · [Chip supply chains](#entry%5FKB189) · [Export controls](#entry%5FKB190) · [AI hardware competition](#entry%5FKB191) · [ASICs](#entry%5FKB192) · [Networking chips](#entry%5FKB193) · [Thermal design power](#entry%5FKB194) · [Compute density](#entry%5FKB195) ### GPUs What is GPUs? A graphics processing unit, or GPU, is a highly parallel processor originally developed for graphics and now widely used for AI and scientific computing. GPUs perform many mathematical operations simultaneously and are typically combined with high-bandwidth memory, fast interconnects, specialized software, and substantial power and cooling infrastructure. **Related definitions:** [AI accelerators](#entry%5FKB182) · [GPU clusters](#entry%5FKB003) · [Inference](#entry%5FKB019) · [Air cooling](#entry%5FKB196) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-181 ### AI accelerators How does AI accelerators affect AI energy demand? An AI accelerator is hardware optimized for machine-learning operations, such as matrix multiplication, tensor processing, or low-precision arithmetic. Accelerators can improve performance per watt for specific workloads, but total energy demand depends on deployment scale, utilization, memory, networking, cooling, and whether lower computing cost increases overall usage. **Related definitions:** [GPUs](#entry%5FKB181)· [HBM memory](#entry%5FKB183) · [Model efficiency](#entry%5FKB024) · [GPU clusters](#entry%5FKB003) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-182 ### HBM memory Why does HBM memory matter for data center power density? High-bandwidth memory, or HBM, is stacked memory designed to move large quantities of data between memory and processors at very high speed. It supports accelerator performance and dense computing, but packaging complexity, power, heat, manufacturing yield, supplier concentration, and capacity constraints can affect system availability and data-center design. **Related definitions:** [AI accelerators](#entry%5FKB182) · [Chip packaging](#entry%5FKB184) · [Air cooling](#entry%5FKB196) · [Inference](#entry%5FKB019) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-183 ### Chip packaging What supply-chain risks affect Chip packaging? Chip packaging connects semiconductor dies to memory, power, cooling, and external signals and increasingly combines multiple components into advanced packages. Supply risks include limited specialized manufacturing capacity, substrates, interposers, bonding equipment, thermal materials, yield, testing, and dependence on a small number of suppliers. **Related definitions:** [HBM memory](#entry%5FKB183) · [Semiconductor fabs](#entry%5FKB185) · [Digital infrastructure supply chain](#entry%5FKB012) · [Immersion cooling](#entry%5FKB198) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-184 ### Semiconductor fabs How could Semiconductor fabs change future electricity demand? A semiconductor fabrication plant, or fab, manufactures chips through repeated deposition, lithography, etching, implantation, cleaning, and testing steps in highly controlled environments. Fabs use substantial electricity, water, chemicals, gases, and capital, so new manufacturing capacity can create significant regional infrastructure demand separate from data centers. **Related definitions:** [Chip packaging](#entry%5FKB184) · [Power efficiency](#entry%5FKB186) · [Water stress](#entry%5FKB201) · [Generative AI](#entry%5FKB017) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-185 ### Power efficiency What is Power efficiency? Power efficiency is the useful computing output delivered per unit of electrical input, measured with workload-appropriate metrics. It can improve through chip architecture, precision, memory, networking, software, utilization, cooling, and power conversion, but comparisons are meaningful only when performance, workload, quality, and system boundaries are consistent. **Related definitions:** [Semiconductor fabs](#entry%5FKB185) · [Moore's Law](#entry%5FKB187) · [Model efficiency](#entry%5FKB024) · [GPU clusters](#entry%5FKB003) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-186 ### Moore's Law How does Moore's Law affect AI energy demand? Moore’s Law describes the historical tendency for transistor density on integrated circuits to increase over time, enabling more computation at lower unit cost. Its practical effect now depends on architecture, packaging, memory, software, and manufacturing economics, and greater efficiency can shift demand downstream by making more AI applications economically viable. **Related definitions:** [Power efficiency](#entry%5FKB186) · [Advanced nodes](#entry%5FKB188) · [Model efficiency](#entry%5FKB024) · [AI compute infrastructure](#entry%5FKB001) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-187 ### Advanced nodes Why does Advanced nodes matter for data center power density? Advanced semiconductor nodes use smaller and more complex manufacturing features to increase density, performance, or energy efficiency. Node labels are not direct physical measurements across manufacturers, and benefits depend on design, voltage, yield, packaging, workload, and cost rather than on the node name alone. **Related definitions:** [Moore's Law](#entry%5FKB187) · [Chip supply chains](#entry%5FKB189) · [Edge AI infrastructure](#entry%5FKB008) · [Closed-loop systems](#entry%5FKB203) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-188 ### Chip supply chains What supply-chain risks affect Chip supply chains? Chip supply chains include design tools, intellectual property, fabrication, materials, manufacturing equipment, packaging, testing, logistics, and specialized labor distributed across countries and companies. Concentration at critical stages can turn geopolitical tension, disasters, export rules, or factory disruptions into constraints on AI infrastructure deployment. **Related definitions:** [Advanced nodes](#entry%5FKB188) · [Export controls](#entry%5FKB190) · [Transformer supply chains](#entry%5FKB271) · [Digital infrastructure supply chain](#entry%5FKB012) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-189 ### Export controls How could Export controls change future electricity demand? Export controls restrict the sale or transfer of specified chips, equipment, software, technology, or services to certain destinations or users. They can redirect hardware supply, encourage domestic alternatives, alter data-center geography, complicate cloud access, and change how firms plan compute capacity and supply-chain resilience. **Related definitions:** [Chip supply chains](#entry%5FKB189) · [AI hardware competition](#entry%5FKB191) · [Digital infrastructure supply chain](#entry%5FKB012) · [Cooling retrofits](#entry%5FKB206) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-190 ### AI hardware competition What is AI hardware competition? AI hardware competition is the contest among chip designers, manufacturers, cloud providers, and system companies to deliver better performance, efficiency, software support, availability, and cost. Competition can diversify supply and accelerate innovation, but switching costs, proprietary ecosystems, manufacturing concentration, and workload compatibility can limit buyer flexibility. **Related definitions:** [Export controls](#entry%5FKB190) · [ASICs](#entry%5FKB192) · [Model efficiency](#entry%5FKB024) · [Cooling retrofits](#entry%5FKB206) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-191 ### ASICs How does ASICs affect AI energy demand? An application-specific integrated circuit, or ASIC, is a chip designed for a particular function or narrow class of workloads. AI ASICs can deliver high efficiency and performance at scale, but require substantial design investment, manufacturing volume, software support, and confidence that workloads will remain compatible through the chip’s useful life. **Related definitions:** [AI hardware competition](#entry%5FKB191) · [Networking chips](#entry%5FKB193) · [AI demand drivers](#entry%5FKB029) · [Air cooling](#entry%5FKB196) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-192 ### Networking chips Why does Networking chips matter for data center power density? Networking chips move data among processors, memory, storage, servers, and facilities. AI clusters depend on switches, network interfaces, optical components, and specialized interconnects because model performance can be limited by communication rather than raw computation, and network equipment adds power, heat, cost, and supply-chain dependencies. **Related definitions:** [ASICs](#entry%5FKB192) · [Thermal design power](#entry%5FKB194) · [AI compute infrastructure](#entry%5FKB001) · [Inference](#entry%5FKB019) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-193 ### Thermal design power What supply-chain risks affect Thermal design power? Thermal design power is a manufacturer-defined measure used to guide the cooling and power design for a component under specified operating assumptions. It is not always equal to maximum or average consumption, so facility designers need actual workload profiles, transient behavior, configuration limits, and measurement data. **Related definitions:** [Networking chips](#entry%5FKB193) · [Compute density](#entry%5FKB195) · [Water usage effectiveness](#entry%5FKB200) · [Digital infrastructure supply chain](#entry%5FKB012) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-194 ### Compute density How could Compute density change future electricity demand? Compute density is the amount of processing capability concentrated in a physical area, rack, facility, or unit of power. Higher density can improve land and network efficiency but increases heat removal, power distribution, fault concentration, maintenance complexity, equipment weight, and the consequences of local infrastructure failure. **Related definitions:** [Thermal design power](#entry%5FKB194) · [Power density](#entry%5FKB011) · [Cooling energy penalty](#entry%5FKB209) · [Liquid cooling](#entry%5FKB197) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-195 SECTION 14 | 15 DEFINITIONS ## Cooling and Water The thermal-management systems, water demands, permitting issues, and climate constraints associated with dense computing facilities. **Terms in this section:** [Air cooling](#entry%5FKB196) · [Liquid cooling](#entry%5FKB197) · [Immersion cooling](#entry%5FKB198) · [Direct-to-chip cooling](#entry%5FKB199) · [Water usage effectiveness](#entry%5FKB200) · [Water stress](#entry%5FKB201) · [Evaporative cooling](#entry%5FKB202) · [Closed-loop systems](#entry%5FKB203) · [Waste heat reuse](#entry%5FKB204) · [Thermal management](#entry%5FKB205) · [Cooling retrofits](#entry%5FKB206) · [Water permitting](#entry%5FKB207) · [Recycled water](#entry%5FKB208) · [Cooling energy penalty](#entry%5FKB209) · [Climate risk](#entry%5FKB210) ### Air cooling What is Air cooling? Air cooling removes heat by moving conditioned air through or around computing equipment. It is mature and widely supported, but very high rack densities can require large airflow, fan energy, containment, lower supply temperatures, or supplemental liquid cooling, and performance varies with climate and facility design. **Related definitions:** [Liquid cooling](#entry%5FKB197) · [Cooling systems](#entry%5FKB037) · [Power density](#entry%5FKB011) · [Modular data centers](#entry%5FKB044) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-196 ### Liquid cooling Why does Liquid cooling matter for AI data centers? Liquid cooling transfers heat from processors or servers to a liquid rather than relying only on air. Direct-to-chip cold plates, rear-door heat exchangers, and immersion systems can support higher density and reduce fan energy, but require compatible equipment, pumps, heat exchangers, water-quality controls, leak management, and new maintenance practices. **Related definitions:** [Air cooling](#entry%5FKB196) · [Immersion cooling](#entry%5FKB198) · [Cooling systems](#entry%5FKB037) · [Edge AI infrastructure](#entry%5FKB008) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-197 ### Immersion cooling How does Immersion cooling affect water and electricity use? Immersion cooling places electronic equipment in a nonconductive fluid that absorbs heat directly. It can support high density and efficient heat transfer, but requires specialized tanks, fluids, hardware compatibility, service procedures, material testing, fluid management, and a business case that justifies changing standard server operations. **Related definitions:** [Liquid cooling](#entry%5FKB197) · [Direct-to-chip cooling](#entry%5FKB199) · [Cooling systems](#entry%5FKB037) · [Edge AI infrastructure](#entry%5FKB008) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-198 ### Direct-to-chip cooling What constraints limit Direct-to-chip cooling? Direct-to-chip cooling uses cold plates attached to processors or other components to carry heat into a liquid loop. Constraints include server compatibility, manifold and piping design, water quality, leak detection, pumps, heat rejection, retrofit complexity, maintenance skills, and the remaining air-cooling load from components not connected to liquid. **Related definitions:** [Immersion cooling](#entry%5FKB198) · [Water usage effectiveness](#entry%5FKB200) · [Cooling systems](#entry%5FKB037) · [Water sustainability](#entry%5FKB265) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-199 ### Water usage effectiveness How should communities evaluate Water usage effectiveness? Water usage effectiveness, or WUE, is a data-center metric that relates annual site water use to information-technology energy consumption. It helps compare water intensity, but results depend on boundaries, climate, cooling design, water source, accounting methods, and whether offsite water used in electricity generation is included. **Related definitions:** [Direct-to-chip cooling](#entry%5FKB199) · [Water stress](#entry%5FKB201) · [Power usage effectiveness](#entry%5FKB036) · [AI load corridors](#entry%5FKB010) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-200 ### Water stress What is Water stress? Water stress occurs when demand for usable water approaches or exceeds available supply or when quality, infrastructure, environmental needs, and competing uses limit access. A data center can have modest global water impact yet create significant local concern in a stressed basin, especially during hot or dry periods. **Related definitions:** [Water usage effectiveness](#entry%5FKB200) · [Evaporative cooling](#entry%5FKB202)· [Water sustainability](#entry%5FKB265) · [Data center permitting](#entry%5FKB041) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-201 ### Evaporative cooling Why does Evaporative cooling matter for AI data centers? Evaporative cooling removes heat by evaporating water, often reducing electricity use compared with compressor-based cooling under suitable conditions. Its performance depends on climate, water availability and quality, treatment, discharge, seasonal operation, and community priorities, creating a tradeoff between direct water use and cooling energy. **Related definitions:** [Water stress](#entry%5FKB201) · [Closed-loop systems](#entry%5FKB203) · [Water sustainability](#entry%5FKB265) · [Cooling systems](#entry%5FKB037) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-202 ### Closed-loop systems How does Closed-loop systems affect water and electricity use? A closed-loop cooling system recirculates a working fluid rather than continuously consuming it. Closed loops can reduce routine water withdrawal, but they still require heat rejection, pumping, treatment, maintenance, and sometimes makeup water, and their total energy and water effects depend on the complete system design. **Related definitions:** [Evaporative cooling](#entry%5FKB202) · [Waste heat reuse](#entry%5FKB204) · [Rack density](#entry%5FKB035) · [Water sustainability](#entry%5FKB265) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-203 ### Waste heat reuse What constraints limit Waste heat reuse? Waste heat reuse captures heat from data centers for buildings, industrial processes, district-energy systems, agriculture, or other uses. Constraints include temperature, distance, timing, infrastructure cost, customer demand, ownership, backup arrangements, and the need to ensure the reuse system does not compromise data-center cooling reliability. **Related definitions:** [Closed-loop systems](#entry%5FKB203) · [Thermal management](#entry%5FKB205) · [Circular economy](#entry%5FKB266) · [Modular data centers](#entry%5FKB044) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-204 ### Thermal management How should communities evaluate Thermal management? Thermal management is the coordinated control of heat generation, transfer, cooling, airflow, liquid systems, temperatures, and equipment limits. It spans chips, servers, racks, rooms, plants, and heat rejection, and must balance reliability, density, energy, water, maintenance, redundancy, and capital cost. **Related definitions:** [Waste heat reuse](#entry%5FKB204) · [Cooling retrofits](#entry%5FKB206) · [Cooling systems](#entry%5FKB037) · [AI campuses](#entry%5FKB014) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-205 ### Cooling retrofits What is Cooling retrofits? A cooling retrofit upgrades an existing facility to support new equipment, higher density, better efficiency, or different cooling technology. Retrofits are constrained by building layout, electrical capacity, downtime, structural limits, piping, controls, commissioning, equipment compatibility, and the need to maintain service during construction. **Related definitions:** [Thermal management](#entry%5FKB205) · [Water permitting](#entry%5FKB207) · [Power density](#entry%5FKB011) · [Modular data centers](#entry%5FKB044) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-206 ### Water permitting Why does Water permitting matter for AI data centers? Water permitting is the legal and administrative approval process for withdrawing, consuming, storing, discharging, or altering water. Data-center requirements may involve municipal service, wells, rights, drought restrictions, treatment, wastewater, stormwater, wetlands, and environmental review, and vary substantially by jurisdiction and cooling design. **Related definitions:** [Cooling retrofits](#entry%5FKB206) · [Recycled water](#entry%5FKB208) · [Data center permitting](#entry%5FKB041) · [Water sustainability](#entry%5FKB265) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-207 ### Recycled water How does Recycled water affect water and electricity use? Recycled water is treated wastewater or other nonpotable water reused for cooling, irrigation, industrial processes, or environmental purposes. It can reduce demand for drinking-quality water, but requires suitable quality, pipelines, contracts, treatment, reliability, health protections, and a comparison of energy, cost, and alternative uses. **Related definitions:** [Water permitting](#entry%5FKB207) · [Cooling energy penalty](#entry%5FKB209) · [Environmental justice](#entry%5FKB269) · [AI load corridors](#entry%5FKB010) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-208 ### Cooling energy penalty What constraints limit Cooling energy penalty? The cooling energy penalty is the electricity consumed by fans, pumps, chillers, cooling towers, controls, and heat-rejection equipment beyond the computing load. It depends on climate, density, temperature set points, redundancy, system efficiency, utilization, maintenance, and the amount of heat that can be rejected without mechanical refrigeration. **Related definitions:** [Recycled water](#entry%5FKB208) · [Climate risk](#entry%5FKB210) · [Power density](#entry%5FKB011) · [Cooling systems](#entry%5FKB037) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-209 ### Climate risk How should communities evaluate Climate risk? Climate risk is the exposure of facilities and infrastructure to acute events and long-term changes such as heat, drought, flooding, wildfire, storms, sea-level rise, and changing water availability. Data-center evaluation should include both direct site hazards and indirect risks to grids, fuel, communications, transportation, suppliers, and communities. **Related definitions:** [Cooling energy penalty](#entry%5FKB209) · [Location strategy](#entry%5FKB005) · [Climate resilience](#entry%5FKB270) · [Cooling systems](#entry%5FKB037) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-210 SECTION 15 | 15 DEFINITIONS ## Energy Storage How batteries, long-duration storage, UPS systems, hybrid resources, and dispatch strategies support reliability, flexibility, and resilience. **Terms in this section:** [Battery storage](#entry%5FKB211) · [Long-duration storage](#entry%5FKB212) · [Thermal storage](#entry%5FKB213) · [UPS systems](#entry%5FKB214) · [Backup batteries](#entry%5FKB215) · [Storage duration](#entry%5FKB216) · [Grid services](#entry%5FKB217) · [Demand charge management](#entry%5FKB218) · [Co-located storage](#entry%5FKB219) · [Storage interconnection](#entry%5FKB220) · [Battery safety](#entry%5FKB221) · [Storage economics](#entry%5FKB222) · [Resilience value](#entry%5FKB223) · [Storage dispatch](#entry%5FKB224) · [Hybrid resources](#entry%5FKB225) ### Battery storage What is Battery storage? Battery storage converts electricity into stored electrochemical energy and later returns it to the electrical system. Batteries can provide backup power, fast frequency response, peak management, energy shifting, and grid services, but duration, degradation, safety, controls, interconnection, replacement, and operating strategy determine actual value. **Related definitions:** [Long-duration storage](#entry%5FKB212) · [Generation interconnection](#entry%5FKB104) · [Flexible loads](#entry%5FKB171) · [Co-located generation](#entry%5FKB097) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-211 ### Long-duration storage Can Long-duration storage support AI data centers? Long-duration energy storage is storage designed to deliver energy for substantially longer periods than typical short-duration batteries, although no single duration definition applies everywhere. It may help cover extended renewable variability, outages, or capacity needs, but cost, efficiency, siting, maturity, cycling, and market compensation vary by technology. **Related definitions:** [Battery storage](#entry%5FKB211) · [Thermal storage](#entry%5FKB213) · [Clean firm power](#entry%5FKB099) · [Non-firm service](#entry%5FKB172) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-212 ### Thermal storage How does Thermal storage affect grid flexibility? Thermal storage stores heating or cooling energy for later use, using chilled water, ice, hot water, phase-change materials, or other media. In data centers it can shift cooling electricity, reduce peak demand, and provide operating flexibility, but space, efficiency, controls, temperature requirements, and integration with cooling systems matter. **Related definitions:** [Long-duration storage](#entry%5FKB212) · [UPS systems](#entry%5FKB214) · [Combined-cycle plants](#entry%5FKB101) · [Frequency response](#entry%5FKB155) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-213 ### UPS systems What are the limits of UPS systems? An uninterruptible power supply, or UPS, provides near-instant power and power conditioning when utility service is interrupted or disturbed. UPS systems bridge the time until generators or other sources respond and protect sensitive equipment, but their energy capacity, redundancy, battery condition, controls, fault behavior, and maintenance impose limits. **Related definitions:** [Thermal storage](#entry%5FKB213) · [Backup batteries](#entry%5FKB215) · [Diesel backup](#entry%5FKB095) · [Customer obligations](#entry%5FKB179) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-214 ### Backup batteries How should developers value Backup batteries? Backup batteries provide short-duration stored energy for continuity during power interruptions, equipment transfers, or generator startup. Developers should value them for avoided outages, power quality, operational flexibility, and possible grid services while accounting for degradation, replacement, fire safety, space, controls, warranty, and the need to preserve emergency reserve. **Related definitions:** [UPS systems](#entry%5FKB214) · [Storage duration](#entry%5FKB216) · [Diesel backup](#entry%5FKB095) · [Critical load](#entry%5FKB162) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-215 ### Storage duration What is Storage duration? Storage duration is the length of time a storage system can discharge at a specified power level before its usable energy is depleted. Duration links power capacity to energy capacity and should be evaluated with efficiency, minimum charge, degradation, recharge time, duty cycle, and the event the system is intended to cover. **Related definitions:** [Backup batteries](#entry%5FKB215) · [Grid services](#entry%5FKB217) · [Minimum demand charges](#entry%5FKB169) · [Diesel backup](#entry%5FKB095) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-216 ### Grid services Can Grid services support AI data centers? Grid services are functions that help operate the power system, such as frequency regulation, reserves, voltage support, peak reduction, congestion relief, and capacity. Storage can provide several services rapidly, but value depends on location, market access, telemetry, performance rules, state of charge, interconnection, and stacking limits. **Related definitions:** [Storage duration](#entry%5FKB216) · [Demand charge management](#entry%5FKB218) · [Ancillary services](#entry%5FKB079) · [Hydropower](#entry%5FKB094) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-217 ### Demand charge management How does Demand charge management affect grid flexibility? Demand-charge management uses operational changes or storage to reduce a customer’s measured peak demand and associated charges. It can improve project economics and lower coincident system stress, but savings depend on tariff design, peak predictability, battery losses and degradation, operational constraints, and whether new peaks are created at other times. **Related definitions:** [Grid services](#entry%5FKB217) · [Co-located storage](#entry%5FKB219) · [Minimum demand charges](#entry%5FKB169) · [On-site generation](#entry%5FKB096) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-218 ### Co-located storage What are the limits of Co-located storage? Co-located storage is installed at the same site or electrical point as generation or load. It can share infrastructure, smooth output, manage peaks, and support resilience, but may face metering, charging-source, market, interconnection, tax, operational, and contractual limits that depend on its configuration. **Related definitions:** [Demand charge management](#entry%5FKB218) · [Storage interconnection](#entry%5FKB220) · [Co-located generation](#entry%5FKB097) · [High load factor customers](#entry%5FKB176) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-219 ### Storage interconnection How should developers value Storage interconnection? Storage interconnection is the process for connecting a battery or other storage system to the grid and defining its charging, discharging, protection, control, and market behavior. Studies must consider both load and generation modes, simultaneous operating limits, fault contribution, inverter controls, network upgrades, and telemetry. **Related definitions:** [Co-located storage](#entry%5FKB219) · [Battery safety](#entry%5FKB221) · [Generation interconnection](#entry%5FKB104) · [Demand response](#entry%5FKB084) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-220 ### Battery safety What is Battery safety? Battery safety is the management of electrical, chemical, thermal, fire, gas, mechanical, and operational hazards across design, transport, installation, use, maintenance, and end of life. It relies on tested equipment, codes, spacing, detection, ventilation, suppression, emergency planning, controls, training, and incident reporting. **Related definitions:** [Storage interconnection](#entry%5FKB220) · [Storage economics](#entry%5FKB222) · [Load shedding](#entry%5FKB157) · [Hydropower](#entry%5FKB094) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-221 ### Storage economics Can Storage economics support AI data centers? Storage economics compare capital, operating, replacement, financing, degradation, efficiency, and interconnection costs with the value of energy shifting, capacity, grid services, resilience, and avoided infrastructure. A project can appear profitable only if revenue streams, dispatch constraints, warranties, and market rules are modeled consistently. **Related definitions:** [Battery safety](#entry%5FKB221) · [Resilience value](#entry%5FKB223) · [Wholesale electricity markets](#entry%5FKB076) · [Co-located generation](#entry%5FKB097) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-222 ### Resilience value How does Resilience value affect grid flexibility? Resilience value is the benefit of maintaining or restoring critical service during disruptions. For storage, it depends on outage probability and duration, protected load, islanding capability, recharge, fuel alternatives, customer outage costs, and whether the system is available when the event occurs rather than already committed to another service. **Related definitions:** [Storage economics](#entry%5FKB222) · [Storage dispatch](#entry%5FKB224) · [Non-firm service](#entry%5FKB172) · [Critical load](#entry%5FKB162) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-223 ### Storage dispatch What are the limits of Storage dispatch? Storage dispatch is the decision of when and how strongly to charge or discharge a storage resource. Limits include state of charge, duration, efficiency, degradation, warranty, network constraints, market commitments, backup reserve, forecasts, controls, and the risk that optimizing one service reduces availability for another. **Related definitions:** [Resilience value](#entry%5FKB223) · [Hybrid resources](#entry%5FKB225) · [Co-located generation](#entry%5FKB097) · [Planning reserve margins](#entry%5FKB159) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-224 ### Hybrid resources How should developers value Hybrid resources? A hybrid resource combines two or more technologies, such as solar and storage, at one site or interconnection point. Hybrids can share equipment and shape output, but developers must value charging rules, capacity accreditation, controls, interconnection limits, tax treatment, degradation, market participation, and operational coordination. **Related definitions:** [Storage dispatch](#entry%5FKB224) · [Renewables](#entry%5FKB093) · [Wholesale electricity markets](#entry%5FKB076) · [Reliability-must-run resources](#entry%5FKB163) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-225 SECTION 16 | 15 DEFINITIONS ## Regulation and Policy The federal, state, regional, and local institutions and rules that govern AI infrastructure, utilities, tariffs, siting, and customer protection. **Terms in this section:** [FERC jurisdiction](#entry%5FKB226) · [State public utility commissions](#entry%5FKB227) · [NERC oversight](#entry%5FKB228) · [RTO and ISO tariffs](#entry%5FKB229) · [Large-load policy](#entry%5FKB230) · [Cost allocation policy](#entry%5FKB231) · [Ratepayer protection](#entry%5FKB232)· [Environmental permitting](#entry%5FKB233) · [Moratoria](#entry%5FKB234) · [Economic development policy](#entry%5FKB235) · [Federal-state coordination](#entry%5FKB236) · [Transmission reform](#entry%5FKB237) · [Data center policy](#entry%5FKB238) · [Tariff design](#entry%5FKB239) · [Regulatory compact](#entry%5FKB240) ### FERC jurisdiction What is FERC jurisdiction? FERC jurisdiction in electricity generally covers interstate transmission, wholesale sales, organized wholesale markets, and specified reliability and hydropower matters under federal law. State and local authorities retain major roles over retail rates, distribution, siting, generation approvals, and utility service, so large-load issues often require coordinated federal-state analysis. **Related definitions:** [State public utility commissions](#entry%5FKB227) · [Large-load service](#entry%5FKB069) · [Contingency analysis](#entry%5FKB161) · [Cooperatives](#entry%5FKB063) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-226 ### State public utility commissions Why does State public utility commissions matter for AI infrastructure? State public utility commissions regulate investor-owned utilities and, depending on state law, may oversee rates, resource planning, service quality, facility approvals, customer protections, mergers, and other matters. Their authority over public power, cooperatives, siting, retail choice, and large-load contracts varies by jurisdiction. **Related definitions:** [FERC jurisdiction](#entry%5FKB226) · [NERC oversight](#entry%5FKB228) · [Public power](#entry%5FKB062) · [Customer obligations](#entry%5FKB179) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-227 ### NERC oversight How does NERC oversight affect utilities and customers? NERC oversight is the development, approval, monitoring, and enforcement of reliability requirements for the North American bulk power system through NERC, regional entities, regulators, and registered organizations. It affects utilities and customers when facilities, models, operations, cybersecurity, or disturbances fall within reliability responsibilities. **Related definitions:** [State public utility commissions](#entry%5FKB227) · [RTO and ISO tariffs](#entry%5FKB229) · [NERC reliability standards](#entry%5FKB151) · [Public power](#entry%5FKB062) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-228 ### RTO and ISO tariffs What questions should regulators ask about RTO and ISO tariffs? RTO and ISO tariffs are publicly filed rules governing transmission service, organized markets, interconnection, planning, settlements, credit, and operating requirements. Regulators should ask whether large-load provisions are transparent, nondiscriminatory, technically justified, aligned with reliability, and capable of allocating costs and uncertainty without shifting unreasonable risk. **Related definitions:** [NERC oversight](#entry%5FKB228) · [Large-load policy](#entry%5FKB230) · [Large-load tariffs](#entry%5FKB166) · [Large-load interconnection](#entry%5FKB136) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-229 ### Large-load policy How could Large-load policy evolve? Large-load policy is the set of laws, tariffs, planning rules, interconnection procedures, customer protections, development incentives, and operating requirements applied to major electricity users. It is likely to evolve toward stronger commitment tests, phased service, financial assurance, transparency, flexibility standards, and coordination among utilities, regulators, and system operators. **Related definitions:** [RTO and ISO tariffs](#entry%5FKB229) · [Cost allocation policy](#entry%5FKB231) · [Large-load interconnection](#entry%5FKB136) · [Reliability accountability](#entry%5FKB165) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-230 ### Cost allocation policy What is Cost allocation policy? Cost-allocation policy establishes principles and methods for assigning infrastructure and service costs among customers, utilities, regions, and beneficiaries. Effective policy distinguishes customer-specific facilities from shared network benefits, addresses uncertainty and stranded assets, and aligns payment responsibility with both causation and demonstrated value. **Related definitions:** [Large-load policy](#entry%5FKB230) · [Ratepayer protection](#entry%5FKB232) · [Cost causation](#entry%5FKB178) · [Network upgrades](#entry%5FKB143) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-231 ### Ratepayer protection Why does Ratepayer protection matter for AI infrastructure? Ratepayer protection is the regulatory objective of preventing existing customers from bearing unreasonable costs or risks created by utility decisions or new customers. For AI infrastructure, it includes credible forecasts, cost causation, financial assurance, transparent contracts, prudent investment, performance monitoring, and remedies when projects fail to materialize. **Related definitions:** [Cost allocation policy](#entry%5FKB231) · [Environmental permitting](#entry%5FKB233) · [Customer protection](#entry%5FKB075) · [Cost causation](#entry%5FKB178) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-232 ### Environmental permitting How does Environmental permitting affect utilities and customers? Environmental permitting applies laws and approvals concerning air, water, wetlands, waste, endangered species, noise, land disturbance, emissions, and other impacts. It affects utilities and customers by shaping project design, location, schedule, mitigation, operating limits, community obligations, and the viability of backup or onsite generation. **Related definitions:** [Ratepayer protection](#entry%5FKB232) · [Moratoria](#entry%5FKB234) · [Large-load service](#entry%5FKB069) · [Phased energization](#entry%5FKB150) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-233 ### Moratoria What questions should regulators ask about Moratoria? A moratorium is a temporary pause or restriction on specified approvals, connections, construction, or development. Regulators should ask what problem it addresses, what projects are covered, how long it lasts, whether exemptions are justified, what evidence will end it, and whether targeted tariffs or standards would manage risk more effectively. **Related definitions:** [Environmental permitting](#entry%5FKB233) · [Economic development policy](#entry%5FKB235) · [Utility governance](#entry%5FKB074) · [Generator interconnection](#entry%5FKB137) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-234 ### Economic development policy How could Economic development policy evolve? Economic-development policy uses incentives, infrastructure support, taxes, grants, workforce programs, or permitting tools to attract investment and jobs. For data centers, policy is likely to evolve toward clearer tests for grid cost, tax value, employment, water, community benefits, emissions, project credibility, and enforceable performance. **Related definitions:** [Moratoria](#entry%5FKB234) · [Federal-state coordination](#entry%5FKB236) · [Cost recovery](#entry%5FKB066) · [Non-firm service](#entry%5FKB172) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-235 ### Federal-state coordination What is Federal-state coordination? Federal-state coordination is the alignment of authorities, data, schedules, and decisions across federal agencies, state commissions, environmental regulators, system operators, utilities, and local governments. It is necessary because transmission, retail service, generation, reliability, siting, water, and economic development often fall under different legal institutions. **Related definitions:** [Economic development policy](#entry%5FKB235) · [Transmission reform](#entry%5FKB237)· [Reliability-must-run resources](#entry%5FKB163) · [Coincident peak](#entry%5FKB177) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-236 ### Transmission reform Why does Transmission reform matter for AI infrastructure? Transmission reform changes planning, permitting, interconnection, cost allocation, technology use, or competition to deliver network capacity more effectively. It matters for AI infrastructure because concentrated load can develop faster than traditional transmission processes, while poorly designed reform can create cost shifts or weaken planning discipline. **Related definitions:** [Federal-state coordination](#entry%5FKB236) · [Data center policy](#entry%5FKB238) · [Network upgrades](#entry%5FKB143) · [Utility procurement](#entry%5FKB073) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-237 ### Data center policy How does Data center policy affect utilities and customers? Data-center policy is the body of government and utility rules addressing siting, taxation, power, water, emissions, labor, reporting, reliability, and community impacts of data centers. It affects customers and utilities by determining development conditions, infrastructure responsibilities, disclosure, incentives, and the balance between economic growth and public risk. **Related definitions:** [Transmission reform](#entry%5FKB237) · [Tariff design](#entry%5FKB239) · [Public power](#entry%5FKB062) · [Emergency operations](#entry%5FKB156) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-238 ### Tariff design What questions should regulators ask about Tariff design? Tariff design converts regulatory principles into specific rates, eligibility rules, contract terms, service classes, and customer obligations. Regulators should ask whether a tariff reflects cost causation, treats similar customers consistently, protects reliability, creates enforceable incentives, addresses project failure, and remains administratively workable. **Related definitions:** [Data center policy](#entry%5FKB238) · [Regulatory compact](#entry%5FKB240) · [Customer protection](#entry%5FKB075) · [Large-load tariffs](#entry%5FKB166) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-239 ### Regulatory compact How could Regulatory compact evolve? The regulatory compact is the broad arrangement under which utilities receive defined rights, service territories, and cost-recovery opportunities in exchange for obligations to serve, invest prudently, and submit to public oversight. It may evolve as large flexible customers, private generation, competitive infrastructure, and new technology blur traditional roles. **Related definitions:** [Tariff design](#entry%5FKB239) · [Investor-owned utilities](#entry%5FKB061) · [Large-load interconnection](#entry%5FKB136) · [Cost recovery](#entry%5FKB066) > Last reviewed: 2026-07-11 | Review frequency: Quarterly | Entry ID: KB-240 SECTION 17 | 15 DEFINITIONS ## AI Governance The standards, controls, oversight practices, and evidence needed to deploy artificial intelligence responsibly in critical energy infrastructure. **Terms in this section:** [Utility AI governance](#entry%5FKB241) · [NIST AI RMF](#entry%5FKB242) · [ISO 42001](#entry%5FKB243) · [Model risk management](#entry%5FKB244) · [Cybersecurity](#entry%5FKB245) · [Operational AI](#entry%5FKB246) · [Human oversight](#entry%5FKB247) · [AI auditability](#entry%5FKB248) · [Explainability](#entry%5FKB249) · [Data governance](#entry%5FKB250) · [Critical infrastructure AI](#entry%5FKB251) · [Vendor governance](#entry%5FKB252) · [AI incident response](#entry%5FKB253) · [Algorithmic accountability](#entry%5FKB254) · [Trustworthy AI](#entry%5FKB255) ### Utility AI governance What is Utility AI governance? Utility AI governance is the system of authority, policies, controls, roles, evidence, and oversight used to manage AI across a utility. It covers use-case approval, data, model risk, cybersecurity, procurement, testing, human review, operations, monitoring, incident response, documentation, and accountability for decisions that affect customers or reliability. **Related definitions:** [NIST AI RMF](#entry%5FKB242) · [Utility AI adoption](#entry%5FKB071) · [NERC oversight](#entry%5FKB228) · [AI safety](#entry%5FKB027) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-241 ### NIST AI RMF Why does NIST AI RMF matter for utilities? The NIST AI Risk Management Framework is voluntary guidance for identifying and managing risks to trustworthy AI across the system life cycle. Its core functions—Govern, Map, Measure, and Manage—help utilities connect enterprise governance with context, evaluation, controls, monitoring, and action rather than treating responsible AI as a checklist. **Related definitions:** [Utility AI governance](#entry%5FKB241) · [ISO 42001](#entry%5FKB243) · [Enterprise AI adoption](#entry%5FKB026) · [Utility governance](#entry%5FKB074) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-242 ### ISO 42001 How should energy companies govern ISO 42001? ISO/IEC 42001 is an international management-system standard for establishing, implementing, maintaining, and continually improving organizational governance of AI. Energy companies can use it to define scope, leadership, risk and impact processes, objectives, controls, supplier management, documentation, monitoring, audit, and corrective action. **Related definitions:** [NIST AI RMF](#entry%5FKB242) · [Model risk management](#entry%5FKB244) · [Utility risk management](#entry%5FKB070) · [AI operations](#entry%5FKB030) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-243 ### Model risk management What standards apply to Model risk management? Model risk management is the disciplined control of risks created by models that are incorrect, misused, poorly understood, outdated, or applied outside their intended context. Standards and practices typically require inventory, ownership, documentation, independent validation, change control, performance monitoring, limitations, human oversight, and escalation. **Related definitions:** [ISO 42001](#entry%5FKB243) · [Cybersecurity](#entry%5FKB245) · [AI operations](#entry%5FKB030) · [Utility risk management](#entry%5FKB070) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-244 ### Cybersecurity What evidence should executives require for Cybersecurity? Cybersecurity is the protection of systems, networks, software, data, identities, and operations from unauthorized access, disruption, manipulation, or destruction. Executives should require architecture, threat models, access controls, secure development, testing, logging, monitoring, incident response, vendor assurance, recovery plans, and evidence that controls operate in practice. **Related definitions:** [Model risk management](#entry%5FKB244) · [Operational AI](#entry%5FKB246) · [AI safety](#entry%5FKB027) · [Utility AI adoption](#entry%5FKB071) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-245 ### Operational AI What is Operational AI? Operational AI is artificial intelligence used within ongoing business or physical-system workflows rather than isolated experimentation. In energy settings it may influence forecasts, maintenance, dispatch support, customer actions, or engineering analysis, so deployment requires validated interfaces, operating limits, fallback, traceability, monitoring, and accountable human decision rights. **Related definitions:** [Cybersecurity](#entry%5FKB245) · [Human oversight](#entry%5FKB247) · [Utility AI adoption](#entry%5FKB071) · [Enterprise AI adoption](#entry%5FKB026) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-246 ### Human oversight Why does Human oversight matter for utilities? Human oversight is the deliberate role assigned to people in supervising, approving, challenging, interrupting, or correcting AI-supported activity. It matters only when reviewers have authority, time, competence, relevant information, and usable controls; a nominal approval step does not provide meaningful oversight if automation bias or workflow design makes intervention unrealistic. **Related definitions:** [Operational AI](#entry%5FKB246) · [AI auditability](#entry%5FKB248) · [AI safety](#entry%5FKB027) · [Reliability entities](#entry%5FKB158) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-247 ### AI auditability How should energy companies govern AI auditability? AI auditability is the ability to reconstruct how an AI system was designed, changed, used, and monitored and how a particular output or decision was produced. It relies on documentation, versioning, data lineage, logs, evaluation records, access histories, approvals, and retention practices appropriate to the consequence of the use case. **Related definitions:** [Human oversight](#entry%5FKB247) · [Explainability](#entry%5FKB249) · [Enterprise AI adoption](#entry%5FKB026) · [Resource adequacy](#entry%5FKB152) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-248 ### Explainability What standards apply to Explainability? Explainability is the degree to which people can understand the factors, logic, limitations, or evidence behind an AI output. The appropriate method depends on the audience and decision: engineers, operators, customers, auditors, and regulators may need different explanations, and understandable output does not by itself prove accuracy or fairness. **Related definitions:** [AI auditability](#entry%5FKB248) · [Data governance](#entry%5FKB250) · [Reliability accountability](#entry%5FKB165) · [Moratoria](#entry%5FKB234) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-249 ### Data governance What evidence should executives require for Data governance? Data governance is the system for assigning ownership and controlling the quality, access, security, lineage, retention, use, sharing, and disposal of data. Executives should require evidence that data are lawful, relevant, representative, protected, documented, monitored, and suitable for the AI system’s intended operating context. **Related definitions:** [Explainability](#entry%5FKB249) · [Critical infrastructure AI](#entry%5FKB251) · [Machine learning](#entry%5FKB016) · [Adequacy vs security](#entry%5FKB164) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-250 ### Critical infrastructure AI What is Critical infrastructure AI? Critical-infrastructure AI is AI used in or affecting systems whose disruption could produce significant public safety, security, economic, or societal consequences. It demands stronger engineering assurance, cybersecurity, operational boundaries, monitoring, human control, incident coordination, recovery, and evidence than low-consequence consumer applications. **Related definitions:** [Data governance](#entry%5FKB250) · [Vendor governance](#entry%5FKB252) · [Critical load](#entry%5FKB162) · [Utility AI adoption](#entry%5FKB071) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-251 ### Vendor governance Why does Vendor governance matter for utilities? Vendor governance is the oversight of third parties that provide AI models, software, data, infrastructure, or services. It includes due diligence, security, performance requirements, data rights, subcontractors, change notification, audit access, incident reporting, service continuity, intellectual property, regulatory support, exit planning, and accountability for integrated outcomes. **Related definitions:** [Critical infrastructure AI](#entry%5FKB251) · [AI incident response](#entry%5FKB253)· [Open models](#entry%5FKB025) · [Reliability accountability](#entry%5FKB165) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-252 ### AI incident response How should energy companies govern AI incident response? AI incident response is the organized process for detecting, containing, investigating, correcting, reporting, and learning from harmful or unexpected AI behavior. Plans should define triggers, owners, logging, evidence preservation, operational fallback, customer and regulator communications, vendor coordination, corrective action, and criteria for returning a system to service. **Related definitions:** [Vendor governance](#entry%5FKB252) · [Algorithmic accountability](#entry%5FKB254) · [Utility AI adoption](#entry%5FKB071) · [Load shedding](#entry%5FKB157) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-253 ### Algorithmic accountability What standards apply to Algorithmic accountability? Algorithmic accountability is the assignment of responsibility for the design, approval, use, monitoring, and consequences of automated systems. It requires identifiable owners, documented purpose, impact assessment, contestability, oversight, records, performance evidence, incident processes, and remedies rather than allowing responsibility to disappear among developers, vendors, and users. **Related definitions:** [AI incident response](#entry%5FKB253) · [Trustworthy AI](#entry%5FKB255) · [Reliability accountability](#entry%5FKB165) · [AI safety](#entry%5FKB027) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-254 ### Trustworthy AI What evidence should executives require for Trustworthy AI? Trustworthy AI is AI that demonstrates appropriate validity, reliability, safety, security, resilience, accountability, transparency, explainability, privacy, and fairness for its context. Executives should require documented evidence, testing, controls, monitoring, and ownership proportional to consequence rather than accepting broad vendor claims or model benchmarks alone. **Related definitions:** [Algorithmic accountability](#entry%5FKB254) · [AI safety](#entry%5FKB027) · [Utility risk management](#entry%5FKB070) · [Frequency response](#entry%5FKB155) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-255 SECTION 18 | 15 DEFINITIONS ## Climate and Sustainability The emissions, water, procurement, resilience, and community questions that determine the environmental performance of AI infrastructure. **Terms in this section:** [Carbon accounting](#entry%5FKB256) · [24/7 clean energy](#entry%5FKB257) · [Renewable matching](#entry%5FKB258) · [Emissions impact](#entry%5FKB259) · [Grid emissions](#entry%5FKB260) · [Scope 2 emissions](#entry%5FKB261) · [Clean power procurement](#entry%5FKB262) · [Additionality](#entry%5FKB263) · [Sustainability claims](#entry%5FKB264) · [Water sustainability](#entry%5FKB265) · [Circular economy](#entry%5FKB266) · [Embodied carbon](#entry%5FKB267) · [Community impacts](#entry%5FKB268) · [Environmental justice](#entry%5FKB269) · [Climate resilience](#entry%5FKB270) ### Carbon accounting What is Carbon accounting? Carbon accounting is the measurement and reporting of greenhouse-gas emissions associated with an organization, activity, product, or asset. For AI infrastructure it may include direct fuel use, purchased electricity, construction, equipment, supply chains, and other value-chain emissions, with results depending on boundaries, methods, factors, time, and attribution. **Related definitions:** [24/7 clean energy](#entry%5FKB257) · [Nuclear power](#entry%5FKB092) · [Water usage effectiveness](#entry%5FKB200) · [UPS systems](#entry%5FKB214) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-256 ### 24/7 clean energy How does 24/7 clean energy apply to AI infrastructure? Twenty-four-seven clean energy is the objective of matching electricity consumption with carbon-free supply in the same location and time interval, rather than only purchasing an equal annual quantity. It requires granular load and supply data, contracts, market rules, transmission awareness, and treatment of hours when clean resources are scarce. **Related definitions:** [Carbon accounting](#entry%5FKB256) · [Renewable matching](#entry%5FKB258) · [Renewables](#entry%5FKB093) · [Hybrid resources](#entry%5FKB225) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-257 ### Renewable matching What evidence is needed to support claims about Renewable matching? Renewable matching links electricity consumption to renewable-energy generation or environmental attributes over a defined period and geography. Credible claims require transparent boundaries, ownership of certificates, avoidance of double counting, suitable time and location matching, treatment of residual supply, and clarity about whether the claim concerns procurement or physical emissions impact. **Related definitions:** [24/7 clean energy](#entry%5FKB257) · [Emissions impact](#entry%5FKB259) · [Environmental permitting](#entry%5FKB233) · [Renewables](#entry%5FKB093) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-258 ### Emissions impact What tradeoffs does Emissions impact create? Emissions impact is the change in greenhouse-gas or air-pollutant emissions caused by an activity relative to a defined baseline. Tradeoffs arise because annual contractual matching, local generation, marginal grid effects, backup power, construction, and supply chains can point in different directions depending on time, place, and methodology. **Related definitions:** [Renewable matching](#entry%5FKB258) · [Grid emissions](#entry%5FKB260) · [Nuclear power](#entry%5FKB092) · [Water stress](#entry%5FKB201) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-259 ### Grid emissions How should companies report Grid emissions? Grid emissions are emissions associated with electricity supplied by the power system. Companies may report average, contractual, residual-mix, or marginal measures, each answering a different question; credible reporting states the method, geography, time period, data source, market instruments, and limitations. **Related definitions:** [Emissions impact](#entry%5FKB259) · [Scope 2 emissions](#entry%5FKB261) · [Co-located storage](#entry%5FKB219) · [Co-located generation](#entry%5FKB097) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-260 ### Scope 2 emissions What is Scope 2 emissions? Scope 2 emissions are indirect greenhouse-gas emissions associated with purchased or acquired electricity, steam, heat, or cooling. Reporting frameworks commonly distinguish location-based and market-based methods, and organizations should disclose contractual instruments, residual mixes, boundaries, and material differences between procurement claims and physical system effects. **Related definitions:** [Grid emissions](#entry%5FKB260) · [Clean power procurement](#entry%5FKB262) · [Evaporative cooling](#entry%5FKB202) · [Data center policy](#entry%5FKB238) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-261 ### Clean power procurement How does Clean power procurement apply to AI infrastructure? Clean-power procurement is the acquisition of electricity or environmental attributes from low- or zero-carbon resources through utility tariffs, power purchase agreements, certificates, ownership, or other structures. Its effect depends on additionality, timing, location, deliverability, contract duration, risk allocation, and what happens when contracted resources are unavailable. **Related definitions:** [Scope 2 emissions](#entry%5FKB261) · [Additionality](#entry%5FKB263) · [Clean firm power](#entry%5FKB099) · [Storage duration](#entry%5FKB216) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-262 ### Additionality What evidence is needed to support claims about Additionality? Additionality is the degree to which a procurement action causes new clean generation or emissions reductions beyond what would otherwise occur. Evidence can include project financing dependence, contract duration, development timing, market conditions, policy treatment, and a transparent counterfactual, but additionality is rarely established by certificate ownership alone. **Related definitions:** [Clean power procurement](#entry%5FKB262) · [Sustainability claims](#entry%5FKB264)· [Clean firm power](#entry%5FKB099) · [Economic development policy](#entry%5FKB235) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-263 ### Sustainability claims What tradeoffs does Sustainability claims create? Sustainability claims describe environmental or social performance, such as renewable use, carbon neutrality, water stewardship, circularity, or community benefit. Tradeoffs arise when a claim emphasizes one metric while overlooking time, location, grid constraints, supply chains, land, water, air quality, or the difference between accounting and physical impact. **Related definitions:** [Additionality](#entry%5FKB263) · [Water sustainability](#entry%5FKB265) · [Water stress](#entry%5FKB201) · [Environmental permitting](#entry%5FKB233) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-264 ### Water sustainability How should companies report Water sustainability? Water sustainability is the responsible management of water withdrawal, consumption, quality, discharge, ecosystem needs, and community access over time. Companies should report local basin conditions, sources, seasonal use, cooling design, recycled water, drought plans, wastewater, indirect water effects, and progress against context-based targets. **Related definitions:** [Sustainability claims](#entry%5FKB264) · [Circular economy](#entry%5FKB266) · [Evaporative cooling](#entry%5FKB202) · [Data center policy](#entry%5FKB238) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-265 ### Circular economy What is Circular economy? A circular economy seeks to reduce extraction and waste by keeping products, components, and materials in productive use through durability, repair, reuse, refurbishment, remanufacturing, and recycling. For AI infrastructure, important issues include servers, chips, batteries, cooling equipment, construction materials, hazardous substances, and secure data destruction. **Related definitions:** [Water sustainability](#entry%5FKB265) · [Embodied carbon](#entry%5FKB267) · [Waste heat reuse](#entry%5FKB204) · [Thermal storage](#entry%5FKB213) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-266 ### Embodied carbon How does Embodied carbon apply to AI infrastructure? Embodied carbon is the greenhouse-gas emissions associated with producing, transporting, constructing, maintaining, and disposing of materials and equipment rather than operating them. Data centers have embodied impacts in concrete, steel, chips, batteries, generators, transformers, and repeated hardware replacement, which can be significant even when electricity is low carbon. **Related definitions:** [Circular economy](#entry%5FKB266) · [Community impacts](#entry%5FKB268) · [Clean firm power](#entry%5FKB099) · [Backup batteries](#entry%5FKB215) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-267 ### Community impacts What evidence is needed to support claims about Community impacts? Community impacts are the local economic, environmental, infrastructure, health, land, water, noise, traffic, tax, and service effects of a project. Evidence should include distribution of benefits and burdens, baseline conditions, cumulative development, enforceable commitments, public participation, monitoring, and remedies rather than relying only on regional investment totals. **Related definitions:** [Embodied carbon](#entry%5FKB267) · [Environmental justice](#entry%5FKB269) · [Economic development policy](#entry%5FKB235) · [Generation siting](#entry%5FKB105) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-268 ### Environmental justice What tradeoffs does Environmental justice create? Environmental justice examines whether environmental harms, risks, participation, and benefits are distributed fairly, particularly for communities that have historically faced disproportionate burdens or limited influence. Tradeoffs arise when infrastructure supports broad economic goals but concentrates pollution, land use, water demand, cost, or reliability risk locally. **Related definitions:** [Community impacts](#entry%5FKB268) · [Climate resilience](#entry%5FKB270) · [Recycled water](#entry%5FKB208) · [Data center policy](#entry%5FKB238) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-269 ### Climate resilience How should companies report Climate resilience? Climate resilience is the ability of an organization or infrastructure system to anticipate, withstand, adapt to, and recover from climate-related hazards. Companies should report scenario assumptions, asset exposure, interdependencies, adaptation measures, investment plans, residual risk, governance, and whether resilience actions protect surrounding systems and communities. **Related definitions:** [Environmental justice](#entry%5FKB269) · [Climate risk](#entry%5FKB210) · [Storage dispatch](#entry%5FKB224) · [Evaporative cooling](#entry%5FKB202) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-270 SECTION 19 | 15 DEFINITIONS ## Supply Chains and Finance The equipment, labor, capital, insurance, credit, and delivery risks that determine whether large infrastructure commitments are credible. **Terms in this section:** [Transformer supply chains](#entry%5FKB271) · [Turbine supply chains](#entry%5FKB272) · [Critical minerals](#entry%5FKB273) · [Construction labor](#entry%5FKB274) · [Capital expenditure](#entry%5FKB275) · [Project finance](#entry%5FKB276) · [Power procurement risk](#entry%5FKB277) · [Infrastructure insurance](#entry%5FKB278) · [Equipment lead times](#entry%5FKB279) · [Permitting risk](#entry%5FKB280) · [Data center financing](#entry%5FKB281) · [Utility balance sheets](#entry%5FKB282)· [Stranded costs](#entry%5FKB283) · [Credit support](#entry%5FKB284) · [Capital discipline](#entry%5FKB285) ### Transformer supply chains What is Transformer supply chains? Transformer supply chains encompass design, electrical steel, copper, insulation, bushings, tap changers, factories, testing, transport, installation, and skilled labor. Constraints arise because transformers are customized, large, difficult to move, and essential across generation, transmission, distribution, and customer projects, making factory capacity and early specification critical. **Related definitions:** [Turbine supply chains](#entry%5FKB272) · [Chip supply chains](#entry%5FKB189) · [Transformer lead times](#entry%5FKB112) · [Digital infrastructure supply chain](#entry%5FKB012) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-271 ### Turbine supply chains Why does Turbine supply chains matter for AI infrastructure delivery? Turbine supply chains include manufacturers, castings, forgings, blades, controls, generators, heat-recovery equipment, service networks, and specialized labor for gas, steam, wind, and hydro machines. Long lead times matter because equipment slots, design changes, construction sequencing, and service agreements can determine when new generation becomes available. **Related definitions:** [Transformer supply chains](#entry%5FKB271) · [Critical minerals](#entry%5FKB273) · [Chip supply chains](#entry%5FKB189) · [Transformer lead times](#entry%5FKB112) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-272 ### Critical minerals How does Critical minerals affect project risk? Critical minerals are materials whose economic or security importance is high and whose supply is vulnerable to disruption. They affect project risk through mining concentration, refining, trade restrictions, permitting, price volatility, environmental impacts, recycling, substitution, and demand from competing sectors. **Related definitions:** [Turbine supply chains](#entry%5FKB272) · [Construction labor](#entry%5FKB274) · [High-voltage lines](#entry%5FKB110) · [Chip supply chains](#entry%5FKB189) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-273 ### Construction labor What evidence should investors require for Construction labor? Construction labor is the skilled and general workforce required to build facilities, grid upgrades, generation, pipelines, cooling, and communications. Investors should require regional labor assessments, contractor capacity, wage and productivity assumptions, training plans, schedule integration, safety performance, and evidence that simultaneous projects will not exceed available crews. **Related definitions:** [Critical minerals](#entry%5FKB273) · [Capital expenditure](#entry%5FKB275) · [Nuclear power](#entry%5FKB092) · [Digital infrastructure supply chain](#entry%5FKB012) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-274 ### Capital expenditure How should developers manage Capital expenditure? Capital expenditure is money invested in long-lived assets such as buildings, servers, substations, transmission, generation, cooling, and equipment. Developers manage it through scope discipline, stage gates, contingencies, procurement strategy, schedule control, financing, standardization, and clear separation of committed spending from optional future phases. **Related definitions:** [Construction labor](#entry%5FKB274) · [Project finance](#entry%5FKB276) · [Competitive transmission](#entry%5FKB108) · [AI infrastructure economics](#entry%5FKB004) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-275 ### Project finance What is Project finance? Project finance funds an asset primarily on the strength of its expected cash flows, contracts, and risk allocation rather than the general credit of a sponsor alone. Lenders examine construction, technology, power, customer, market, regulatory, operating, insurance, and counterparty risks and impose covenants, reserves, and security accordingly. **Related definitions:** [Capital expenditure](#entry%5FKB275) · [Power procurement risk](#entry%5FKB277) · [AI infrastructure economics](#entry%5FKB004) · [Advanced nodes](#entry%5FKB188) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-276 ### Power procurement risk Why does Power procurement risk matter for AI infrastructure delivery? Power-procurement risk is the possibility that electricity cost, availability, carbon attributes, location, timing, or contract performance differs from project needs. It matters because a data center may commit to long-lived facilities while generation, market prices, grid congestion, regulation, and computing demand remain uncertain. **Related definitions:** [Project finance](#entry%5FKB276) · [Infrastructure insurance](#entry%5FKB278) · [AI infrastructure economics](#entry%5FKB004) · [Nuclear power](#entry%5FKB092) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-277 ### Infrastructure insurance How does Infrastructure insurance affect project risk? Infrastructure insurance transfers specified construction, property, liability, business-interruption, cyber, environmental, and other risks to insurers subject to limits, exclusions, deductibles, and conditions. Insurance affects project risk through coverage availability, pricing, engineering requirements, claims history, catastrophe exposure, and whether losses are actually insurable. **Related definitions:** [Power procurement risk](#entry%5FKB277) · [Equipment lead times](#entry%5FKB279) · [Firm power](#entry%5FKB098) · [High-voltage lines](#entry%5FKB110) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-278 ### Equipment lead times What evidence should investors require for Equipment lead times? Equipment lead time is the period from specification and order through manufacturing, testing, transport, installation, and commissioning. Investors should require supplier quotations, factory-slot evidence, design maturity, logistics plans, alternatives, escalation terms, and schedule contingency rather than relying on generic industry estimates. **Related definitions:** [Infrastructure insurance](#entry%5FKB278) · [Permitting risk](#entry%5FKB280) · [Transformer lead times](#entry%5FKB112) · [Chip supply chains](#entry%5FKB189) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-279 ### Permitting risk How should developers manage Permitting risk? Permitting risk is the possibility that approvals are delayed, denied, conditioned, challenged, or made more costly. Developers manage it through jurisdictional mapping, early engagement, complete applications, realistic schedules, site alternatives, environmental and community analysis, legal strategy, design flexibility, and avoidance of premature construction commitments. **Related definitions:** [Equipment lead times](#entry%5FKB279) · [Data center financing](#entry%5FKB281) · [Generation siting](#entry%5FKB105) · [High-voltage lines](#entry%5FKB110) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-280 ### Data center financing What is Data center financing? Data-center financing includes corporate capital, debt, project finance, real-estate structures, joint ventures, leases, securitization, and customer precommitments used to fund facilities and equipment. Financing depends on tenant credit, utilization, power availability, construction cost, technology cycles, residual value, contracts, and the separation between real estate and computing assets. **Related definitions:** [Permitting risk](#entry%5FKB280) · [Utility balance sheets](#entry%5FKB282) · [AI infrastructure economics](#entry%5FKB004) · [Critical path management](#entry%5FKB120) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-281 ### Utility balance sheets Why does Utility balance sheets matter for AI infrastructure delivery? Utility balance sheets record the assets, liabilities, equity, and financing capacity used to support utility investment. Rapid AI infrastructure delivery can require substantial capital, affecting credit metrics, borrowing, customer rates, regulatory approvals, construction risk, and the utility’s ability to finance other system needs. **Related definitions:** [Data center financing](#entry%5FKB281) · [Stranded costs](#entry%5FKB283) · [Nuclear power](#entry%5FKB092) · [ASICs](#entry%5FKB192) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-282 ### Stranded costs How does Stranded costs affect project risk? Stranded costs are investments or contractual commitments that cannot be fully recovered because demand, technology, policy, market conditions, or customer behavior changes. Large-load projects can create stranded risk when utilities build dedicated or shared infrastructure before customer demand and payment obligations are sufficiently secure. **Related definitions:** [Utility balance sheets](#entry%5FKB282) · [Credit support](#entry%5FKB284) · [AI infrastructure economics](#entry%5FKB004) · [Transmission cost allocation](#entry%5FKB109) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-283 ### Credit support What evidence should investors require for Credit support? Credit support is financial protection provided through guarantees, letters of credit, cash collateral, surety bonds, parent commitments, or other instruments. Investors and utilities should evaluate issuer strength, enforceability, amount, duration, replenishment, ranking, transfer, termination, and whether support remains available through construction and operation. **Related definitions:** [Stranded costs](#entry%5FKB283) · [Capital discipline](#entry%5FKB285) · [Competitive transmission](#entry%5FKB108) · [ASICs](#entry%5FKB192) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-284 ### Capital discipline How should developers manage Capital discipline? Capital discipline is the consistent allocation of money to projects whose expected value and strategic importance justify their risk. It requires realistic forecasts, comparable alternatives, stage gates, independent challenge, contingency, accountability, post-investment review, and willingness to delay or stop projects when evidence weakens. **Related definitions:** [Credit support](#entry%5FKB284) · [Transmission planning](#entry%5FKB106) · [AI infrastructure economics](#entry%5FKB004) · [Latency and geography](#entry%5FKB013) > Last reviewed: 2026-07-11 | Review frequency: Semiannual | Entry ID: KB-285 SECTION 20 | 15 DEFINITIONS ## AIxEnergy Concepts Original AIxEnergy frameworks for understanding the operating, institutional, and economic consequences of AI-driven electricity demand. **Terms in this section:** [Shadow Grid](#entry%5FKB286) · [Cognitive Grid](#entry%5FKB287) · [Phantom Data Centers](#entry%5FKB288) · [Infrastructure Intelligence](#entry%5FKB289) · [Commitment Discipline](#entry%5FKB290) · [Dispatchable Flexibility](#entry%5FKB291) · [Infrastructure Readiness](#entry%5FKB292) · [Electricity Customer Bill of Rights](#entry%5FKB293) · [Megawatt Economy](#entry%5FKB294) · [Speed to Power](#entry%5FKB295) · [Credible Deliverability](#entry%5FKB296) · [AI Load Corridors](#entry%5FKB297) · [Grid as institution](#entry%5FKB298) · [Operational proof](#entry%5FKB299) · [Power-to-intelligence conversion](#entry%5FKB300) ### Shadow Grid What is Shadow Grid? The Shadow Grid is an AIxEnergy concept describing the largely unseen network of proposed, reserved, contingent, and partially committed power infrastructure that forms around anticipated AI demand before physical load, generation, and network upgrades fully materialize. It includes queue positions, land options, supply claims, provisional utility plans, and capacity held against uncertain projects. **Related definitions:** [Cognitive Grid](#entry%5FKB287) · [AI infrastructure strategy](#entry%5FKB015) · [Phantom data centers](#entry%5FKB168) · [Data center load forecasts](#entry%5FKB167) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-286 ### Cognitive Grid Why does AIxEnergy use the concept of Cognitive Grid? The Cognitive Grid is an AIxEnergy framework for an electric system in which sensing, forecasting, optimization, machine reasoning, and institutional decision processes are integrated without surrendering engineering validation, auditability, cybersecurity, or human accountability. The concept emphasizes that intelligence must improve the grid’s operating institutions, not merely add algorithms. **Related definitions:** [Shadow Grid](#entry%5FKB286) · [Phantom Data Centers](#entry%5FKB288) · [Utility AI adoption](#entry%5FKB071) · [Phantom data centers](#entry%5FKB168) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-287 ### Phantom Data Centers How does Phantom Data Centers help explain AI-energy infrastructure? Phantom Data Centers is an AIxEnergy concept for proposed facilities that influence forecasts, queues, infrastructure plans, or public expectations without enough evidence of financing, customers, equipment, site readiness, and executable schedules. The framework distinguishes legitimate development uncertainty from demand claims too weak to justify durable utility investment. **Related definitions:** [Cognitive Grid](#entry%5FKB287) · [Infrastructure Intelligence](#entry%5FKB289) · [Phantom data centers](#entry%5FKB168) · [Cloud infrastructure](#entry%5FKB007) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-288 ### Infrastructure Intelligence What operational evidence supports Infrastructure Intelligence? Infrastructure Intelligence is an AIxEnergy framework for combining technical data, commercial evidence, regulatory context, supply-chain information, and operating conditions to determine what infrastructure can actually be delivered. It moves analysis beyond maps and announcements toward verified capacity, timing, dependencies, ownership, and risk. **Related definitions:** [Phantom Data Centers](#entry%5FKB288) · [Commitment Discipline](#entry%5FKB290) · [Utility AI adoption](#entry%5FKB071) · [Planning reserve margins](#entry%5FKB159) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-289 ### Commitment Discipline How should decision-makers apply Commitment Discipline? Commitment Discipline is an AIxEnergy principle requiring infrastructure decisions to scale with demonstrated commercial and operational commitment. Forecasts, queue positions, and announcements should not trigger the same investment as financed projects with site control, milestones, contracts, financial assurance, and accountable counterparties. **Related definitions:** [Infrastructure Intelligence](#entry%5FKB289) · [Dispatchable Flexibility](#entry%5FKB291) · [Developer credibility](#entry%5FKB180) · [AI load corridors](#entry%5FKB010) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-290 ### Dispatchable Flexibility What is Dispatchable Flexibility? Dispatchable Flexibility is an AIxEnergy concept for demand flexibility that is visible, contractually defined, measurable, controllable, and available to the institution responsible for reliability. It is more rigorous than claiming that a load could theoretically move; the capability needs telemetry, operating rules, notice, duration, compensation, testing, and consequences for nonperformance. **Related definitions:** [Commitment Discipline](#entry%5FKB290) · [Infrastructure Readiness](#entry%5FKB292) · [Curtailable service](#entry%5FKB173) · [Frequency response](#entry%5FKB155) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-291 ### Infrastructure Readiness Why does AIxEnergy use the concept of Infrastructure Readiness? Infrastructure Readiness is an AIxEnergy framework for evaluating whether a project or region can convert ambition into dependable operating capacity. It integrates power deliverability, interconnection, land, permits, equipment, water, networks, financing, workforce, contracts, construction, and institutional responsibility rather than treating any single favorable input as proof of readiness. **Related definitions:** [Dispatchable Flexibility](#entry%5FKB291) · [Electricity Customer Bill of Rights](#entry%5FKB293) · [Infrastructure readiness](#entry%5FKB009) · [Developer credibility](#entry%5FKB180) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-292 ### Electricity Customer Bill of Rights How does Electricity Customer Bill of Rights help explain AI-energy infrastructure? The Electricity Customer Bill of Rights is an AIxEnergy framework for protecting existing customers, new developers, utilities, communities, and regulators during large-load growth. It emphasizes transparent costs, fair risk allocation, reliable service, credible commitments, understandable decisions, community safeguards, and enforceable obligations across the infrastructure-development process. **Related definitions:** [Infrastructure Readiness](#entry%5FKB292) · [Megawatt Economy](#entry%5FKB294) · [Utility planning](#entry%5FKB064) · [Customer obligations](#entry%5FKB179) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-293 ### Megawatt Economy What operational evidence supports Megawatt Economy? The Megawatt Economy is an AIxEnergy concept describing a market in which access to dependable electrical capacity becomes a central determinant of digital growth, industrial location, investment value, and strategic advantage. In this economy, megawatts are not a background utility input; they are scarce, scheduled, financed, and competed-for productive capacity. **Related definitions:** [Electricity Customer Bill of Rights](#entry%5FKB293) · [Speed to Power](#entry%5FKB295) · [Compute as strategic resource](#entry%5FKB006) · [Phantom data centers](#entry%5FKB168) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-294 ### Speed to Power How should decision-makers apply Speed to Power? Speed to Power is an AIxEnergy framework for the time required to deliver usable, dependable electrical service to a project. It includes studies, permits, equipment, network upgrades, generation, contracts, financing, construction, testing, and operating approvals; announced generation or nominal grid proximity alone does not establish speed to power. **Related definitions:** [Megawatt Economy](#entry%5FKB294) · [Credible Deliverability](#entry%5FKB296) · [Infrastructure readiness](#entry%5FKB009) · [Developer credibility](#entry%5FKB180) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-295 ### Credible Deliverability What is Credible Deliverability? Credible Deliverability is an AIxEnergy standard for judging whether promised infrastructure capacity can be provided at the stated location, quantity, quality, cost, and time. It requires evidence across engineering, commercial commitment, supply chains, permits, construction, contracts, and institutional accountability rather than relying on nameplate resources or aspirational schedules. **Related definitions:** [Speed to Power](#entry%5FKB295) · [AI Load Corridors](#entry%5FKB297) · [Infrastructure readiness](#entry%5FKB009) · [Utility procurement](#entry%5FKB073) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-296 ### AI Load Corridors Why does AIxEnergy use the concept of AI Load Corridors? AI Load Corridors is an AIxEnergy concept for regions where data-center development, fiber, power supply, transmission, land, incentives, and specialized services create self-reinforcing clusters of AI demand. The framework highlights both scale advantages and concentrated risks involving grid capacity, water, cost allocation, permitting, community impact, and common-mode failure. **Related definitions:** [Credible Deliverability](#entry%5FKB296) · [Grid as institution](#entry%5FKB298) · [AI load corridors](#entry%5FKB010) · [Minimum demand charges](#entry%5FKB169) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-297 ### Grid as institution How does Grid as institution help explain AI-energy infrastructure? Grid as Institution is an AIxEnergy framework emphasizing that the electric grid is not only wires, generators, and control systems. It is also a set of utilities, regulators, markets, standards, contracts, planning processes, incentives, and public obligations that determine how physical capability is interpreted, financed, allocated, and operated. **Related definitions:** [AI Load Corridors](#entry%5FKB297) · [Operational proof](#entry%5FKB299) · [Reliability accountability](#entry%5FKB165) · [Phantom data centers](#entry%5FKB168) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-298 ### Operational proof What operational evidence supports Operational proof? Operational Proof is an AIxEnergy standard requiring capabilities to be demonstrated in the conditions that matter before they are relied upon for reliability, finance, or policy. A claim of flexibility, resilience, efficiency, or intelligence should be supported by telemetry, tests, performance history, controls, ownership, and consequences for failure. **Related definitions:** [Grid as institution](#entry%5FKB298) · [Power-to-intelligence conversion](#entry%5FKB300) · [Reliability accountability](#entry%5FKB165) · [Utility risk management](#entry%5FKB070) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-299 ### Power-to-intelligence conversion How should decision-makers apply Power-to-intelligence conversion? Power-to-Intelligence Conversion is an AIxEnergy framework for examining how electricity, chips, cooling, networks, software, utilization, and capital combine to produce useful AI output. It shifts attention from raw megawatts or compute purchases to the efficiency, value, timing, and system consequences of converting physical power into economic and institutional intelligence. **Related definitions:** [Operational proof](#entry%5FKB299) · [AI infrastructure strategy](#entry%5FKB015) · [NERC reliability standards](#entry%5FKB151) · [Load shape](#entry%5FKB174) > Last reviewed: 2026-07-11 | Review frequency: Annual | Entry ID: KB-300 End of AIxEnergy Dictionary ### Map Layer Catalog URL: https://www.aixenergy.io/aixenergy-map-layer-catalog/ Last updated: 2026-08-16T20:30:22.000Z [AIxEnergy Map](https://www.aixenergy.io/the-moratorium-map-how-the-ai-boom-is-redrawing-the-geography-of-power-and-permission/) brings together proprietary intelligence, public infrastructure data, market information, environmental constraints, and decision-oriented analytical layers in a single geospatial platform. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/08/aixenergy-map-2026-08-16--12-.png) The current production catalog includes 83 map layers across eight categories, spanning AI and large-load projects, site screening, transmission and interconnection, electricity markets, generation, environmental constraints, policy, fuel infrastructure, and connectivity. The objective is to make the relationships between those datasets useful for site selection, infrastructure planning, market intelligence, investment diligence, and large-load strategy, spanning AI and large-load projects, site screening, transmission and interconnection, electricity markets, generation, environmental constraints, policy, fuel infrastructure, and connectivity. 83Production overlays 8Layer categories 2,339Projects 2,276Map-plottable 4,902Curated sources 98Geographies [Projects & facilities](#projects-facilities)[AIxEnergy screening](#aixenergy-screening)[Transmission & interconnection](#transmission-interconnection)[Generation & markets](#generation-markets)[Environment & land](#environment-land)[Context & policy](#context-policy)[Fuel & renewables](#fuel-renewables)[Connectivity & fiber](#connectivity-fiber) **AIx** \= proprietary, merged, or calculated. **Public** \= available to any visitor. **Showcase** \= included across guest/free/journal/map access. **Journal** \= Journal + Map. **Map** \= full Map seat. CORE \= one of the 15 highest-value overlays. [Open AIxEnergy Map ↗](https://map.aixenergy.io/?ref=aixenergy.io) | Rank | Layer | Source / type | Access | Layer details | | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- | ------------- | -------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **1\. Projects & facilities · 3 layers** Curated AI infrastructure (status fill, PCP border) and published existing US facilities. | | | | | | 1 | [ProjectsCORE](#layer-projects) projects | AIx | Showcase | What it showsCurated AI-infrastructure markers from the private Project Keystone workbook, serialized through scripts/build\_data.py under an explicit public-field whitelist. Marker **fill** is normalized public\_status (Operational, Energizing, Under construction, Construction-ready, Advanced development, Planned, Announced / early planning, Canceled). Marker **border** is Project Completion Probability (PCP-1.1.0). Click a marker for Project Intelligence: overview, infrastructure, diligence, evidence, power, environment, site, and market — plus ranked advantages/constraints with “show on map” actions.Decision useWhere documented AI / large-load infrastructure is, how far it has progressed, how complete the evidence is, and how first energization screens. Use filters (status, company, geography, business model, evidence, PCP band, MW, capex) and shareable URLs. Pair with demand, queues, HV, water, and policy — the atlas is the reason those surfaces are commercially interesting.What it is notA complete census of every rumor. Parcel-accurate siting. Underwriting. A COD, financing outcome, or credit rating. The probability that announced IT load or capex will be fully delivered. Linked queue IDs are associated published requests, not proof the campus holds that interconnection position. Coordinates are approximate (site, city, regional, campus, or portfolio centroids). | | 2 | [Existing US data centers (IM3)](#layer-im3-datacenters) im3-datacenters | PNNL | Public | What it showsPNNL IM3 Open Source Data Center Atlas — an OSM-derived inventory of published existing US facilities. Point markers at continental zoom; footprints from zoom 8+.Decision useContrast the announced / pipeline atlas against the broader operating US footprint — “what is already here?” versus “what is being proposed?”What it is notThe AIxEnergy project list. Entitlement, capacity, IT load, or AIxEnergy research status. | | 3 | [Project labels](#layer-project-labels) project-labels | AIx | Showcase | What it showsCurated project names beside markers at closer zoom so dense clusters remain readable.Decision useReading campus names when presenting or sharing a regional view.What it is notA separate dataset. | | **2\. AIxEnergy screening · 11 layers** Proprietary indices for site, climate, water, permitting, tariffs, fiber, BTM, grid-shock, and U.S. state attractiveness. Missing inputs are omitted and remaining weights renormalize. Blanks are never coerced to zero. **Not parcel feasibility.** | | | | | | 1 | [Composite site score (US 0.25°)CORE](#layer-site-score-us) site-score-us CSS-US-0.25-4.0.0 | AIx | Showcase | What it showsCSS-4.0.0 on a 0.25° US land grid (0–100; higher = more favorable relative suitability). Click a cell for the research dossier and factor breakdown. On-demand Parcel Research pages public county GIS for the in-cell inventory and ranks a top-100 shortlist joined to cell power and fiber signals.Decision useFirst-pass physical screen for a 100–500 MW campus. Site Screening view is cell-first (atlas off) so the score is readable. Then pin-drop or parcel research inside a shortlisted cell.What it is notParcel feasibility, buildable acreage, a survey, entitlement, interconnect rights, or bankable siting. Published queues inside power access are **generation/storage-dominated**, not data-center large-load queues. Cell aggregates are not POI-accurate wait times. | | 2 | [Composite site score (EU 0.25°)CORE](#layer-site-score-eu) site-score-eu CSS-EU-0.25-4.0.0 | AIx | Map | What it showsCSS-4.0.0 on the Europe bidding-zone densify grid. Same 0–100 language and cell dossier as the US product.Decision useComparative European shortlists alongside CCI, SFCI, and market layers.What it is notParcel feasibility, an ENTSO-E planning model, or interconnect rights. | | 3 | [Large-load tariffCORE](#layer-large-load-tariff) large-load-tariff LLCCI | AIx | Journal | What it showsLLCCI applicability areas with embedded tariff-record metadata for hyperscale / large-load rate constructs. Click an area for the tabbed dossier of curated tariff evidence.Decision useWhether a special large-load rate construct appears to apply before you hire rate counsel.What it is notA binding quote, negotiated hyperscale rate, service-territory guarantee, or complete tariff library. | | 4 | [BTM configurationCORE](#layer-btm-configuration) btm-configuration BTM-CFG-1.1.0 | AIx | Journal | What it showsBTM-CFG-1.1.0 categorical heatmap of the lowest firm $/MWh mix for a flat 24/7 AI load on a closed 24-hour typical day. Popup: winning archetype and a mix vector per 100 MW IT at 100% load factor.Decision useWhich firm mix screens cheapest when the grid alone is expensive or tight. Pair with GDPI, US gas pipelines, and C&I retail.What it is notA dispatch model, capacity-expansion MILP, TMY coincidence study, PPA, interconnect rights, firm laterals, global buildable acres, or bankable LCOE. | | 5 | [Permitting frictionCORE](#layer-permitting-jurisdictions) permitting-jurisdictions PFI | AIx | Journal | What it showsPFI relative entitlement-friction score by ADM2-scale jurisdiction (county / equivalent). Higher friction = comparatively harder permitting screen.Decision useCompare jurisdictions early in site selection — before local counsel is hired.What it is notA legal opinion, zoning certainty, entitlement forecast, or substitute for counsel. | | 6 | [Cooling water accessCORE](#layer-water-territories) water-territories CWAF | AIx | Journal | What it showsCWAF relative difficulty of cooling-water access by water-utility territory polygon.Decision useFlag territories where cooling-water access may be harder relative to peers. Cooling Water Access view keeps the project atlas on.What it is notA water right, well permit, will-serve letter, plant water balance, or guarantee of supply. | | 7 | [BTM Gas Favorability](#layer-btm-gas-favorability) btm-gas-favorability BTM-GAS-1.0.0 | AIx | Journal | What it showsBTM-GAS-1.0.0 pipeline-gated spark-spread heatmap in USD/MWh. Green = industrial retail exceeds all-in BTM gas (attractive for on-site gas). Red = BTM gas is more costly than the prevailing C&I rate.Decision useWhether behind-the-meter gas screens cheaper than industrial retail on a typical day — the binary companion to BTM-CFG’s mix question.What it is notA PPA, heat-rate guarantee, firm gas capacity, laterals study, negotiated large-load tariff, or bankable LCOE. | | 8 | [Cooling climate](#layer-cooling-climate) cooling-climate CCI | AIx | Journal | What it showsCCI 0–100 thermal-suitability raster (higher = more favorable for cooling / free-cooling). Styled XYZ PNG tiles z0–z7.Decision useWhether climate favors air, evaporative, or water-intensive cooling — portfolio screen, not mechanical design.What it is notA facility energy model, PUE forecast, ASHRAE design study, live weather, climate-change projection, or a statement about power, water rights, or policy. | | 9 | [Synthetic fiber corridors](#layer-fiber-corridor) fiber-corridor SFCI-1.2.0 | AIx | Journal | What it showsSFCI-1.2.0 relative corridor-capacity index (0–100) on a land-aware 0.5° graph. Popup: capacity index, corridor score, hub-access proxy (km), corroboration flag.Decision useWhether digital-infrastructure access screens as a **corridor**, not a last-mile claim. Fiber Availability view. Used as a CSS factor (weight 0.125).What it is notLit fiber, carrier routes, ownership, latency, or measured bandwidth (Gbps). | | 10 | [U.S. state attractiveness](#layer-us-state-attractiveness) us-state-attractiveness USPA-1.0.0 | AIx | Journal | What it showsUSPA-1.0.0 0–100 relative screen and dense rank **#1–#50** across the 50 U.S. states. Color stretched across the score range. Click a state for narrative, pillar meters, and calculation. DC is computed for disclosure when inputs exist but is **not ranked**.Decision useState-level attractiveness **before** cell- and project-level work. Then go to CSS and the atlas inside the shortlisted state.What it is notA site score, a county ranking, parcel feasibility, interconnection position, negotiated tariffs, or project-success probability. | | 11 | [Grid shock events](#layer-grid-shock) grid-shock Grid Shock v0.5 | AIx | Journal | What it showsCurated atlas of documented large load-transfer / grid-shock episodes. City-scale references (not exact substations). Sized and colored by observed MW. Markers and rings are cartographic emphasis.Decision useWhere the grid has already absorbed a sudden large transfer — historical context for PJM, Dublin, and peer events.What it is notA substation-level reliability study, voltage contour, or campus footprint. | | **3\. Transmission & interconnection · 11 layers** HV/MV corridors, ISO day-ahead constraints, nodal LMP, interchange, queue stress, and published interconnection requests. | | | | | | 1 | [Large-load interconnection requestsCORE](#layer-large-load-queue) large-load-queue | AIx | Showcase | What it showsClassified national layer ≥50 MW (LLQ-1.3.0). Magenta markers with ice-white strokes sit on the teal-to-crimson generation-queue heatmap; names from zoom 6\. Market-sizing slider scales bubbles by requested MW (blank MW is not zero). Popup **record class** says which kind of feature you clicked. ERCOT and Dominion published residuals render as light territory fills.Decision useWhere competing large load is published — and **which class** the feature is. Analysis: projects near a published request (≤50 km) or inside a residual polygon.What it is notA complete US utility-portal harvest. A residual polygon is not a named request. An announced campus is not a utility queue row. | | 2 | [Interconnection requests (queue)CORE](#layer-ix-queue) ix-queue | AIx | Showcase | What it showsPublished ISO/RTO/utility **generation and storage** queue requests. Teal = lighter pressure; gold through crimson = jammed clusters (same cool-to-hot language as nodal DA LMP). Heatmap below zoom 8; inspectable points when closer. Click a request for the interconnection dossier (overview, member requests, stress, grid, evidence) and printable report.Decision useRelative access/stress screen and published-request proximity. Analysis can screen **PCP × queue** — higher-completion-probability campuses that also sit near published queue activity (timing risk, not a queue seat).What it is notA queue position, POI-accurate wait time, deliverability study, or the large-load layer. | | 3 | [US nodal DA LMPCORE](#layer-us-nodal-lmp) us-nodal-lmp | AIx | Showcase | What it showsDay-ahead nodal LMP at HIFLD substations ≥69 kV, colored USD/MWh (teal cheap → crimson expensive). Subtle 0.25° cells where load-side congestion or ISO-basis premium is elevated (moderate / high / severe only). Unpainted cells are **unknown, not unstressed**. Unmatched pnodes omitted.Decision usePrice context at matched nodes next to constraints and the pipeline. Cheap is not available capacity.What it is notWall-to-wall nodal coverage, a feeder tariff, line loading, ATC, interconnect rights, or a licensed EMS/CIM map. Constraints are not prices. | | 4 | [Transmission constraints / interfacesCORE](#layer-transmission-constraints) transmission-constraints | AIx | Map | What it showsISO day-ahead binding constraints and shadow prices, name-matched to HIFLD substations ≥115 kV **in the same ISO/RTO**. HV corridor lines only when from and to both match a HIFLD ≥115 kV line within \~5 km. Otherwise a substation pin or a jittered ISO-region **gray dot** (unlocated fallback — not a constrained POI). Straight station-to-station chords are not drawn. A name hit in another ISO is rejected.Decision useDocumented congestion **exposure** near a site or campus.What it is notMeasured line loading, ATC, MV/LV distribution congestion, an EMS/CIM map, or a constrained POI (gray dots). | | 5 | [US HV gridCORE](#layer-us-hv) us-hv | HIFLD | Public | What it showsHIFLD transmission lines and substations at **≥345 kV**, simplified for web-map performance.Decision usePlausible extra-high-voltage access. Near is not interconnected.What it is notInterconnection, spare MVA, CEII ratings, a power-flow model, or a NERC compliance model. | | 6 | [US power flows (interchange + ties)](#layer-transmission-flows) transmission-flows | AIx | Map | What it showsEIA-930 hourly **actual** BA-to-BA interchange (Form 930), not day-ahead schedules. Clock is hour-ending HE 1–24 in the reporting BA’s local time; the latest published hour is labeled. Schematic arrows on shared BA borders. Averages: 7-day, 30-day, YTD, trailing-12-month, weekday on-peak (HE 7–22), off-peak, summer / winter / spring, and a 30-day typical-hour (diurnal) mean. Arrowheads follow the latest hour; a reversing average is shown on the label. ISO/RTO labels (CAISO, ERCOT, ISO-NE, MISO, NYISO, PJM, SPP, plus Ontario IESO) show that BA’s net for the selected clock.Decision useWhich regions are leaning on which neighbors — energy tightness around AI load.What it is notCampus deliverability, ATC, a power-flow model, an EMS map, or line ratings. Schematic arrows are not electrical paths. Blank MW is unknown. | | 7 | [Planned transmission](#layer-planned-transmission) planned-transmission | HIFLD | Public | What it showsHIFLD under-construction / proposed lines ≥115 kV plus optional curated ISO plan extracts (RTEP / MTEP / RPG shapefiles dropped in private-source/planned-transmission/).Decision useWhether a corridor is already under construction — a timing signal next to today’s HV grid and binding constraints.What it is notA certain 2028–2032 grid. Proposed is not constructed capacity. | | 8 | [LBNL Queued Up](#layer-lbnl-queued-up) lbnl-queued-up | LBNL | Public | What it showsLawrence Berkeley National Lab Queued Up — county-level generation and storage queue MW summaries.Decision useCompare generation/storage queue intensity by county (order-of-magnitude FoM).What it is notData-center or large-load queues. Individual queue positions. | | 9 | [Europe HV grid](#layer-europe-hv) europe-hv | OSM | Public | What it showsPyPSA-Eur / OSM transmission and substations at **≥220 kV**.Decision useEuropean extra-high-voltage context beside bidding zones and prices.What it is notAn ENTSO-E planning model, TSO map, or interconnection study. | | 10 | [Global electric grid (OSM)](#layer-osm-power-lines) osm-power-lines | OSM | Public | What it showsOpenStreetMap power=line features worldwide at **≥220 kV**. Complements curated US/EU HV extracts.Decision useWorldwide HV screening backdrop (Global Power Prices / Global Power System views).What it is notA complete global transmission inventory, utility asset register, power-flow model, or interconnect rights. | | 11 | [US MV/LV grid (OSM)](#layer-osm-mv-power-lines) osm-mv-power-lines | OSM | Public | What it showsOpenStreetMap mid/low-voltage lines via OpenInfraMap vector tiles, filtered to **10 kV ≤ voltage < 220 kV** (the ≥220 kV OSM overlay is separate).Decision useSubtransmission / distribution context when zoomed into a candidate pad.What it is notA utility GIS extract or complete distribution model. | | **4\. Generation & markets · 12 layers** Electricity demand, operating and planned generation, utility territories, and wholesale / retail price surfaces. | | | | | | 1 | [Electricity DemandCORE](#layer-electricity-demand) electricity-demand | AIx | Showcase | What it showsEDM-1.4.0 hex demand nodes: annual MWh, average MW, estimated coincident peak MW, sector mix (residential, commercial, industrial, existing data centers, transportation, other), confidence, and — in 2030 — proposed and **expected** AI load. Click a hex for the demand dossier; nearby atlas projects within \~50 km are listed. Year rail, Growth view, and 2030 AI scenario.Decision useWhere existing peak, organic growth, and expected AI load stack — and which hexes would be overwhelmed by the announced pipeline.What it is notMeters, feeders, a utility allocation, or interconnect rights. | | 2 | [C&I retail electricity pricesCORE](#layer-ci-retail-prices) ci-retail-prices | AIx | Map | What it shows2024 USD/MWh heat map of typical commercial and industrial retail rates. Click a jurisdiction for values and source notes.Decision useTypical C&I retail cost by utility / jurisdiction — the retail side of every BTM and CSS screen.What it is notA negotiated large-load rate. Averages can mask LLCCI constructs. | | 3 | [Utility service territories](#layer-utility-territories) utility-territories | HIFLD | Public | What it showsApproximate retail utility footprints (EIA-861 / HIFLD lineage), simplified for web display.Decision useWho serves the site — the geography for tariff, demand, and large-load residual screens.What it is notA legal franchise map, will-serve, or rate schedule. | | 4 | [US power plants (EIA-860M)](#layer-eia-us-power-plants) eia-us-power-plants | EIA | Public | What it showsPlant-aggregated EIA-860M inventory. Existing ≥50 MW = filled circles; planned ≥20 MW = hollow rings; retired ≥50 MW since 2002 = muted. Fuel color on existing/planned; size ∝ disclosed capacity. Optional color by capacity factor, CO₂ intensity, or annual generation — **gray is unknown, not zero**. Click for EIA-923 plant/prime-mover net generation and electric fuel (latest complete year) plus EPA eGRID annual CO₂ / NOx / SO₂ rates (lb/MWh) when Plant ID joins.Decision useSupply mix, output, and emissions context around a site.What it is notA production-cost model. | | 5 | [Energy storage](#layer-energy-storage) energy-storage | AIx | Map | What it showsEIA-860M batteries and pumped storage plus active published interconnection-queue BESS/hybrid requests. Heatmap below zoom 8; circles from zoom 5\. MWh blank when unpublished.Decision useOperating, planned, and queued storage near the pipeline.What it is notA new storage database or a duration guarantee. LBNL county storage\_mw stays on Queued Up. | | 6 | [US day-ahead prices](#layer-us-da-prices) us-da-prices | AIx | Map | What it showsUSD/MWh heatmap of ISO **zone/hub averages** with labeled centroids — not nodal settlement.Decision useComparative wholesale cost context at zone/hub scale.What it is notA hyperscale tariff, nodal LMP series, forward curve, or bilateral quote. | | 7 | [US ISO / RTO](#layer-us-iso-rto) us-iso-rto | AIx | Map | What it showsCurated market-region footprints from public ISO/RTO boundary materials.Decision useWhich organized market the site sits in.What it is notA nodal model or the official ISO footprint for filings. Confirm with the ISO. | | 8 | [Europe day-ahead prices](#layer-europe-da-prices) europe-da-prices | AIx | Map | What it showsEUR/MWh heatmap of bidding-zone averages with labeled centroids (Energy-Charts / Fraunhofer-style public feeds).Decision useEuropean wholesale context.What it is notA trading feed, imbalance price, PPA, or industrial tariff. | | 9 | [Europe bidding zones](#layer-europe-bidding-zones) europe-bidding-zones | ENTSO-E | Public | What it showsENTSO-E community bidding-zone boundaries — the spatial scaffold for European DA prices and CSS-EU densify.Decision useZone orientation and price overlays.What it is notA price or a complete market-rules reference. | | 10 | [International day-ahead prices](#layer-intl-da-prices) intl-da-prices | AIx | Map | What it showsRegional spot heatmap where public feeds exist, with labeled centroids.Decision useNon-US/EU wholesale context where a feed exists.What it is notA universal price surface, trading feed, or bilateral quote. | | 11 | [Generation points](#layer-power-plants) power-plants | AIx | Map | What it showsGlobal generation points **≥100 MW**: US EIA-860M existing + WRI Global Power Plant Database non-US. Capacity and fuel where disclosed.Decision useProximity to large existing generation outside the detailed EIA-860M US layer.What it is notA complete small-plant inventory. | | 12 | [Plant labels](#layer-eia-us-power-plant-labels) eia-us-power-plant-labels | EIA | Public | What it showsEIA-860M plant names beside markers when zoomed in.Decision useReading plant names on a regional plate.What it is notA separate dataset. | | **5\. Environment & land · 17 layers** Population, buildings, water, hazard, land cover, and siting-constraint screens. Ranked by how often the overlay kills or reshapes a 100–500 MW shortlist. | | | | | | 1 | [Water stress](#layer-water-stress) water-stress | WRI | Public | What it showsWRI Aqueduct **baseline water stress** at watershed scale.Decision useWhether evaporative or water-intensive cooling is even plausible at basin scale.What it is notA water right, cooling-water capacity, facility water use, or future climate hydrology. | | 2 | [Flood hazard (FEMA NFHL)](#layer-flood-nfhl) flood-nfhl | FEMA | Public | What it showsFEMA National Flood Hazard Layer Flood Hazard Zones.Decision useUS flood-zone screening for pad and insurance friction.What it is notSFHA delineation, FIRM determination, or flood insurance rating. | | 3 | [NWI wetlands](#layer-nwi-wetlands) nwi-wetlands | USFWS | Public | What it showsUSFWS National Wetlands Inventory classed presence: estuarine, freshwater emergent/forested, pond, lake, riverine, other.Decision useWetland-presence screen that can exclude a CSS cell.What it is notA Clean Water Act jurisdictional delineation or field-verified boundary. | | 4 | [Slope](#layer-slope) slope | USGS | Public | What it showsUSGS 3DEP Slope Map (flat gray → yellow → red-brown), proxied as PNG tiles.Decision useGradeability triage for a campus pad.What it is notA topographic survey, cut/fill, or pad engineering study. | | 5 | [Groundwater basins / aquifers](#layer-groundwater-basins) groundwater-basins | USGS | Public | What it showsUSGS Principal Aquifers (2003) outcrop and subcrop. Color is rock group. Aqueduct BWS is sampled at each aquifer **centroid** for basin-stress context only.Decision useWhether a principal aquifer is present — the third water question after basin stress and territory friction.What it is notA water right, well permit, AMA overlay, or confirmed industrial yield. Centroid BWS is not groundwater availability. | | 6 | [Wildfire hazard potential](#layer-wildfire-whp) wildfire-whp | USFS | Public | What it showsUSFS RMRS Wildfire Hazard Potential Classified **2023**.Decision useLandscape wildfire potential for campus and corridor risk.What it is notA seasonal forecast, ignition outlook, or First Street property-level product. | | 7 | [EPA nonattainment areas](#layer-epa-nonattainment) epa-nonattainment | EPA | Public | What it showsEPA OAQPS nonattainment polygons (NAAQS / Green Book) — ozone, PM2.5, and other criteria pollutants.Decision useWhether Clean Air Act nonattainment adds permitting friction.What it is notA permit decision, BACT determination, PSD applicability opinion, or facility ambient monitoring. | | 8 | [Protected areas](#layer-protected) protected | OSM | Public | What it showsOpenStreetMap parks and reserves as a conservation-sensitive-land proxy.Decision useFlag likely constrained areas early.What it is notAn authoritative protected-area inventory or a legal setback. | | 9 | [Natural hazard](#layer-natural-hazard) natural-hazard | ThinkHazard | Public | What it showsThinkHazard! (GFDRR / World Bank) multi-hazard levels joined to administrative units.Decision useMulti-hazard first screen.What it is notSite-specific engineering risk, insurance pricing, or catastrophe modeling. | | 10 | [USA Structures (ORNL / FEMA)](#layer-usa-structures) usa-structures | ORNL | Public | What it showsBuildings larger than 450 sq ft. Below zoom 12: buildings-per-km² mosaic/heatmap from count-only stats (/api/tiles/buildings-stats). Zoom 12+: occupancy-colored footprints (/api/tiles/buildings-preview), largest first.Decision useExisting built fabric and occupancy mix around a candidate pad.What it is notParcel, title, or legal boundaries. | | 11 | [Population density](#layer-population) population | WorldPop | Public | What it showsWorldPop **2020** density heatmap via MapTiler tiles.Decision useLabor, community, and load-context screening.What it is notA census block, commuting study, or growth forecast. | | 12 | [Quaternary faults (USGS)](#layer-quaternary-faults) quaternary-faults | USGS | Public | What it showsUSGS Quaternary Fault and Fold Database National layer (show:21).Decision useFault-proximity screen.What it is notA seismic design study or design ground-motion analysis. | | 13 | [Industrial / commercial land-use](#layer-land-use) land-use | OSM | Public | What it showsOpenStreetMap industrial and commercial polygons via MapTiler OpenMapTiles.Decision useAlready-industrialized land context.What it is notZoning entitlement or parcel land-use codes. | | 14 | [Airport exclusion buffers](#layer-airport-exclusion) airport-exclusion | AIx | Map | What it showsAIxEnergy **5 / 10 / 15 km** rings around medium and large OurAirports sites.Decision useAirport-proximity friction for height and community screens.What it is notFAA/EASA obstacle surfaces, height limits, or permitting determinations. | | 15 | [Grid carbon intensity](#layer-grid-carbon-intensity) grid-carbon-intensity | Ember | Public | What it showsEmber annual country electricity intensity (gCO₂eq/kWh) choropleth.Decision useBuyer Scope 2 / diligence screening.What it is notReal-time dispatch, balancing-authority intensity, or metered Scope 2. | | 16 | [Canopy height](#layer-canopy) canopy | ETH | Public | What it showsETH Global Canopy Height **2020** (Lang et al. 2023; CC BY 4.0) — satellite-estimated tree-canopy top height in meters.Decision useVegetation / clearing context.What it is notCorridor LiDAR or a NERC FAC-003 clearance determination. | | 17 | [Cities & towns](#layer-cities) cities | OSM | Public | What it showsOpenStreetMap place names via MapTiler OpenMapTiles.Decision usePlace-name orientation at mid zooms.What it is notA population ranking or labor-market product. | | **6\. Context & policy · 8 layers** Policy atlases, pushback evidence, European protected sites, and basemap context. Policy and pushback carry the value; roads and rail are orientation. | | | | | | 1 | [Policy constraintsCORE](#layer-policy) policy | AIx | Map | What it showsEvidence-graded constraint records joined to jurisdiction geometries: moratoria, pauses, restrictive postures, and related kinds. Features carry evidence metadata for the policy dossier.Decision useFormal policy / moratorium / restriction footprints before announcement.What it is notA legal opinion, exhaustive law, or political probability. Evidence grade is not the chance a rule survives. | | 2 | [Public pushback](#layer-pushback) pushback | AIx | Map | What it showsCurated protests, hearings, lawsuits, and opposition events tied to AI / data-center development, styled by recency.Decision useDocumented community or legal opposition — a social-license signal from public evidence.What it is notA complete news archive or a sentiment index. | | 3 | [EU Natura 2000](#layer-eu-natura-2000) eu-natura-2000 | EEA | Public | What it showsEEA Natura 2000 protected sites (SAC / SPA — Habitats and Birds Directives).Decision useEuropean protected-site friction on a shortlist.What it is notA permitting determination or Appropriate Assessment. | | 4 | [Policy labels](#layer-policy-labels) policy-labels | AIx | Map | What it showsJurisdiction names on constraint polygons when zoomed in.Decision useReading the constraint geography without opening every feature.What it is notA separate dataset. | | 5 | [Political boundaries](#layer-boundaries) boundaries | MapTiler | Public | What it showsCountry, state/province, and county lines from MapTiler / OpenMapTiles.Decision useOrientation and jurisdiction context. On in most Views.What it is notA cadastral or disputed-boundary product. | | 6 | [Water features](#layer-water-features) water-features | OSM | Public | What it showsRivers, lakes, and oceans from OpenStreetMap via MapTiler.Decision useSurface-water orientation.What it is notA water-right map, bathymetry, or floodplain. | | 7 | [Roads](#layer-roads) roads | OSM | Public | What it showsHighways and local roads from OpenStreetMap via MapTiler.Decision useAccess and logistics context.What it is notA traffic or overweight-permit study. | | 8 | [Rail corridors](#layer-rail) rail | OSM | Public | What it showsOpenStreetMap rail lines — detail from zoom 8+.Decision useRail / ROW context for logistics and fiber-corridor proxies.What it is notOwnership, siding capacity, or freight rights. | | **7\. Fuel & renewables · 16 layers** Gas corridors, coal and gas delivery prices, and renewable resource / plant screening. **Not interconnect rights.** Fuel-price fabrics (GDPI, CDPI) and corridor presence rank above resource rasters because they change a BTM or thermal-generation go / no-go. | | | | | | 1 | [Natural gas delivery prices](#layer-gas-delivery-prices) gas-delivery-prices | AIx | Map | What it showsGDPI heatmap: US citygate + EU industrial region fills, hub and citygate point labels in **USD/MMBtu**, schematic corridor-basis spines. Click for USD/MMBtu, HHV MWh, basis versus Henry Hub, cadence, source, and nearest pipeline snap.Decision useFuel-price context for BTM and industrial siting. Gates BTM-GAS (nearest GDPI price; else hub within 400 km).What it is notFirm delivered meter prices. Corridor spines are schematic — not metered routes or interconnect rights. Pipeline snap is proximity only. | | 2 | [Coal delivery prices](#layer-coal-delivery-prices) coal-delivery-prices | AIx | Map | What it showsCDPI — companion coal-price fabric to GDPI in a common **USD/MMBtu** language (also tonne, short ton, and HHV MWh). US delivered-to-electric choropleth fills; producer-country FOB fills; seaborne and US basin hub markers; schematic rail/seaborne spines. Year/month **vintage slider** restyles hubs, fills, and spines from coal-price-history/timeline.json using the latest observation on or before the selected date. Click for price, vintage, observed date, industrial SEDS where attached, basis versus Newcastle, cadence, and source.Decision useThermal-fuel price context beside GDPI — seaborne FOB versus US delivered-electric, and how that spread has moved.What it is notMine-specific contracts, firm rail/terminal rights, or daily API2 / API4 / NEWC prints. Energy Institute markers are annual S&P-based averages republished by OWID. EIA weekly coal commodity spots are S&P-proprietary and are **not redistributed**. Withheld EPM cells stay blank unless a documented fallback exists — never coerced to zero. Corridor spines are schematic. | | 3 | [US gas pipelines](#layer-gas-us) gas-us | EIA | Public | What it showsEIA / HIFLD interstate and intrastate transmission corridors.Decision useUS fuel-corridor proximity for behind-the-meter screens.What it is notA nomination, interconnect, pressure/capacity model, or engineering alignment. | | 4 | [Global gas pipelines](#layer-gas-global) gas-global | GEM | Public | What it showsGlobal Energy Monitor GGIT transmission routes.Decision useFuel-corridor presence worldwide.What it is notFirm capacity, lateral rights, or nomination availability. | | 5 | [US gas compressor stations](#layer-gas-compressors) gas-compressors | HIFLD | Public | What it showsHIFLD compressor stations along transmission corridors.Decision useCompression / delivery-point context.What it is notAvailable horsepower, spare capacity, or a gas-quality/offtake study. | | 6 | [GWA US wind (100 m)](#layer-gwa-us-wind) gwa-us-wind | GWA | Public | What it showsGlobal Wind Atlas mean wind speed at **100 m** for CONUS (\~250 m source; regional PNG pyramid z0–z6).Decision useCONUS wind screen at a consistent global methodology.What it is notMicro-siting, IEC class certification, measured site wind, or bankable yield. | | 7 | [GSA US solar (PVOUT)](#layer-gsa-us-solar) gsa-us-solar | GSA | Public | What it showsGlobal Solar Atlas **PVOUT** for CONUS (Solargis / World Bank).Decision useCONUS PV-output screen.What it is notInterconnect rights, plant design basis, or bankable production. | | 8 | [GWA EU wind (100 m)](#layer-gwa-eu-wind) gwa-eu-wind | GWA | Public | What it showsSame GWA 100 m mean wind-speed product clipped to Europe (regional PNG; \~150 tiles).Decision useEuropean wind screen.What it is notA project resource assessment or grid connection. | | 9 | [GSA EU solar (PVOUT)](#layer-gsa-eu-solar) gsa-eu-solar | GSA | Public | What it showsGSA PVOUT clipped to Europe (regional PNG; \~113 tiles).Decision useEuropean PV-output screen.What it is notInterconnect rights or engineered yield. | | 10 | [NREL denser US wind](#layer-nrel-wind) nrel-wind | NREL | Public | What it showsNREL Wind Toolkit fields served via Esri Living Atlas — a denser CONUS wind-resource screen than the GWA PNG pyramid.Decision useHigher-resolution US wind-resource screening when GWA’s \~250 m climatology is too coarse.What it is notMeasured site wind, wake-adjusted yield, or interconnectable capacity. | | 11 | [NSRDB GHI](#layer-nsrdb-ghi) nsrdb-ghi | NREL | Public | What it showsNREL National Solar Radiation Database **global horizontal irradiance** at \~4 km.Decision useSolar-resource screen complementary to GSA PVOUT (irradiance vs. modeled PV output).What it is notPV layout yield, soiling, or interconnection. | | 12 | [GWA global wind (PMTiles)](#layer-gwa-global-wind) gwa-global-wind | GWA | Public | What it showsNear-global Global Wind Atlas 100 m via Worker/R2 PMTiles (/api/tiles/pmtiles/gwa-wind-100m/{z}/{x}/{y}).Decision useWorldwide wind backdrop outside US/EU packs.What it is notWall-to-wall met-mast truth, site measurement, or LCOE. | | 13 | [GSA global solar (PMTiles)](#layer-gsa-global-solar) gsa-global-solar | GSA | Public | What it showsNear-global Global Solar Atlas PVOUT via Worker/R2 PMTiles (/api/tiles/pmtiles/gsa-pvout/{z}/{x}/{y}).Decision useWorldwide solar backdrop outside US/EU packs.What it is notA bankable resource study or site-design irradiance. | | 14 | [US wind turbines](#layer-us-wind-turbines) us-wind-turbines | USGS | Public | What it showsUSGS US Wind Turbine Database (USWTDB) observed turbine locations.Decision useExisting wind-plant footprint context.What it is notA queue, a PPA, capacity-factor performance, or planned builds. | | 15 | [US solar PV plants](#layer-us-solar-plants) us-solar-plants | USGS | Public | What it showsUSGS US Large-Scale Solar Photovoltaic Database (USPVDB) plant locations and footprints where published.Decision useExisting utility-scale solar context.What it is notA queue, a PPA, or the planned pipeline. | | 16 | [NREL reV wind supply curve](#layer-rev-supply) rev-supply | NREL | Public | What it showsNREL reV LCOT / all-in LCOE centroids — approximate supply-cost geography.Decision useIndicative wind supply-cost geography.What it is notA site-specific LCOE, PPA offer, bid, or transmission deliverability study. | | **8\. Connectivity & fiber · 5 layers** IX / cloud proximity, published fibre routes, and broadband availability. Synthetic fiber corridors (screening category) already carry the proprietary corridor-capacity claim; these layers are the published evidence underneath and around it. | | | | | | 1 | [Connectivity terminals](#layer-connectivity) connectivity | AIx | Map | What it showsMerge of PeeringDB IXPs/facilities with curated cable landings and cloud-region points. The hub set used for proximity screening and SFCI pathing (3,252 terminals in the current SFCI build).Decision useDigital-infrastructure proximity for latency-sensitive campuses.What it is notA complete carrier inventory, a cross-connect, or a cloud on-ramp contract. | | 2 | [Cloud / IX distance rings](#layer-connectivity-rings) connectivity-rings | AIx | Map | What it showsAIxEnergy **25 / 50 / 100 km** rings from curated cloud regions and IXPs.Decision useDistance bands to cloud and interconnection hubs.What it is notNetwork latency, fiber path length, or available capacity. | | 3 | [Open fibre routes (OFDS)](#layer-ofds-fiber) ofds-fiber | OFDS | Public | What it showsDigitized public operator fibre maps from the Open Fibre Data Standard corpus.Decision usePublished fibre-route proximity — corroboration for SFCI, not a substitute.What it is notA complete national fibre map or a lit wave. | | 4 | [Tower density](#layer-tower-density) tower-density | AIx | Map | What it showsAIxEnergy 0.25° density grid aggregated from OSM telecom towers.Decision useRelative terrestrial-wireless / infra activity.What it is notCoverage quality, 5G capacity, or a backhaul path. | | 5 | [US broadband availability](#layer-us-broadband) us-broadband | FCC | Public | What it showsCounty-level FCC availability fraction. **October 2020 vintage** in the current packaging.Decision useCoarse last-mile availability context only.What it is notCurrent FCC fabric, FTTP, enterprise Ethernet, lit circuits, or dark fiber. | **Screening discipline:** Blank numeric fields remain unknown rather than being converted to zero. Proximity to infrastructure, queues, tariffs, water, or policy geography is a screening signal—not proof of interconnection, capacity, entitlement, legal rights, or underwriting. ### Map Capabilities URL: https://www.aixenergy.io/map-capabilities/ Last updated: 2026-08-17T10:54:54.000Z AIxEnergy Map brings the documented AI infrastructure pipeline, proprietary screening, electricity demand, grid infrastructure, interconnection activity, markets, water and policy into one professional decision environment — so teams can move from a question to a defensible screen. [Request a demonstration ](mailto:bowens@aixenergy.io?subject=AIxEnergy%20Map%20Demonstration) [Explore AIxEnergy Map → ](https://map.aixenergy.io/?ref=aixenergy.io) The decision problem ## The answer does not live in one dataset. Public GIS can show a transmission line. A market terminal can show a price. Neither can tell you which documented AI campuses sit on that corridor, how complete the underlying evidence is, or how first energization screens under competing risks. AIxEnergy Map connects the project pipeline to power infrastructure, demand, interconnection activity, markets, water, policy and proprietary analytics so those relationships can be examined together. You are not buying 83 layers. You are buying six capabilities. 2,339 documented AI and large-load infrastructure project records 4,902 public-source citations supporting the project atlas 98 geographies represented across the project corpus Six capabilities ## From market question to defensible screen. Each capability begins with a decision professionals already need to make. The layers, models and evidence sit underneath that decision rather than becoming the product story themselves. 01 ## See the documented pipeline. What is actually being proposed — and how far has it progressed? AIxEnergy Map publishes a curated atlas of documented AI and large-load infrastructure. Marker fill represents normalized public project status. Marker border represents Project Completion Probability (PCP-1.1.0) — AIxEnergy's competing-risks screen of first energization versus cancellation versus remaining in the pipeline. Click a campus to open Project Intelligence: evidence, power, environmental, site and market context, together with ranked findings that can be returned directly to the map. Existing U.S. facilities from the PNNL IM3 atlas sit beside the development pipeline so operating stock is not confused with what has merely been proposed. **Decision discipline:**A project link is not a queue position. Approximate locations remain identified as approximate. Blank fields remain blank. Typical decisions Underwrite an announced campus. See who is already developing nearby. Rank a state before ranking individual cells. Share a project profile as a reusable evidence reference. Primary users Infrastructure investors, private equity, banks, hyperscalers, developers, utility large-load teams, economic-development organizations and consultants. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/08/aixenergy-map-us-projects-overview-2026-08-16--3-.png) 02 ## Screen a campus before you fly. The expensive studies should start after the cheap ones are finished. Composite Site Score (CSS-4.0.0) combines power access, cooling climate, fiber, cooling-water access, C&I electricity price, natural hazard and permitting friction on a 0.25° U.S. grid. Hard exclusions screen out cells with no nearby high-voltage substation, steep slope or wetland cover. Click a cell for its research dossier. Drop a pin for a live siting pre-read. Then stack cooling-water territory friction, permitting, large-load tariff applicability, flood and wetlands before committing expensive field diligence. Europe uses the same core screening recipe on a 0.25° densified grid, with queue congestion and U.S.-specific terrain filters omitted. **A high score is a screen, not a site.**AIxEnergy Map does not say a location can interconnect, has secured water or is entitled. It identifies where the first-pass evidence is strong enough to justify deeper work — and where it is not. Typical decisions Build a 200–400 MW greenfield shortlist. Eliminate water-infeasible locations. Test the institutional stack. Compare European markets using a consistent screening language. Primary users Hyperscaler site-selection teams, data-center developers, consultants, engineering firms, industrial large-load customers and economic-development organizations. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/08/aixenergy-map-site-screening-2026-08-16.png) 03 ## Read two queues and the stressed grid. There are two queues on this map. They are not the same object. Cheap is not available capacity. Generation and storage interconnection requests render as a teal-to-crimson stress heatmap. Large-load requests render separately in magenta through AIxEnergy's classified national screen (LLQ-1.3.0), combining BPA named queue GIS, announced campuses of at least 50 MW, named public filings, and ERCOT and Dominion residual totals. Residual totals remain territory polygons. Points are not invented from aggregate data. Alongside them sit HIFLD extra-high-voltage transmission, ISO day-ahead binding constraints matched to substations, nodal day-ahead LMP at matched HIFLD nodes, and public under-construction transmission corridors. **Interpretation matters:**Unpainted LMP is unknown, not unstressed. An unlocated constraint is not a constrained point of interconnection. Near a queue request is not a queue position. Typical decisions Compare competing generation and competing load. Read price next to documented congestion. Time a transmission corridor. Screen likely energization against published queue pressure. Primary users Utility and transmission-owner large-load teams, hyperscalers, generation developers, transmission developers, storage developers and investors. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/08/aixenergy-map-interconnection-requests-2026-08-16.png) 04 ## See demand shock before the planning cycle. Expected 2030 load is probability-weighted. The unadjusted announcement book is not the forecast. Electricity Demand (EDM-1.4.0) is a bottom-up U.S. demand model calibrated to EIA-861\. Hexes show who consumes electricity, how organic demand is changing and, in 2030, where expected first-energization load would stack on existing local peak. Operational campuses use recorded operating or energized IT MW when that evidence exists. Announced IT capacity is never presented as operating load, and proposed load never enters the 2024 baseline. Click a demand hex and the map identifies nearby named atlas projects within roughly 50 km, allowing a high proposed-load share to be examined against actual project names, evidence grades and statuses rather than an anonymous demand smear. **The central distinction:**organic electricity-demand growth and expected AI infrastructure load are separate signals and should remain separate in planning discussions. Typical decisions Brief a large-load process before the next planning cycle. Separate organic growth from AI shock. Test a utility territory under incremental load. Show a board how proposed campuses compare with existing peak. Primary users Utility and transmission planners, regulators, public planners, infrastructure investors, consultants and economic-development organizations. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/08/aixenergy-map-electricity-demand-2026-08-16.png) 05 ## Price the power and the BTM alternative. Cheap solar is not cheap 24/7 power. High AI load factor changes the comparison. C&I retail prices, ISO zone and hub day-ahead prices, nodal LMP, gas delivery prices (GDPI) and coal delivery prices (CDPI) share the same decision environment. Behind-the-Meter Configuration identifies the lowest screened firm $/MWh configuration for a flat 24/7 AI load among four alternatives: grid, gas, gas plus renewables, or renewables plus storage plus grid. BTM Gas Favorability answers a narrower question: where does pipeline-gated gas screen cheaper than industrial retail electricity on a typical day? The analysis can then be viewed beside gas pipelines, gas delivery prices and large-load tariff constructs. **Screening, not quotation:**A favorable BTM result does not mean fuel is contracted, a gas lateral exists, an air permit is available, or a commercial power agreement can be executed at the screened value. Typical decisions Test BTM optionality. Compare retail, zone and nodal power economics. Evaluate gas and coal on a common fuel-cost basis. Understand where a large-load tariff construct appears to apply. Primary users Industrial and campus energy teams, developers considering on-site firm power, IPP and hybrid developers, investors, utility rate teams and utility large-load teams. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/08/aixenergy-map-behind-the-meter-power-2026-08-16.png) 06 ## Leave with a packet, not a screenshot. A View replaces layers on purpose. Repeatability is the product. A 40-layer map is not. Sixteen curated Views apply the layer pack, camera and basemap required for one decision. Analysis screens the project atlas against registered conditions and reports pass, fail or unknown, together with the dominant eliminating criterion — including missing data. Charts quantify what is actually on the map. Pin-drop analysis, CSS cell dossiers and, in supported U.S. cells, parcel research turn a coordinate into a structured pre-read. Executive PDFs and an untitled map plate turn the interactive analysis into material that can leave the application. My Views stores zoom, basemap and layer configuration to the signed-in account, while share URLs can recover the camera, project or scene. **Unknown is a coverage gap, not zero.**Analysis will not tell a buyer that a project is interconnected. It can show which announced campuses sit near documented constraints or elevated proposed-share demand — and which screens fail because required evidence is missing. Typical decisions Run a registered multi-layer question. Hand the match set to Charts. Save the scene. Export the Analysis PDF. Deliver a state or regional decision packet that another professional can reopen. Primary users Consultants, engineering firms, utility planning and large-load teams, investment committees, developer steering committees, economic-development organizations and regulators. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/08/aixenergy-map-cooling-water-access-2026-08-16--1-.png) Sample outputs ## See what leaves the map. AIxEnergy Map is designed to produce more than an interactive screen. Project diligence, site screening and multi-layer analysis can move directly into professional decision processes through structured executive reports. The examples below show how map-based intelligence is converted into reusable professional outputs for project diligence, site evaluation and executive analysis. [Hyperion – Richland Parish · AIxEnergy Project Intelligence ProfileScreening is based on the project's mapped coordinate, which may represent an approximate project loHyperion – Richland Parish · AIxEnergy Project Intelligence Profile.pdf170 KBdownload-circle](https://www.aixenergy.io/content/files/2026/08/Hyperion-----Richland-Parish----AIxEnergy-Project-Intelligence-Profile.pdf "Download") [AIxEnergy\_Site\_Screening\_Fairfax\_Virginia\_site-cell-155--310\_20260807Regional screening analysisAIxEnergy\_Site\_Screening\_Fairfax\_Virginia\_site-cell-155--310\_20260807.pdf562 KBdownload-circle](https://www.aixenergy.io/content/files/2026/08/AIxEnergy%5FSite%5FScreening%5FFairfax%5FVirginia%5Fsite-cell-155--310%5F20260807.pdf "Download") [AIxEnergy\_Pin\_Siting\_205-north-litchfield-road-goodyear-arizona-85338\_2026-08-17AIxEnergy\_Pin\_Siting\_205-north-litchfield-road-goodyear-arizona-85338\_2026-08-17.pdf509 KBdownload-circle](https://www.aixenergy.io/content/files/2026/08/AIxEnergy%5FPin%5FSiting%5F205-north-litchfield-road-goodyear-arizona-85338%5F2026-08-17.pdf "Download") [AIxEnergy-Analysis-triple-constraint-2026-08-17Triple constraint screening analysisAIxEnergy-Analysis-triple-constraint-2026-08-17.pdf182 KBdownload-circle](https://www.aixenergy.io/content/files/2026/08/AIxEnergy-Analysis-triple-constraint-2026-08-17.pdf "Download") Who uses AIxEnergy Map ## Different organizations. The same physical system. Eight buyer groups sit on the same map. They do not need eight products. They need the six capabilities in different sequences. **Hyperscalers & data-center developers** Screen markets and sites before committing land, engineering resources and development capital. **Utilities & transmission owners** See incoming large-load pressure against demand, grid and interconnection conditions. **Infrastructure investors, PE & banks** Test teaser MW, evidence maturity, location quality and execution context before accepting a development narrative. **IPP, transmission & storage developers** Examine load growth, generation backlog, constraints and the grid required to support future development. **Industrial & campus energy teams** Compare retail power, wholesale conditions and behind-the-meter alternatives before entering an EPC or commercial process. **Consultants & engineering firms** Replace repeatedly reconstructed layer packages with a screen clients can reopen and reuse. **Economic-development organizations** Test whether a claimed “data-center-ready” market survives project, grid, demand and water scrutiny. **Regulators & public planners** Examine proposed large-load growth against documented projects, queues, demand and power-system conditions. ## Posts ### Community-Scale Data Centers Need a Contract, Not a Carveout URL: https://www.aixenergy.io/community-scale-data-centers-need-a-contract-not-a-carveout/ Last updated: 2026-10-05T05:28:24.000Z On July 14, New York created a 50 megawatt line for one specific purpose. Executive Order 62 directed the Department of Environmental Conservation to hold in abeyance certain discretionary permit, approval, and license applications for covered data centers that were pending before DEC and had not been determined complete before the order. Local permits were expressly excluded. The order defines covered data centers, for its own purposes, in part by whether a facility or group of facilities on the same or contiguous sites consumes or can consume at least 50 megawatts.1 Article Summary 0:00 /28.26449 1× That number is already being mistaken for something larger than it is. It is a policy threshold inside a particular New York process, not a finding that 49 megawatts is electrically inconsequential or that projects below the line deserve automatic approval. Other institutions draw different lines because they are solving different problems. AEP Ohio applies its specialized Schedule DCT process to new data center loads of 25 megawatts or more. ERCOT’s current Batch Zero process addresses eligible load facilities of 75 megawatts or greater. The North American Electric Reliability Corporation (NERC) is developing new registration criteria for computational loads, while the Federal Energy Regulatory Commission (FERC) has directed NERC to finalize registry criteria and initial reliability standards by the end of 2026.2 The useful conclusion is that megawatts are a triage variable. They can determine which pathway a project enters. They cannot replace evidence about the point of interconnection, the expected ramp, the system’s available capacity, on-site generation, power quality, water and emissions, upgrade costs, or who bears the risk if the forecast fails. ## The Demand Is Large-the Decisions Are Local A 2024 US Department of Energy report prepared with Lawrence Berkeley National Laboratory estimated that US data centers consumed about 176 terawatt-hours of electricity in 2023, or 4.4 percent of national electricity use. It projected 325 to 580 terawatt-hours by 2028, equivalent to roughly 6.7 to 12 percent of US electricity use under the report’s scenarios.3 Federal institutions are responding, but they are not creating one national siting code. On June 18, 2026, FERC issued Federal Power Act section 206 show-cause orders to the six regional grid operators under its jurisdiction and their transmission owners. The orders require them to justify existing tariffs or propose changes addressing large-load study processes, transmission cost shifting, co-location, flexible transmission service, and electrically proximate generation. FERC also stated that the orders do not displace state authority over generation siting and permitting or state public utility commission authority over retail rates, terms, and conditions.4 NERC is addressing a different layer: bulk-power-system reliability. Its May 2026 Level 3 alert called for actions involving computational-load modeling, studies, instrumentation, commissioning, operations, protection, and control. In July, FERC ordered NERC to complete new computational-load registry criteria and initial reliability standards by December 31, 2026\. At this article’s cutoff, those registry criteria and standards were still under development, not mandatory final rules.5 These actions reinforce the same point. A data center can be small relative to a hyperscale campus and still be consequential to the utility, substation, feeder, water system, or county that must absorb it. The material question is not whether a project fits a marketing label. It is what the project obligates other institutions to build, reserve, finance, operate, and defend. ## A Predictable Path Developers are right to object when a project can spend months moving among a utility, transmission owner, county, water authority, air agency, and other offices without a clear sequence. But the remedy is not to waive review for projects below an arbitrary size. It is to define the evidence that makes a project ready for review and the commercial consequences that follow once capacity is reserved. | **From Site Control to Firm Power: What Data Center Development Looks Like on the Ground** **Rich Demon, CEO East Energy Renewables** East Energy Data, an affiliate of East Energy Renewables, entered data center development with an advantage many newcomers lack: years of experience building energy infrastructure in rural communities. Its agricultural-waste-to-energy projects have sometimes been controversial, but the company has built durable local relationships around all four of its operating sites. When EED began pursuing data centers in early 2025, it therefore expected community acceptance to be manageable. The company deliberately avoided hyperscale projects. It targeted facilities of 8 to 49 MW, generally near its existing operations, and tried to size each project around power that could be delivered without major transmission upgrades. Preferred sites were established industrial areas requiring less than 50 acres, with substantial buffers, reclaimed wastewater for cooling where available, and onsite natural-gas generation for redundancy. The first projects showed how difficult even that approach can be. At one site, EED withdrew after encountering sustained public opposition. The town adopted a one-year data center moratorium a week later. At a second site, the company went further, spending approximately $35,000 on a utility impact study, limiting the project to 8 MW to fit the capability of a nearby substation, and beginning land procurement. That city also enacted a moratorium, and EED shelved the project. A third project has followed a different path. In mid-2025, EED selected a 160-acre site in a newly designated rural industrial zone, with more than 100 acres retained as forested buffer. It executed a memorandum of understanding with the county and initiated a roughly $170,000, three-part utility impact study involving the local electric cooperative and the larger organizations responsible for supplying it. The study subsequently indicated that approximately 15 MW could be available within 12 months and 50 MW within 36 months. Yet even then the project was not secure. As EED's site control expired, two other parties sought control of the surrounding 600 acres for hyperscale development, and the county initially declined to renew EED's agreement. EED responded by making the development case in local terms: a smaller footprint, separation from neighboring properties, potentially more than $600 million of investment at 50 MW, local employment, community-college apprenticeships, commitments to hire and promote locally, and recurring support for a local service organization. The county ultimately renewed the MOU, creating a path toward an option and eventual purchase if the remaining development work is completed. The experience illustrates why acreage and nominal electric capacity are poor proxies for data center readiness. A viable project requires power that can actually be delivered on an acceptable schedule, local institutions willing to support the development, and community benefits residents regard as credible. It also exposes a broader industry problem. The rush to control large parcels can begin before developers understand power availability or local conditions. Those speculative efforts create expectations, competition, and sometimes opposition that affect every project that follows. For EED, the lesson has been clear: community engagement is no longer a workstream that follows site selection. It has become part of the infrastructure required to make a data center project executable. | | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | AEP Ohio provides a useful example of that architecture. Its data center tariff became effective July 23, 2025\. For requests of 25 megawatts or more, a customer must control the property, provide a specific location, load ramp, and final load, and pay a study fee. AEP Ohio says it will make reasonable efforts to complete load studies within 45 days when regional transmission upgrades are not needed and 60 days when they are. Those are stated effort targets, not guarantees of energization.6 After the study, AEP Ohio presents a Letter of Agreement and an Electric Service Agreement. The Letter of Agreement can require reimbursement of 100 percent of buildout costs if the customer cancels or delays the project by more than 12 months before the target energization date. The service agreement begins on actual energization, not on a forecast date.7 The tariff also converts a forecast into a longer-term commitment. The ramp may not exceed four years; contract capacity floors rise during that period; the initial contract term equals the ramp period plus eight years; collateral can equal 50 percent of total minimum charges for customers that do not satisfy specified credit and liquidity tests; and minimum billing provisions continue to apply even when actual demand is lower.8 Those percentages are not a universal template. Their importance is institutional. A utility that is asked to reserve scarce capacity and trigger infrastructure construction can require evidence of readiness and place the risk of cancellation, delay, and underutilization on the customer that caused the commitment. ## Contract the Load the Grid Must Actually Serve A community-scale pathway should therefore center on the quantity that matters to electric service: maximum grid withdrawal at the point of interconnection, together with a staged ramp and the conditions under which that withdrawal may change. Nameplate campus size remains relevant, but it is not a substitute for the load the utility must plan to serve. On-site generation complicates that calculation. A project that proposes 60 megawatts of computing load and 25 megawatts of behind-the-meter generation may expect to withdraw only 35 megawatts in normal operation. That does not mean the grid can automatically be studied as a permanent 35 megawatt load. The contract has to address generator outages, transition states, export, islanding, protection, and the equipment that prevents net grid withdrawals from exceeding the agreed limit. AEP Ohio’s tariff illustrates the distinction. When behind-the-meter generation is used for netting, the customer must maintain equipment that can instantaneously curtail load by at least the output of the behind-the-meter generation so that loss of that generation does not cause the customer to exceed contract capacity. The same tariff aggregates certain affiliated or commonly owned loads when applying its minimum-demand rules and can aggregate multiple facilities at a single location when their combined load crosses the 25 megawatt threshold.9 That is a stronger model than a carveout. It governs the system that will actually exist, rather than the way a developer divides parcels, phases buildings, or describes the campus. ## Flexibility as a Performance Obligation Data centers may have batteries, deferrable computing, workload migration, or on-site generation. None of those capabilities has system value merely because it exists. Flexibility becomes useful to the grid when the contract identifies the trigger, notice, baseline, response time, duration, rebound behavior, measurement method, test regime, and consequence for nonperformance. The economics are also tariff-specific. In MISO, the current planning-reserve-margin requirement calculation uses a coincident-peak-demand forecast as an input. MISO is separately considering reforms that would allocate resource-adequacy obligations using hours that better reflect reliability risk rather than relying only on the current peak-based approach.10 In ERCOT, Four Coincident Peak calculations still identify each distribution service provider’s load during the system peak intervals for June, July, August, and September. Texas is also reviewing transmission-cost allocation, so the durability of today’s 4CP incentives should not be assumed. Whether and how those wholesale signals reach an individual data center customer depends on the applicable retail tariff or contract.11 A hypothetical 40 megawatt project that can curtail 20 megawatts might create meaningful value if it performs during the interval that sets a cost or capacity obligation. The same capability may be worth far less if it responds in the wrong hour, rebounds into another constraint, or is unavailable during the system event that matters. No publication-grade dollar value can be derived from nameplate demand and an average dollars-per-kilowatt-month charge alone. A defensible structure is therefore two-part: firm service for the demand the utility must stand ready to supply, and a separately defined flexible block that earns a credit, accelerated treatment, or other benefit only after it is tested. Repeated nonperformance should remove the benefit or increase security if the governing tariff or contract authorizes that remedy. ## Behind the Meter Does Not Mean Outside the Rules The much larger Susquehanna nuclear example shows why proximity to generation is not a jurisdictional shortcut. Talen Energy’s first-quarter 2026 filing says its revised agreement with Amazon Web Services is expected to provide up to 1,920 megawatts through 2042 in a front-of-the-meter model and that the transition to the revised agreement occurred in April 2026.12 The project is not a community-scale analogue. It is evidence for a narrower proposition: a load located next to generation still has to be fitted into transmission, service, and reliability arrangements. The same caution applies to the Public Utility Regulatory Policies Act. FERC describes a qualifying small power production facility as a generating facility of 80 megawatts or less whose primary energy source is renewable energy, biomass, waste, or geothermal resources, subject to the agency’s site and certification rules. Qualifying cogeneration is a different category and requires sequential production of electricity or mechanical energy and useful thermal energy under applicable standards.13 A conventional natural-gas generator does not become a PURPA small power production qualifying facility merely because it is below 80 megawatts. A project may have a separate cogeneration case, but that requires a project-specific legal and engineering analysis. In either case, qualifying-facility status does not erase every interconnection, retail-service, or environmental requirement.14 Air permitting is similarly equipment- and jurisdiction-specific. The US Environmental Protection Agency notes that most Clean Air Act permitting programs are administered by state and local air agencies, while stationary engines and turbines can also be subject to federal New Source Performance Standards or hazardous-air-pollutant requirements depending on the equipment and operating profile.15 ## Community Benefits Should Survive the Announcement Community benefits need the same discipline as power commitments. Taxes owed under law, construction payroll, permanent employment, infrastructure built primarily for the customer, and negotiated community contributions are different categories. They should be reported separately. New York’s Executive Order 62 directed Empire State Development to create a Community Investment Framework for local negotiations. The policy outline released with the order describes community funds, infrastructure, labor, public services, transparency, and other possible commitments. It also floated a recommended negotiation baseline of $1 million per megawatt of anticipated utility demand. That figure is a policy proposal and negotiation starting point, not a generally binding statewide payment requirement.16 The order separately directed the Department of Public Service to form a data center interconnection working group around 'beneficiary pays' principles and to consider mechanisms that could protect other customers from stranded-grid costs. Community investment and electric-system cost responsibility are related policy questions, but they are not the same transaction.17 Linn County, Iowa, offers a local example of converting some obligations into rules. In February 2026, the county adopted requirements for large data centers in unincorporated Linn County that include a water study, a water-use agreement with monitoring and reporting, and an economic-development agreement outlining community benefits. Effective July 1, the county then imposed an 18-month moratorium on accepting new applications to rezone property to its EU-3 Large-Scale Data Center Zoning District while it reviews cumulative impacts. The moratorium does not apply inside incorporated cities.18 The lesson is not that every county should copy Linn County or that New York’s proposed contribution level is the right number. It is that community promises become credible when they name a beneficiary, amount or formula, date, metric, reporting duty, enforcement authority, and consequence for failure—and when those obligations survive a sale or material change in the project. ## The Earned-Predictability Package A smaller project should be eligible for a standardized or expedited track only after it becomes measurable. The package below is a proposed governance design, not a statement that every jurisdiction already has authority to impose every term. Exact remedies must be grounded in the applicable tariff, contract, permit, ordinance, and state or federal law. | Commitment | Required evidence and ongoing verification | Potential remedies for nonperformance | | ---------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------- | | **Project readiness** | Documented site control, a project-specific load ramp, an equipment procurement plan, financing milestones, and deadlines for meeting readiness requirements. | Loss of queue position or reserved capacity, where permitted by the governing process. | | **Grid demand and load ramp** | Studies covering expected demand at the point of interconnection and each stage of the load ramp; metering and telemetry to verify actual demand. | Deferral of capacity increases, curtailment, additional financial security, or other service remedies. | | **Cost responsibility** | Study deposits, upgrade cost estimates, collateral, minimum billing obligations, and explicit allocation of cancellation and stranded-cost exposure. | Draws on financial security, exit charges, or reimbursement of assigned costs. | | **Committed load flexibility** | A defined curtailment quantity, activation trigger, baseline, notice period, response time, duration, and rebound plan; performance testing and event telemetry. | Reduction or withdrawal of flexibility credit or expedited treatment; additional financial security where authorized. | | **Power quality and on-site generation** | Equipment models, ride-through requirements, protection settings, export limits, procedures for outages and transitions between grid and on-site supply, commissioning tests, and notification of material changes. | Required mitigation, operating limits, restudy, or other reliability remedies. | | **Community commitments** | Named beneficiaries; payment amounts or formulas; measurable employment and water-use commitments; reporting, audit, and financial security provisions; and obligations that survive changes in ownership or control. | Draws on escrow, clawbacks, damages, or permit-related remedies expressly established by the governing instrument. | In return, the project should receive something concrete: a standard application, a defined study track, published evidence requirements, target schedules or written reasons for delay, staged energization where technically and legally feasible, and clear transition rules when policy changes. Predictability is valuable. It should be exchanged for commitments that make the project easier to study and safer for other customers to rely on. A size threshold can still be part of that regime. Below a stated level, projects that satisfy impact screens could enter an expedited track. A project that fails a screen would move to fuller review rather than automatically die. Larger projects with unusually strong controls could receive staged treatment without being deemed harmless. That is the useful meaning of community scale: not a legal safe harbor, but a development strategy in which the project accepts bounded, locally legible obligations. The developer gets a clearer path. The utility gets a credible load forecast, security, and operational limits. The community gets measurable commitments instead of a benefits list. Regulators keep the authority to look at cumulative effects. The distinction is between a customer that asks everyone else to plan around a forecast and a project that makes the forecast enforceable. A carveout says the number is small enough to trust. A contract says what will be measured, who carries the risk, and what happens when the project does something different. That is the harder bargain. It is also the one most likely to make smaller data centers genuinely easier to approve. ## Sign up for AIxEnergy The AIxEnergy Journal provides executive intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, reliability, investment, and policy. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## Notes **1\.** Office of the Governor of New York, “No. 62: Establishing a Temporary Moratorium on Data Centers in New York While the State Develops Higher Standards for Data Center Development and Benefits Blueprint to Support Localities,” Executive Order 62, July 14, 2026, https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data centers-new-york-while-state-develops. **2\.** AEP Ohio, “Data Center Tariff,” effective July 23, 2025, accessed August 17, 2026, https://www.aepohio.com/company/about/rates/data center-tariff/; Electric Reliability Council of Texas, “Large Load Integration,” Batch Zero Interconnection Study Process, accessed August 17, 2026, https://www.ercot.com/services/rq/large-load-integration; North American Electric Reliability Corporation, “Project 2026-02 Computational Loads,” accessed August 17, 2026, https://www.nerc.com/standards/reliability-standards-under-development/2026-02-computational-loads; NERC, “FERC Sets Year-End Deadline for NERC to Finalize Registry Criteria and Standards for Computational Loads,” July 16, 2026, https://www.nerc.com/newsroom/ferc-sets-year-end-deadline-for-nerc-to-finalize-registry-criteria-and-standards-for-computational-loads. **3\.** US Department of Energy, “DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers,” December 20, 2024, summarizing Lawrence Berkeley National Laboratory, 2024 United States Data Center Energy Usage Report, https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data centers. **4\.** Federal Energy Regulatory Commission, “FERC Launches Aggressive Targeted Action to Speed Large Load Integration,” June 18, 2026, https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration; FERC, “Fact Sheet | FERC Takes Action to Supercharge America’s Grid for Efficiency, Reliability, and a Bold Energy Future,” June 18, 2026, https://www.ferc.gov/news-events/news/fact-sheet-ferc-takes-action-supercharge-americas-grid-efficiency-reliability-and. **5\.** North American Electric Reliability Corporation, “Computational Load Modeling, Studies, Instrumentation, Commissioning, Operations, Protection, and Control,” Level 3 Essential Actions Alert, May 4, 2026, https://www.nerc.com/globalassets/programs/bpsa/alerts/level-3-computational-load-alert.pdf; NERC, “FERC Sets Year-End Deadline for NERC to Finalize Registry Criteria and Standards for Computational Loads,” July 16, 2026, https://www.nerc.com/newsroom/ferc-sets-year-end-deadline-for-nerc-to-finalize-registry-criteria-and-standards-for-computational-loads. **6\.** AEP Ohio, “Data Center Tariff,” “Mandatory Process for Customers Interested in Moving Forward with Data Center Projects in AEP Ohio’s Service Territory,” accessed August 17, 2026, https://www.aepohio.com/company/about/rates/data center-tariff/. **7\.** AEP Ohio, “Data Center Tariff,” steps 3–4 of the Schedule DCT service-request process, accessed August 17, 2026, https://www.aepohio.com/company/about/rates/data center-tariff/. **8\.** AEP Ohio, “Data Center Tariff,” “Load Ramp Period,” “Contract Term,” “Collateral Requirement,” and “Minimum Demand Charges,” accessed August 17, 2026, https://www.aepohio.com/company/about/rates/data center-tariff/. **9\.** AEP Ohio, “Data Center Tariff,” “Aggregation” and “Behind-the-Meter (BTM) Generation,” accessed August 17, 2026, https://www.aepohio.com/company/about/rates/data center-tariff/. **10\.** Midcontinent Independent System Operator, “How Is Planning Reserve Margin Requirement (PRMR) Calculated,” updated July 6, 2026, https://help.misoenergy.org/knowledgebase/article/KA-01065/en-us; MISO, “RASC: PRMR Allocation Design (RASC-2019-2 and 2020-4),” April 15, 2026, https://www.misoenergy.org/engage/stakeholder-feedback/2026/rasc-prmr-allocation-design-rasc-2019-2-and-2020-4-20260415/. **11\.** Electric Reliability Council of Texas, “ERCOT Four Coincident Peak Calculations,” accessed August 17, 2026, https://www.ercot.com/mktinfo/data\_agg/4cp; Public Utility Commission of Texas, Project No. 58484, “Evaluation of Transmission Cost Recovery,” filings through August 17, 2026, https://interchange.puc.texas.gov/Search/Filings?ControlNumber=58484. **12\.** Talen Energy Corporation, Quarterly Report on Form 10-Q for the quarter ended March 31, 2026, definition of “AWS PPA,” filed with the US Securities and Exchange Commission, https://www.sec.gov/Archives/edgar/data/1622536/000162253626000036/tln-20260331.htm. **13\.** Federal Energy Regulatory Commission, “PURPA Qualifying Facilities,” accessed August 17, 2026, https://www.ferc.gov/qf; see also 18 C.F.R. §§ 292.203–292.207. **14\.** FERC, “PURPA Qualifying Facilities.” Qualifying-facility status and the federal QF categories do not by themselves eliminate project-specific interconnection, retail-service, state-implementation, or environmental requirements. https://www.ferc.gov/qf. **15\.** US Environmental Protection Agency, “Clean Air Act Resources for Data Centers,” accessed August 17, 2026, https://www.epa.gov/stationary-sources-air-pollution/clean-air-act-resources-data centers. **16\.** Empire State Development, “Policy Outline for the New York State Data Center Community Investment Framework,” July 14, 2026, https://esd.ny.gov/sites/default/files/media/document/CIF-Policy-Outline-071426.pdf. **17\.** Office of the Governor of New York, Executive Order 62, sections 3–4, July 14, 2026, https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data centers-new-york-while-state-develops. **18\.** Linn County, Iowa, “Data Centers in Unincorporated Linn County,” updated 2026, accessed August 17, 2026, https://www.linncountyiowa.gov/1862/Data centers-in-Unincorporated-Linn-Coun. ### Dependable Enough to Plan On URL: https://www.aixenergy.io/dcflex/ Last updated: 2026-10-05T05:26:58.000Z Global electricity demand is expected to grow by an average of 3.6 percent annually between 2026 and 2030, adding about 1,100 terawatt-hours each year, roughly 50 percent more annual demand growth than over the previous decade, according to the International Energy Agency (IEA). Through 2030, electricity consumption is projected to grow at least 2.5 times as fast as overall energy demand, underscoring how much this moment is being driven by electrification and not simply broader economic growth¹. Article Summary 0:00 /64.313469 1× This growth is driven by rising industrial use of electricity, accelerating uptake of electric vehicles², expanded air conditioning use, and the AI-fueled increase in data center power demand³. In the United States, which is leading the global data center surge, EPRI scenarios estimate that data centers could account for roughly 9 to 17 percent of US electricity consumption by 2030, up from about 4 to 5 percent in 20244. We are moving at speed into the age of electricity. The question that is top of mind for many is, “How will we meet this moment?” ## **Under what conditions can flexibility create more headroom?** As demand from AI and data centers grows at unprecedented speed, flexibility is becoming the third leg of the speed-to-power stool, alongside generation and transmission.⁵ Flexibility allows large loads to reduce or shift consumption during periods of grid stress through various methods, including on-site generation, on-site battery storage, on-site thermal storage, standby generation, uninterruptible power supply, and/or shifting of chiller and compute workloads. These mechanisms are not interchangeable: they differ in speed, duration, notice, location, emissions profile, and contractual form. The question is whether flexibility can create additional system headroom and, in some cases, reduce or defer infrastructure needs when the response is available at the right location, time, duration, and level of reliability. Under what conditions can flexibility also shorten interconnection timelines and accelerate time-to-power, subject to the utility study and the contractual arrangement that governs the commitment? The industry has moved beyond asking whether selected data center loads can flex toward understanding what makes a specific flexibility commitment dependable enough to count in a planning study that carries a compliance obligation. We propose five criteria for evaluating whether a flexibility commitment is dependable enough to receive planning credit: **Firm.**The obligation to reduce must be contractual, not discretionary, with defined depth, duration, notification time, and frequency, and with consequences for non-performance. **Deliverable.**The reduction must demonstrably relieve the relevant electrical constraint. System-level flexibility does not automatically translate into relief of a specific post-contingency overload. **Coincident.**The commitment must be available in the hours and under the conditions in which the constraint binds, which are not necessarily the hours of highest system load. **Verifiable.**Performance must be measured against a defined baseline, reported, and auditable, on the same footing as any other resource on which reliability depends. **Modeled.**The load must be represented in planning and operational models with validated dynamic behavior, including response to voltage and frequency disturbances and recovery or rebound demand following a flexibility event. These criteria are a practical test for evaluating planning credit. They also align with the direction of recent reliability work. In May 2026, NERC issued a Level 3 Alert addressing computational-load modeling, studies, instrumentation, commissioning, operations, protection, and control. In July, FERC directed NERC to develop and submit new or modified reliability standards and registry revisions for computational loads by December 31, 2026.⁶ That record reinforces why demonstration success and planning credit are related, but not the same. ## **When is flexibility real?** When EPRI’s Data Center Flexible Load Initiative (DCFlex) was founded in November 2024, a central question was how data center flexibility could be demonstrated, repeated, and incorporated into utility planning. Since then, DCFlex has advanced a portfolio of demonstrations at data center facilities around the world⁷. Exhibit 1 Evidence From the Field Data Center Flexibility Is Moving Into Real-World Operations The evidence is beginning to move beyond modeling and controlled laboratory work into operating data centers and commercial utility programs. Phoenix, Arizona — AI workload flexibility In May 2025, EPRI, working with Emerald AI, tested a software-based approach that enables AI data centers to operate as flexible grid resources. The demonstration used a 256-GPU cluster running representative AI workloads in a hyperscale cloud facility and reduced power consumption by 25 percent for three hours during a Salt River Project system peak, while maintaining AI quality-of-service guarantees.8 Lenoir, North Carolina — Utility-integrated demand response EPRI’s DCFlex demonstration with Google and Duke Energy is examining computational flexibility alongside the operational and communications protocols needed to make data-center demand response usable by utilities.9 Google has since announced demand-response agreements with multiple U.S. utilities representing roughly 1 GW of contracted capability—commercial arrangements that extend the concept beyond any single pilot site.10 Beyond individual sites — Multiple forms of flexibility Other DCFlex demonstrations are testing a broader range of operating strategies. In Chicago, ComEd, Constellation, Emerald AI, and NVIDIA are evaluating data-center flexibility in a major power market. Near London, National Grid, Nebius, and Emerald AI observed load reductions of up to 40 percent on evaluated computing clusters, with tests ranging from several minutes to nine hours.11 Other work has examined geospatial workload shifting between Ashburn, Virginia, and Chicago, as well as lower-carbon backup generation. At a Compass data center near Dallas, a field evaluation found hydrotreated vegetable oil produced approximately 40 percent lower hydrocarbon emissions than diesel under U.S. Environmental Protection Agency reference methods, while lifecycle analysis indicated roughly 80 percent lower greenhouse gas intensity than fossil diesel. These demonstrations illustrate a range of flexibility pathways across compute, facility systems, and backup power and show that flexibility strategies can be tailored to the availability, scale, and duration of grid events. Public project materials document meaningful load response under tested conditions while maintaining service-level commitments for the evaluated workloads. Detailed metrics should be read from the underlying EPRI and partner project reports, including EPRI’s DCFlex technical publications. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/09/aixenergy_dcflex_map_print-1.png) ## **Moving from pilots to scale** While DCFlex’s demonstrations have highlighted the benefits of flexibility, they are only the first step. The utility and data center developer industries still need a common language to define flexibility. That’s why, in March 2026, DCFlex launched [Flex MOSAIC™](https://dcflex.epri.com/flex-mosaic?ref=aixenergy.io), collaborating with more than 70 utilities, system operators, technology providers, and hyperscalers12. The Flex MOSAIC framework defines five classes of flexibility based on the magnitude, timing, duration, and frequency of a load’s response. The purpose of establishing these defined classes is to replace bespoke, project-by-project interconnection negotiations with shared, performance-based descriptions. Without a common vocabulary, every interconnection is a custom negotiation, which can slow down speed-to-power. Leveraging insights from the Flex MOSAIC framework, the DCFlex team is developing tools for base features and reference-design support elements. By year-end 2026, the team expects to roll out the final reference design, as well as structural incentive-program designs and tools for adoption of the framework. By establishing a shared language, transparent facility performance expectations, and repeatable grid responses, the framework can help utilities, system operators, regulators, and data center developers make faster, more consistent, and more confident decisions when evaluating flexibility. The DCFlex team recently released an interactive simulation on its website where users can test the usefulness of Flex MOSAIC from the standpoint of a data center developer, interconnection-study engineer, tariff/program designer, or long-term planner. I invite you to try the simulation and to move from simulations to reality by applying the lessons learned across the energy ecosystem. ## **Flexibility and speed to power** Flexibility can reduce the need for premature investment in additional power plants and transmission lines if the industry has tools to evaluate when that potential is real. As part of the DCFlex effort, the EPRI team developed a practical industry framework to help power-system planners evaluate how much additional load, particularly from rapidly growing data centers, can be integrated without expanding generation, storage, or transmission infrastructure13. The [Headroom Framework](https://headroom.epri.com/?ref=aixenergy.io) provides a uniform, stepwise approach to evaluate how increasing data center flexibility, via Flex MOSAIC flexibility classes, may contribute usable headroom under specific system conditions: resource adequacy, transmission topology, operating constraints, contingencies, timing, and location. At EPRI, we believe these tools can improve information exchange between system planners and data center developer communities about where system flexibility is most valuable and what type of flexibility is prioritized. With the right signaling, connection requests become more aligned with system needs, which can streamline the interconnection process. ## **Collaboration and the path to scale** Thanks to DCFlex and its more than 70 collaborators, including hyperscalers, utilities, independent system operators, power producers, technology providers, consultants, and finance stakeholders, what exists today is a shared language integrated with demonstrations leading to flexibility strategies and dynamic planning tools. Even with these advances, a trust gap remains. Because DCFlex demonstration projects have validated the physics of flexibility under defined conditions, stakeholders can use a common language from the earliest stages of negotiations and interconnection planning through rate design and long-term planning. The next step is turning individual results into a track record: aggregating dispatched-event outcomes across many facilities and evaluating them against consistent baselines, the same way generation resources have been assessed for decades. That is how flexibility earns trust and a place in planning, not just in pilots. As DCFlex progresses, large-load flexibility is becoming more viable as a planning resource. The path to fuller utilization will require ongoing industry-wide collaboration and recognition that gains in efficiency can override one-off solutions created by individual utilities and hyperscalers, making outcomes more predictable, effective, and affordable. We have shown that data centers can flex. The five-part test offers a practical way to determine if that flexibility is dependable. That is how we meet this moment: with flexibility that is not just possible, but dependable enough to plan on. [*Arshad Mansoor*](https://www.linkedin.com/in/arshad-mansoor-a2090b165/?ref=aixenergy.io) *is president and CEO of* [*EPRI*](https://www.epri.com/?ref=aixenergy.io)*. The institute is an independent, nonprofit energy research and development organization. EPRI does not advocate for policy or commercial outcomes.* ## Sign up for AIxEnergy The AIxEnergy Journal provides executive intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, reliability, investment, and policy. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## **Notes** ¹ International Energy Agency (IEA), “Demand,” in *Electricity 2026* (IEA, February 2026), [https://www.iea.org/reports/electricity-2026/demand](https://www.iea.org/reports/electricity-2026/demand?ref=aixenergy.io). ² IEA, *Global EV Outlook 2026* (IEA, May 2026), [https://www.iea.org/reports/global-ev-outlook-2026](https://www.iea.org/reports/global-ev-outlook-2026?ref=aixenergy.io); IEA, “Demand,” in *Electricity 2026*, [https://www.iea.org/reports/electricity-2026/demand](https://www.iea.org/reports/electricity-2026/demand?ref=aixenergy.io). ³ IEA, *World Energy Outlook 2024* (IEA, 2024), “Executive Summary,” [https://www.iea.org/reports/world-energy-outlook-2024/executive-summary](https://www.iea.org/reports/world-energy-outlook-2024/executive-summary?ref=aixenergy.io); IEA, “The World Is Moving at Speed into the Age of Electricity,” accessed September 14, 2026, [https://www.iea.org/spotlights/the-world-is-moving-at-speed-into-the-age-of-electricity](https://www.iea.org/spotlights/the-world-is-moving-at-speed-into-the-age-of-electricity?ref=aixenergy.io). 4 Electric Power Research Institute (EPRI), *Powering Intelligence 2026: Updated Scenarios of U.S. Data Center Electricity Use and Power Strategies* (EPRI, February 2026), [https://powering-intelligence.epri.com/](https://powering-intelligence.epri.com/?ref=aixenergy.io). ⁵ Zachary Skidmore, “EPRI Launches Data Center Flexibility Framework to Speed Up Grid Connections,” *Data Center Dynamics*, April 23, 2026, [https://www.datacenterdynamics.com/en/news/epri-launches-data-center-flexibility-framework-to-speed-up-grid-connections/](https://www.datacenterdynamics.com/en/news/epri-launches-data-center-flexibility-framework-to-speed-up-grid-connections/?ref=aixenergy.io). ⁶ North American Electric Reliability Corporation (NERC), “Computational Load Modeling, Studies, Instrumentation, Commissioning, Operations, Protection, and Control,” Level 3 Essential Action Alert, May 4, 2026, [https://www.nerc.com/globalassets/programs/bpsa/alerts/level-3-computational-load-alert.pdf](https://www.nerc.com/globalassets/programs/bpsa/alerts/level-3-computational-load-alert.pdf?ref=aixenergy.io); Federal Energy Regulatory Commission, “Reliability Standard(s) Pertaining to Computational Load Integration,” Docket No. RD26-7-000, July 16, 2026, [https://www.ferc.gov/media/e-1-rd26-7-000](https://www.ferc.gov/media/e-1-rd26-7-000?ref=aixenergy.io). ⁷ EPRI, “DCFlex Demonstrations,” accessed September 14, 2026, [https://dcflex.epri.com/demonstrations](https://dcflex.epri.com/demonstrations?ref=aixenergy.io) ⁸ Philip Colangelo et al., “AI Data Centres as Grid-Interactive Assets,” *Nature Energy* 11, no. 2 (2026): 254–61, [https://doi.org/10.1038/s41560-025-01927-1](https://doi.org/10.1038/s41560-025-01927-1?ref=aixenergy.io). ⁹ EPRI Journal, “Flexible Loads, Resilient Grids,” August 20, 2025, [https://eprijournal.com/flexible-loads-resilient-grids/](https://eprijournal.com/flexible-loads-resilient-grids/?ref=aixenergy.io). 10 Michael Terrell, “A New Milestone for Smart, Affordable Electricity Growth,” Google, March 19, 2026, [https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/demand-response-data-center-milestone/](https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/demand-response-data-center-milestone?ref=aixenergy.io). 11 EPRI, *Data Center Flexible Load Initiative (*[*DCFlex)*](https://dcflex.epri.com/flex-mosaic?ref=aixenergy.io)*: Flex MOSAIC and Demonstration Summary*, n.d., 3–4, accessed September 14, 2026, [https://restservice.epri.com/publicattachment/98143](https://restservice.epri.com/publicattachment/98143?ref=aixenergy.io). 12 EPRI, “An Open Letter to the Power and Digital Infrastructure Industries: Accelerating Time-to-Power for Data Centers and Large Loads Through a Flexibility Framework,” accessed September 14, 2026, [https://dcflex.epri.com/flex-mosaic/open-letter](https://dcflex.epri.com/flex-mosaic/open-letter?ref=aixenergy.io). 13 EPRI, *A Proposed Framework to Assess Headroom for Integrating Data Centers into Regional Power Systems*, report no. 3002036339 (EPRI, 2026), [https://headroom.epri.com/](https://headroom.epri.com/?ref=aixenergy.io). ### AI in Grid Planning: The Burden of Proof URL: https://www.aixenergy.io/ai-in-grid-planning-the-burden-of-proof/ Last updated: 2026-09-27T21:16:27.000Z AI can accelerate planning work, but speed is not an evidentiary standard. The more an AI output can shape an infrastructure decision, the more utilities should require traceability, independent validation, reproducibility, controlled change, and accountable engineering judgment. _This post is for subscribers only._ ### Capital Is Sorting the AI Power Buildout by Risk URL: https://www.aixenergy.io/capital-is-sorting-the-ai-power-buildout-by-risk/ Last updated: 2026-09-27T18:18:50.000Z The financing boom is real, but venture rounds, project loans, corporate debt, and behind-the-meter partnerships are funding different risks. That distinction will help determine which megawatts arrive first. _This post is for subscribers only._ ### The Power Crunch Is Moving Inside the Data Center URL: https://www.aixenergy.io/the-power-crunch-is-moving-inside-the-data-center/ Last updated: 2026-09-27T18:19:00.000Z AI rack density is forcing a redesign of the electrical path from utility service to accelerators, with consequences for usable power, equipment choices and project economics. _This post is for subscribers only._ ### New York and Texas Are Rewriting the Terms of Data Center Growth URL: https://www.aixenergy.io/new-york-and-texas-are-rewriting-the-terms-of-data-center-growth/ Last updated: 2026-09-27T21:16:49.000Z New York and Texas have both put the brakes on parts of the data center development process. For an industry accustomed to competing on the speed of construction and access to power, that is a consequential change. It also deserves a more careful reading than the word moratorium usually receives. The two states are using different authorities, interrupting different approvals, and asking different questions about what should happen before projects proceed.1 New York offers the more comprehensive approach to the relationship between a data center and the place that hosts it. Its policy combines environmental assessment, electricity infrastructure planning, and guidance for negotiating community investments. Texas is using access to the Electric Reliability Council of Texas (ERCOT) grid to obtain and verify information about proposed projects and their effects. Both approaches recognize a basic economic problem: private investment decisions can create public infrastructure obligations well before the promised economic benefits arrive. The strongest case for New York is the way these responsibilities are being considered together. A data center can be a valuable investment, but the size of its construction budget does not establish whether it is a good investment for its host community. That requires examining what the project consumes, what it contributes, who pays for the infrastructure, and who remains responsible if the original development plan changes. New York is creating a process in which those questions can shape the terms of development. The relevant measure of success is durable economic value supported by infrastructure that works. Announced megawatts and development spending are useful indicators of commercial interest. They are inadequate measures of whether a project will benefit the people who live around it. Governor Kathy Hochul issued Executive Order 62 on July 14\. Its immediate restriction applies to discretionary Department of Environmental Conservation approvals for construction or expansion of covered data centers whose applications DEC had not deemed complete before that date. The definition generally reaches facilities, including groups on the same or contiguous sites, capable of consuming at least 50 megawatts. Specified manufacturing, research, education, and medical uses are excluded. Local government approvals are outside the permitting hold.2 New York has not ordered every data center to stop operating, and the order is not a universal prohibition on every form of construction or approval. The governor described the pause as lasting up to a year. The operative order ties its conclusion to submission of the final Generic Environmental Impact Statement and findings; its separate water review carries a twelve-month reporting deadline. A project schedule should reflect that distinction.3 Texas took a different route on August 3\. Governor Greg Abbott directed the Public Utility Commission of Texas and ERCOT to verify and audit data centers advancing through ERCOT’s interconnection process before additional projects receive approval to move forward. The directive makes compliance a condition of connection. Its immediate point of control is the interconnection process, rather than a New York-style hold on specified state environmental permits.4 A developer needs to know which approval is affected, which authority controls it, and what evidence or administrative action allows the next step. Treating both states as a single category marked “closed” conceals those differences. It also obscures the opportunity for a serious developer to use the intervening period to improve its project. ## Why Project Credibility Matters Texas’s information request now has a defined scope and schedule. ERCOT is seeking responses from developers of covered data centers of at least 25 megawatts pursuing interconnection, with responses due October 12 and a public report planned by December 10\. Its September 18 reference guide includes virtual currency mining and identifies covered facilities that have not yet energized. The reporting milestone should not be mistaken for a guaranteed date of connection or approval.5 The Texas review reaches well beyond the engineering of a grid connection. It asks about public subsidies, ownership, electricity requirements, onsite generation, water sources, cooling technology, and effects on neighboring communities. Abbott reinforced the water component on September 14 by directing the Texas Water Development Board to enforce reporting requirements and coordinate with ERCOT.6 This is useful because infrastructure planning becomes less reliable when a utility or grid operator cannot establish what a proposed customer intends to build, how it will operate, or how much of its stated demand is credible. Better information can help distinguish a funded development from an option on future capacity. It can also expose assumptions about water availability or onsite power that have not yet been converted into executable plans. The scale of Texas’s reported queue helps explain the urgency. Abbott’s August announcement cited approximately 474 gigawatts of connection requests, with data centers accounting for roughly 90 percent of the new power requests. Those figures describe requests, not a forecast of realized demand. Reading them as inevitable electricity consumption would confuse developer interest with infrastructure that will actually be built.7 AIxEnergy | POLICY COMPARISON ### New York and Texas: Two Different Pauses New York links environmental review, grid planning, and community investment. Texas uses the ERCOT interconnection process to verify projects and their impacts. Swipe horizontally to compare both states. | Policy dimension | New York | Texas | | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Policy instrument | **Executive Order 62, July 14, 2026.** A targeted DEC permitting hold linked to a broader state review.[1](#aixe-source-1) | **Governor’s directive, August 3, 2026.** Verification and audit of data centers advancing through ERCOT interconnection before additional approvals to move forward.[2](#aixe-source-2) | | Approval affected | Specified discretionary DEC approvals for construction or expansion when applications were not deemed complete before July 14\. Local approvals are outside the hold.[1](#aixe-source-1) | Advancement through the ERCOT interconnection process. This is a grid-access intervention, with compliance required before connection.[2](#aixe-source-2) | | Coverage and scale | Generally, covered facilities or groups on the same or contiguous sites capable of consuming **50 MW or more**. Specified manufacturing, research, education, and medical uses are excluded.[1](#aixe-source-1) | The implementing information request covers specified, not-yet-energized facilities of **25 MW or more**, including virtual currency mining. This is the RFI threshold, not an equivalent statewide permitting definition.[3](#aixe-source-3) | | Timing and next milestone | The order ties the end of the hold to submission of the final Generic Environmental Impact Statement and findings. A separate water review has a twelve-month reporting deadline.[1](#aixe-source-1) | RFI responses are due **October 12, 2026**; a public report is planned by **December 10**. The report date is not a guaranteed approval or connection date.[3](#aixe-source-3) | | Scope of review | A public environmental review considers energy, water, air quality, noise, and disproportionate effects on disadvantaged communities, coordinated with the state’s large-load proceeding.[1](#aixe-source-1) | Project-level information includes ownership, subsidies, electricity needs, onsite generation, water, cooling, and neighboring-community impacts.[3](#aixe-source-3) | | Ratepayer protection | The broader policy effort considers a Grid Acceleration Fund, clean supply, and safeguards against speculative-load infrastructure costs. These mechanisms remain under development or consideration.[4](#aixe-source-4) | The audit complements Senate Bill 6, which addresses financial commitments, site control, similar service requests, interconnection cost recovery, and certain emergency obligations.[5](#aixe-source-5) | | Host-community investment | A voluntary Community Investment Framework recommends **$1 million per MW of utility demand** as a negotiating benchmark. Guidance includes state-required grid modernization investments within that benchmark; it is not an automatic statewide fee.[6](#aixe-source-6) | The audit requests information about subsidies and community impacts. The cited directive and RFI do not establish a comparable statewide per-MW community investment benchmark.[2](#aixe-source-2),[3](#aixe-source-3) | | Potential contributionAIxEnergy analysis | **A coordinated basis for responsible development:** connect infrastructure obligations, environmental performance, and host-community benefits before major commitments are locked in. | **More credible infrastructure planning:** improve the information used to assess project demand, resource requirements, and readiness to connect. | The New York advantage | AIxEnergy analysisNew York gives environmental review, grid planning, and community negotiation a common place in the development process. That coordination can help developers structure more durable projects and give host communities a stronger basis for securing lasting value. **Policy snapshot: September 20, 2026.** The two actions affect different approvals and are not blanket bans on every data center project. Potential benefits are assessments of policy design, not measured outcomes. **Sources** 1. Kathy Hochul, [Executive Order 62](https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data-centers-new-york-while-state-develops?ref=aixenergy.io), July 14, 2026. 2. Greg Abbott, [letter to Thomas Gleeson and Pablo Vegas concerning data centers](https://gov.texas.gov/uploads/files/press/Thomas%5FGleeson%5FPablo%5FVegas%5FData%5FCenters%5FDirective%5FLetter%5Fto%5FPUCT%5FERCOT%5FAugust%5F2026%5F.pdf?ref=aixenergy.io), August 3, 2026. 3. ERCOT, [“Data Center Impact RFI”](https://www.ercot.com/about/legal/data-center-impact-rfi?ref=aixenergy.io); and [“State and Community Impact RFI: Question Reference,”](https://www.ercot.com/files/docs/2026/09/18/State-and-Community-Impact-RFI%5FQuestion-Reference-9.18.26.pdf?ref=aixenergy.io) September 18, 2026. 4. Office of Governor Kathy Hochul, [“First Statewide Moratorium on New Hyperscale Data Centers Launched by Governor Kathy Hochul,”](https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul?ref=aixenergy.io) July 14, 2026. 5. Texas Legislature, [Senate Bill 6, enrolled text](https://capitol.texas.gov/tlodocs/89R/billtext/html/SB00006F.htm?ref=aixenergy.io), 89th Legislature, Regular Session, 2025. 6. Empire State Development, [“Community Investment Framework”](https://esd.ny.gov/communityinvestmentframework?ref=aixenergy.io) and [guidance document](https://esd.ny.gov/sites/default/files/media/document/CIFGuidanceDocument.pdf?ref=aixenergy.io), accessed September 20, 2026. There is a reason this distinction matters to customers who have no involvement in the AI industry. Consider a hypothetical utility asked to serve a campus that expects to grow from an initial phase to several hundred megawatts. The utility may need to commit to equipment or network reinforcement before the later phases are financed. If the campus stops expanding, the equipment does not disappear. The financial question is whether the developer’s commitments cover the resulting exposure or whether other customers inherit it. Texas already has a statutory foundation for addressing that problem. Senate Bill 6 directs the development of large-load interconnection standards covering financial commitments, site control, and disclosure of substantially similar service requests. It also addresses interconnection cost recovery and certain emergency operating obligations. The August audit therefore sits alongside a broader electricity policy effort; it would be inaccurate to portray Texas as concerned only with paperwork or unconcerned with ratepayers.8 ## New York's Broader Approach New York’s advantage is the breadth of the work it has joined to the permitting pause. The July announcement placed the environmental review alongside consideration of a Grid Acceleration Fund, new clean electricity supply, and protection against infrastructure costs associated with speculative demand. These are mechanisms being developed or considered, not benefits already demonstrated or uniform charges already imposed.9 When a project creates a new infrastructure requirement, its commercial obligations should reflect that requirement. If a customer wants capacity reserved years ahead of actual use, that reservation has consequences for the utility’s investment decisions and for other customers seeking service. Connecting financial commitments to the development schedule gives both sides a more credible basis for planning. The same principle applies to flexibility. A data center that can reliably reduce demand during constrained hours may impose a different system burden from one that requires uninterrupted grid service at its maximum contracted demand. But planners need to understand the operating commitment. How much load can be reduced, with how much notice, for how long, and under whose control? What happens if the facility cannot deliver? Flexibility becomes valuable to the power system when these questions have enforceable answers. For technology suppliers, this creates a productive direction for innovation. Cooling improvements, storage, control systems, and workload management should be evaluated against the constraints they actually relieve. A proposal becomes more persuasive when its energy and water claims are measurable under local operating conditions. The commercial opportunity is to make development easier to support through demonstrated performance. ## A Stronger Bargain New York’s Community Investment Framework makes the broader policy particularly tangible. Released in September, it gives local governments a voluntary basis for negotiating benefits with developers. Its recommended benchmark is $1 million for each megawatt of utility demand. The state illustrates the benchmark with a 50-megawatt project and a recommended $50 million community investment. The amount is a negotiating benchmark, not an automatically assessed statewide tax or a uniform fee due at permitting.10 That is a consequential change in how the economic development discussion can proceed. Large capital expenditures do not automatically produce a proportionate flow of local income or public revenue. Equipment spending, construction employment, permanent employment, and municipal receipts have different beneficiaries and different time horizons. A host community needs to understand each of them before concluding that a large investment announcement is sufficient compensation for its commitments of land and infrastructure. The framework gives that community a more informed starting position. Empire State Development identifies good-neighbor commitments, labor standards, transparency, and targeted investments as its core elements. Localities can adapt the approach to their own priorities, including public infrastructure, broadband, education, and child care.11 This is one of the most defensible strengths of the New York approach. A small municipality may negotiate only one project of this kind, while a sophisticated developer can draw on experience across many markets. Providing a common structure and practical guidance can reduce that imbalance without requiring every town to make the same bargain. The state supplies analytical support; the community retains responsibility for determining which benefits matter locally. The implementation details are as important as the headline benchmark. ESD’s guidance calls for clear responsibilities, payment schedules, attention to ongoing maintenance, and planning for expansion or abandonment. It advises communities to use counsel to translate negotiated terms into binding agreements. The guidance also describes the benchmark as inclusive of grid modernization investments required by New York State, a detail that matters when estimating a project’s obligations.12 For a developer, these distinctions belong in the financial model early. A negotiated community investment, a utility infrastructure contribution, and an operating electricity bill serve different purposes. They should be accounted for explicitly, with any overlap resolved in the applicable agreements. Treating them as an undifferentiated surcharge makes it harder to compare sites and easier to misunderstand what the community is receiving. For a town, the value of an agreement depends on whether it survives changes in ownership and development plans. An attractive promise has limited value if the party making it will not own the operating asset or if payment depends on a later phase that never arrives. My view is that a well-structured agreement should make those contingencies visible before construction begins. This is where careful public policy can improve the quality of a private transaction. New York also connects local benefits to a broader environmental assessment. Executive Order 62 directs a public process examining energy and water, air quality, noise, and disproportionate effects on disadvantaged communities. It places that work within the Department of Public Service’s large-load interconnection proceeding and provides for consultation across relevant agencies.13 That institutional design addresses a familiar problem in infrastructure development. A project can look acceptable when each decision is considered separately while creating a difficult cumulative result. Water availability can constrain a cooling design. A change in cooling can alter electricity demand. Onsite generation can change the air permitting question without resolving the need for backup grid service. Reviewing these relationships together gives policymakers and developers a better chance of identifying a workable design before major commitments become difficult to change. Onsite power illustrates why the distinction between New York and Texas matters. Texas’s audit is organized around projects seeking access to ERCOT. New York’s permitting restriction depends on the covered facility and its need for affected DEC approvals. A proposal to supply power onsite therefore does not, by itself, answer the New York permitting question.14 More broadly, self-supply is a design choice with its own fuel, operating, and permitting requirements. Its value depends on what it actually delivers, including during equipment outages or stressed system conditions. ## Other States Are Taking Different Paths Cleveland approved a three-month pause on new standalone, principal-use data center projects in July while evaluating impacts and seeking public input. That is a municipal land-use response, narrower in geography and different in authority from either state action.15 It illustrates why a national count of moratoriums tells an investor less than a clear account of what each restriction covers. Maine provides an example of the importance of distinguishing proposals from enacted policy. Governor Janet Mills vetoed its proposed statewide pause in April, objecting to the absence of an exemption for a project in Jay. The vetoed proposal should not be listed as an operative statewide moratorium.16 Massachusetts has taken another path. Its September executive action makes local approval and community benefits part of the conditions for covered projects to advance through state permitting. This is a conditional development framework rather than a general moratorium.17 It reinforces the growing importance of host-community consent without making the Massachusetts and New York approaches interchangeable. These examples should be read as different institutional responses to a shared planning problem. Governments need enough information and authority to connect private development to public infrastructure and local priorities. A temporary permit hold, a grid audit, and a community agreement requirement each address part of that problem. Their effectiveness depends on what follows from them. ## A Foundation for Growth? For New York, the strongest opportunity is to turn this coordinated effort into a more predictable basis for responsible development. A credible applicant should be able to identify the relevant standards, understand its financial responsibilities, and work with the host community on commitments that can be delivered. A utility should be able to distinguish between demand supported by binding obligations and demand that remains speculative. Residents should be able to see how a project affects their community and what they will receive in return. This is an assessment of policy design, not a claim that the eventual results have already been measured. Lower electricity bills, more reliable project delivery, and lasting local investments must be demonstrated through implementation. The advantage visible today is that New York has given those objectives a common place in the development process, with the community framework already providing a practical tool for negotiations. Texas’s audit can improve the quality of information entering infrastructure decisions. New York’s broader approach can improve the terms on which those decisions are made. The distinction matters for the next phase of AI infrastructure investment. A successful project must be capable of securing power, carrying its infrastructure obligations, and maintaining the confidence of the community around it. New York is making those responsibilities explicit at a point when developers can still incorporate them into the project. That is a strong foundation for growth that can endure. ## **Notes** 1\. Kathy Hochul, [Executive Order No. 62, “Establishing a Temporary Moratorium on Data Centers in New York While the State Develops Higher Standards for Data Center Development and Benefits Blueprint to Support Localities,”](https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data-centers-new-york-while-state-develops?ref=aixenergy.io) July 14, 2026; Greg Abbott, [letter to Thomas Gleeson and Pablo Vegas,](https://gov.texas.gov/uploads/files/press/Thomas%5FGleeson%5FPablo%5FVegas%5FData%5FCenters%5FDirective%5FLetter%5Fto%5FPUCT%5FERCOT%5FAugust%5F2026%5F.pdf?ref=aixenergy.io) August 3, 2026, 1–2. 2\. Hochul, [Executive Order No. 62,](https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data-centers-new-york-while-state-develops?ref=aixenergy.io) §§ 1 and 6. 3\. Office of Governor Kathy Hochul, [“First Statewide Moratorium on New Hyperscale Data Centers Launched by Governor Kathy Hochul,”](https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul?ref=aixenergy.io) July 14, 2026; Hochul, Executive Order No. 62, §§ 1 and 5; Scott D. Ellis, [“New York Just Pressed Pause on Large Data Center Permitting,”](https://www.foley.com/insights/publications/2026/08/new-york-just-pressed-pause-on-large-data-center-permitting/?ref=aixenergy.io) Foley & Lardner, August 3, 2026\. The public announcement describes an up-to-one-year process; the order’s operative permitting provision uses completion of the environmental review as its endpoint. 4\. Abbott, [letter to Gleeson and Vegas,](https://gov.texas.gov/uploads/files/press/Thomas%5FGleeson%5FPablo%5FVegas%5FData%5FCenters%5FDirective%5FLetter%5Fto%5FPUCT%5FERCOT%5FAugust%5F2026%5F.pdf?ref=aixenergy.io) August 3, 2026, 1–2. 5\. Electric Reliability Council of Texas, [“ERCOT Request for Information Regarding Data Center State and Community Impacts,”](https://www.ercot.com/about/legal/data-center-impact-rfi?ref=aixenergy.io) accessed September 20, 2026; ERCOT, [“Data Center State and Community Impact RFI—Question Reference Guide,”](https://www.ercot.com/files/docs/2026/09/18/State-and-Community-Impact-RFI%5FQuestion-Reference-9.18.26.pdf?ref=aixenergy.io) version 1.0, September 18, 2026, 1–2\. The guide identifies the RIOO response portal as the controlling document. 6\. Abbott, letter to Gleeson and Vegas, August 3, 2026, 2; Greg Abbott, [letter to L’Oreal Stepney and Bryan McMath,](https://gov.texas.gov/uploads/files/press/TWDB%5FLetter.pdf?ref=aixenergy.io) September 14, 2026, 1–2. 7\. Office of the Texas Governor, [“Governor Abbott Directs Comprehensive Data Center Audit,”](https://gov.texas.gov/news/post/governor-abbott-directs-comprehensive-data-center-audit?ref=aixenergy.io) August 3, 2026\. Queue figures are the governor’s reported figures at that date and are not a forecast of realized load. 8\. Texas Legislature, [Senate Bill 6,](https://capitol.texas.gov/tlodocs/89R/billtext/html/SB00006F.htm?ref=aixenergy.io) 89th Legislature, Regular Session (2025), enrolled text, §§ 1–2, adding Texas Utilities Code §§ 35.004(c-1)–(c-2) and 37.0561\. The statutory starting threshold for the interconnection standards is 75 MW, subject to a lower threshold set by the commission; this differs from the 25 MW scope of ERCOT’s September 2026 data center RFI. 9\. Office of Governor Kathy Hochul, [“First Statewide Moratorium”;](https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul?ref=aixenergy.io) Hochul, Executive Order No. 62, §§ 3–4\. The order directs consideration of the fund and related mechanisms; it does not itself establish a final contribution schedule. 10\. Office of Governor Kathy Hochul, [“Governor Hochul Announces Strongest Community Investment Framework in the Nation to Protect Communities from Exploitative Data Center Development,”](https://www.governor.ny.gov/news/governor-hochul-announces-strongest-community-investment-framework-nation-protect-communities?ref=aixenergy.io) September 15, 2026. 11\. Empire State Development, [“Community Investment Framework,”](https://esd.ny.gov/communityinvestmentframework?ref=aixenergy.io) accessed September 20, 2026. 12\. Empire State Development, [New York State Guidance for Host Communities Implementing the Community Investment Framework for Data Center Development,](https://esd.ny.gov/sites/default/files/media/document/CIFGuidanceDocument.pdf?ref=aixenergy.io) accessed September 20, 2026, 1–6, especially 5–6 on obligations and fund administration and 6 on the treatment of grid modernization investment. 13\. Hochul, [Executive Order No. 62,](https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data-centers-new-york-while-state-develops?ref=aixenergy.io) §§ 1 and 7. 14\. Hochul, [Executive Order No. 62,](https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data-centers-new-york-while-state-develops?ref=aixenergy.io) §§ 1 and 6; Abbott, letter to Gleeson and Vegas, August 3, 2026, 1–2\. Comparison of the instruments’ respective scopes is the author’s analysis. 15\. City of Cleveland, [“Data Center Public Survey,”](https://www.clevelandohio.gov/city-hall/departments/economic-development/data-centers?ref=aixenergy.io) accessed September 20, 2026. 16\. Aditya Soni and Mrinmay Dey, [“Maine Governor Blocks First US State Freeze on New Data Centers,”](https://www.reuters.com/legal/litigation/maine-governor-rejects-first-us-state-freeze-new-data-centers-2026-04-24/?ref=aixenergy.io) Reuters, April 24, 2026, updated April 25, 2026. 17\. David J. Murphy, [“Massachusetts Just Gave Towns a Veto Over Data Centers. Here’s What Developers Need to Know,”](https://murphypc.com/news/massachusetts-just-gave-towns-a-veto-over-data-centers-heres-what-developers-need-to-know/?ref=aixenergy.io) Murphy PC, accessed September 20, 2026\. This discussion of the September 8 executive action relies on the published legal analysis; Massachusetts’s official order page was inaccessible during source review. ### Google’s 100-Megawatt Capacity Deal Signals a New Phase in AI Power Procurement URL: https://www.aixenergy.io/googles-100-megawatt-capacity-deal-signals-a-new-phase-in-ai-power-procurement/ Last updated: 2026-09-08T15:57:37.000Z The strategic change is not that a virtual power plant can provide demand response. It is that an AI company can now procure accredited capacity through a distributed portfolio, turning flexibility from a utility program into a corporate infrastructure contract. _This post is for subscribers only._ ### The Next Bottleneck Has a Serial Number: Washington Has Added Equipment Provenance to the AI Power Race URL: https://www.aixenergy.io/the-next-bottleneck-has-a-serial-number-washington-has-added-equipment-provenance-to-the-ai-power-race/ Last updated: 2026-09-08T02:43:06.000Z The race to build artificial-intelligence infrastructure has been organized around a deceptively simple question: where are the available megawatts? Developers have searched for generation, transmission headroom, substations, natural gas, water, land and permitting pathways. Executive Order 14420 adds another question that may prove just as consequential: can the equipment required to deliver those megawatts be legally procured, securely operated and reliably serviced? President Donald Trump signed the order on August 26, declaring a national emergency over foreign involvement in equipment used by the U.S. bulk-power system. It authorizes the Department of Energy to prohibit the acquisition, importation, transfer or installation of foreign-produced equipment associated with entities determined to present unacceptable national-security, cybersecurity or supply-disruption risks. The order covers not only transformers, generators and circuit breakers, but also grid-connected inverters, battery energy storage systems, protective relays, backup generators, industrial controls, firmware, software, maintenance services and remote-access capabilities.[1](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fn-1) The security rationale is credible. The implementation risk is equally real. If DOE develops a technically grounded, risk-tiered system supported by domestic manufacturing and trusted allied supply chains, the order could close material vulnerabilities. If it relies primarily on country of origin, it could deepen equipment shortages, increase market concentration and delay the grid expansion needed to support the AI buildout. ## What the Order Actually Changes Executive Order 14420 is not an immediate prohibition on all imported grid equipment. A covered transaction must involve foreign-produced equipment or an associated component, service or digital capability connected to a “Covered Foreign Entity,” and DOE must determine that the transaction creates an undue or unacceptable risk. China is among the jurisdictions captured through the order’s incorporation of U.S. arms-embargo and sanctions policies, although DOE may designate additional governments, companies or persons.[2](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fn-2) The Secretary of Energy can license otherwise prohibited transactions, negotiate mitigation measures and establish a list of prequalified equipment and vendors. DOE must publish implementing rules within 120 days, placing the initial deadline around December 24, 2026\. Until those rules define the affected countries, companies, products and transaction classes, the legal perimeter will remain uncertain. The commercial effects begin before the rules are complete. Transactions initiated after August 26 may eventually be subject to prohibition, while equipment installed before the order can be identified, isolated, monitored, secured, disconnected, replaced or removed. DOE must consider reliability, safety, replacement availability and continuity of service before ordering removal, but utilities and developers cannot assume that installed equipment is permanently grandfathered. This is a more expansive successor to Executive Order 13920, issued during the first Trump administration in May 2020.[3](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fn-3) The implementing action under that order was ultimately limited to selected Chinese equipment serving critical defense facilities and was revoked in April 2021.[4](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fn-4) The new order covers a wider set of technologies, explicitly reaches software and lifecycle services, and gives DOE clearer authority to address equipment already in operation. ## Two Risks Hidden Inside One Policy The order combines two distinct national-security problems. The first is control risk: the possibility that a foreign government or affiliated vendor could exploit hardware, firmware, software or remote access to manipulate equipment. The second is continuity risk: the possibility that geopolitical conflict, sanctions, export controls or industrial concentration could interrupt the supply of equipment, replacement components or maintenance services. Those risks require different responses. Control risk is reduced through product testing, signed firmware, access restrictions, network segmentation, credential management, software bills of materials, independent monitoring and the elimination of unnecessary vendor connections. Continuity risk is reduced through diversified sourcing, interchangeable designs, domestic and allied manufacturing, strategic inventories, repair capacity and long-term procurement commitments. A blanket equipment restriction can address one risk while worsening the other. Removing a vendor with dangerous remote-control capabilities may reduce control risk. Removing multiple suppliers before substitutes are available may increase continuity risk by reducing spare-equipment availability and lengthening restoration and construction schedules. The distinction matters because the order invokes AI and data-center growth as evidence that the threat has become more acute. Load growth increases the economic consequences of a major grid disruption, but it does not necessarily increase the probability that any individual transformer or control system has been compromised. AI infrastructure changes the consequence side of the risk equation more directly than the vulnerability side. That distinction should influence implementation. Equipment serving a nationally significant transmission corridor, nuclear facility, military installation or multi-gigawatt data-center cluster presents a different consequence profile from similar equipment at a smaller and more isolated facility. A serious regime would account for system consequence, digital controllability and substitutability rather than treating every product within a broad category as equivalent. ## Why Origin Alone Is Not Enough The order uses foreign production and foreign control as thresholds for federal action, but neither is a complete measure of security. A product assembled in the United States may still contain foreign control boards, communications modules, code libraries, firmware, cloud services or administrative dependencies. Domestic assembly does not establish who controls the product after installation. The reverse is also true. Equipment manufactured by a trusted allied company under transparent security and servicing practices may present less operational risk than domestically assembled equipment with opaque software and persistent vendor access. The relevant question is not simply where the enclosure was built, but who can alter the device, reach it remotely, withhold essential services or prevent its owner from operating it independently. The order partly recognizes this problem by including software, firmware, maintenance services and remote access. Its domestic-manufacturing provisions nevertheless risk turning a complex engineering assessment into a politically convenient country-of-origin test. That would be easier to administer, but it would not necessarily identify the most consequential vulnerabilities. FERC and NERC already operate mandatory supply-chain risk-management standards for covered bulk-electric-system cyber assets.[5](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fn-5) Those standards do not capture every device or large-load facility implicated by the present buildout, but they provide an institutional foundation. DOE should build on that framework rather than creating a separate vendor regime with different definitions, evidence standards and mitigation requirements. ## The Collision with Equipment Scarcity The order arrives during the most difficult grid-equipment market in decades. DOE reported in August that lead times for critical equipment can exceed two years and that prices for some transformers have increased four- to ninefold over the past five years. The department has announced a program of up to $375 million to expand domestic production, encourage equipment standardization and reduce reliance on imports, but manufacturing capacity cannot be created on a 120-day regulatory schedule.[6](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fn-6) The Government Accountability Office has identified long manufacturing lead times, limited capacity, labor and material shortages, transportation constraints and excessive customization as major threats to transformer resilience. As of May 2026, GAO’s recommendation that DOE establish an actionable transformer supply-chain plan remained open.[7](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fn-7) The federal government is therefore imposing a new security screen before it has resolved the underlying availability problem. This creates a stock-and-flow conflict. Restrictions on new transactions affect the flow of equipment into the system, while potential replacement orders create additional demand from the installed stock. If existing equipment must compete with new substations and power plants for the same limited pool of approved replacements, remediation could displace expansion rather than accompany it. The federal AI strategy points in the opposite direction. A July 2025 executive order sought to accelerate large data-center projects and specifically identified transmission lines, substations, transformers, switchgear, gas turbines and backup power systems as components upon which those projects depend.[8](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fn-8) Federal policy is simultaneously trying to accelerate construction and tighten access to the equipment required for construction. Those objectives are not inherently incompatible, but they must be modeled together. Restricting insecure equipment while expanding secure production could improve long-term resilience. Restricting equipment faster than secure supply can respond will produce higher costs, longer energization schedules and greater competition for a smaller set of vendors. ## Data Centers Cannot Escape Through the Meter The order formally applies to the bulk-power system, defined to include transmission facilities operating at 69 kilovolts or above while excluding local distribution. Many hyperscale campuses connect at transmission voltage or require new substations, switchyards and protective equipment within that perimeter. Even where the data center itself is not directly regulated, the utility facilities required to serve it may be. Developers may therefore bear costs associated with equipment they neither select nor own. Utilities generally procure the transformers, breakers, relays and controls used for interconnection, but network-upgrade costs and schedule consequences can be assigned to the connecting customer. A utility decision to disqualify a vendor, repeat a solicitation or redesign a substation can delay a data center without changing the project’s nominal interconnection position. Behind-the-meter generation does not remove the exposure. A campus powered by gas turbines, fuel cells, batteries or a microgrid still requires transformers, switchgear, inverters, controls, backup systems and replacement components. The order expressly includes several of those categories, making equipment provenance relevant to the very configurations being promoted as alternatives to delayed utility service. The market response will be uneven. Hyperscalers can reserve manufacturing capacity, conduct detailed supplier audits, negotiate cybersecurity requirements and finance equipment years before energization. Independent developers, merchant generators, storage companies and smaller utilities have less purchasing leverage and less ability to absorb redesign or delay. A prequalified-vendor list could therefore become a new concentration point. Regulatory approval would increase demand for a limited number of manufacturers, and early production slots from those companies could acquire substantial option value. Large buyers would reserve compliant capacity earlier, leaving smaller market participants to accept longer schedules or greater regulatory uncertainty. ## From Available Megawatts to Secure Megawatts The order should be treated as a constraint on deliverable capacity rather than as a generic cybersecurity flag. The relevant quantity is the secure megawatt: the amount of load that can be physically delivered by a given date using equipment that is available, qualified and serviceable under the applicable security regime. At any point in time, secure capacity is bounded by the lowest of five quantities: network-ready capacity, firm supply capacity, site-ready load, compliant equipment capacity and regulatory clearance. More generation or transmission headroom does not increase the secure megawatt if qualified transformers or control systems remain the binding constraint. The same logic applies to timing. A project’s secure energization date is the latest of its network-completion date, power-supply date, site-readiness date, compliant-equipment delivery date and regulatory-clearance date. This makes the equipment issue analytically comparable to permitting, interconnection and construction rather than leaving it in a separate cybersecurity discussion. That treatment reveals how a national rule can produce geographically uneven effects. Utilities use different equipment specifications, approved-vendor lists, procurement practices and substation designs. Corridors also differ in their dependence on new transmission, imported inverters, battery systems, gas turbines and behind-the-meter configurations. A national restriction will therefore propagate through regional project pipelines in different ways. ## The Cost-Causation Problem The order does not determine who will pay for compliance. That omission will become important as utilities identify equipment requiring additional monitoring, isolation, replacement or procurement changes. Costs could be assigned to individual connecting customers, included in network-upgrade charges, recovered through utility rates or absorbed by equipment vendors, depending on ownership, tariff structure and the timing of the federal determination. Prospective projects will be easier to address because utilities can incorporate new requirements into interconnection agreements and procurement specifications. Existing equipment will be harder because replacement may benefit the broader system while being triggered by a national-security determination unrelated to the original prudence of the investment. Assigning those costs entirely to one customer would often be difficult to justify. Data-center tariffs and large-load interconnection rules will eventually need to address this issue. If a customer specifies or owns the affected behind-the-meter equipment, direct cost assignment may be appropriate. If the utility owns a shared substation or replaces equipment serving multiple customers, broader allocation may better reflect cost causation. Ignoring the allocation question will not make it disappear. It will move the issue into rate cases, interconnection disputes, contract negotiations and project financing, where uncertainty itself becomes a cost. ## A Better Implementation Path DOE should classify equipment according to system consequence, digital controllability, ownership and remote access, supply-chain concentration, replacement availability and the effectiveness of available mitigation. High-consequence equipment with persistent access controlled by an adversarial entity should face the strongest restrictions. Passive components, isolated equipment and products with independently verifiable controls should not automatically receive the same treatment. The government should also distinguish trusted allied sourcing from strategic dependence on adversarial jurisdictions. Hardware and software bills of materials, signed firmware, controlled administrative credentials, independent logging, vulnerability disclosure, patching obligations and termination rights for remote services would provide stronger evidence of security than domestic assembly alone. Utilities should retain the practical ability to operate essential equipment if a vendor or country becomes unavailable. Replacement requirements must be sequenced around actual manufacturing and restoration needs. Equipment presenting an active control threat may require rapid isolation or replacement, while equipment presenting primarily a future continuity risk may be addressed through monitoring, spares and phased retirement. Treating both categories as emergencies would consume scarce replacements without necessarily reducing the highest risks first. The 69-kV boundary also requires attention. Coordinated batteries, inverters, generators and large loads connected below transmission voltage can influence bulk-system behavior even if no individual device meets the formal threshold. DOE, FERC, NERC, CISA, state regulators and utilities will need a common approach to aggregated distribution-side risk rather than a jurisdictional gap. ## Conclusion Executive Order 14420 recognizes that energy security depends on more than fuel supply and generation capacity. The electric system is also a network of manufactured devices, embedded software, service relationships and geopolitical dependencies. Ignoring those dependencies would leave a material weakness in the infrastructure supporting the AI economy. The order can still become misguided if domestic origin substitutes for technical evidence or if restrictions advance faster than secure supply. A resilient policy must reduce adversarial control while preserving enough equipment availability to expand, maintain and restore the grid. Those are joint constraints, not competing talking points. The AI power race is therefore entering a new phase. The next gigawatt will not go simply where land is cheap, projects are announced or electrical headroom appears available. It will go where the full corridor can procure, qualify, install and service the equipment required to turn theoretical capacity into secure megawatts. ## Notes 1. Donald J. Trump, “Declaring a National Emergency to Secure the United States Bulk-Power System,” Executive Order 14420, August 26, 2026, White House, [https://www.whitehouse.gov/presidential-actions/2026/08/declaring-a-national-emergency-to-secure-the-united-states-bulk-power-system/](https://www.whitehouse.gov/presidential-actions/2026/08/declaring-a-national-emergency-to-secure-the-united-states-bulk-power-system/?ref=aixenergy.io). [↩](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fnref-1) 2. U.S. Department of State, Directorate of Defense Trade Controls, “Country Policies,” accessed August 27, 2026, [https://www.pmddtc.state.gov/ddtc\_public?id=ddtc\_public\_portal\_country\_landing](https://www.pmddtc.state.gov/ddtc%5Fpublic?id=ddtc%5Fpublic%5Fportal%5Fcountry%5Flanding&ref=aixenergy.io). [↩](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fnref-2) 3. Donald J. Trump, “Securing the United States Bulk-Power System,” Executive Order 13920, May 1, 2020, *Federal Register* 85, no. 86 (May 4, 2020): 26595–99, [https://www.federalregister.gov/documents/2020/05/04/2020-09695/securing-the-united-states-bulk-power-system](https://www.federalregister.gov/documents/2020/05/04/2020-09695/securing-the-united-states-bulk-power-system?ref=aixenergy.io). [↩](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fnref-3) 4. U.S. Department of Energy, “Revocation of Prohibition Order Securing Critical Defense Facilities,” *Federal Register* 86, no. 76 (April 22, 2021): 21308–9, [https://www.federalregister.gov/documents/2021/04/22/2021-08483/revocation-of-prohibition-order-securing-critical-defense-facilities](https://www.federalregister.gov/documents/2021/04/22/2021-08483/revocation-of-prohibition-order-securing-critical-defense-facilities?ref=aixenergy.io). [↩](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fnref-4) 5. Federal Energy Regulatory Commission, *Supply Chain Risk Management Reliability Standards*, Order No. 850, 165 FERC ¶ 61,020, Docket No. RM17-13-000, October 18, 2018, [https://www.ferc.gov/media/order-no-850](https://www.ferc.gov/media/order-no-850?ref=aixenergy.io). [↩](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fnref-5) 6. U.S. Department of Energy, Office of Electricity, “Strengthening America’s Grid Supply Chain,” August 10, 2026, [https://www.energy.gov/oe/articles/strengthening-americas-grid-supply-chain](https://www.energy.gov/oe/articles/strengthening-americas-grid-supply-chain?ref=aixenergy.io). [↩](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fnref-6) 7. U.S. Government Accountability Office, *Electricity Grid: DOE Could Better Support Industry Efforts to Ensure Adequate Transformer Reserves*, GAO-23-106180 (Washington, DC: Government Accountability Office, August 2023), updated May 2026, [https://www.gao.gov/products/gao-23-106180](https://www.gao.gov/products/gao-23-106180?ref=aixenergy.io). [↩](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fnref-7) 8. Donald J. Trump, “Accelerating Federal Permitting of Data Center Infrastructure,” Executive Order 14318, July 23, 2025, White House, [https://www.whitehouse.gov/presidential-actions/2025/07/accelerating-federal-permitting-of-data-center-infrastructure/](https://www.whitehouse.gov/presidential-actions/2025/07/accelerating-federal-permitting-of-data-center-infrastructure/?ref=aixenergy.io). [↩](https://chatgpt.com/c/6a904df5-77b8-83e8-b086-31e1385b3a62?ref=aixenergy.io#user-content-fnref-8) ### The Geography of AI: Infrastructure Corridors and the New Data Center Map URL: https://www.aixenergy.io/the-next-geography-of-ai-infrastructure-corridors-and-the-new-data-center-map/ Last updated: 2026-09-08T02:43:32.000Z The next gigawatt will go not where land is cheapest or ambition loudest, but where the full corridor—grid, land, equipment, capital, and institutions—can execute. _This post is for paying subscribers only._ ### The Grid’s Newest Contingency Is Demand That Disappears URL: https://www.aixenergy.io/the-grids-newest-contingency-is-demand-that-disappears/ Last updated: 2026-09-08T15:58:49.000Z AIxEnergy has identified 14 discrete large-load disturbances and three documented event series across PJM, ERCOT, and Ireland. Together, they reveal the reliability problem that is created when individually rational protection systems don't coordinate. _This post is for subscribers only._ ### A.I. Data Centers Planned Across the Parched U.S. West URL: https://www.aixenergy.io/thirsty-a-i-data-centers-planned-throughout-drought-stricken-u-s-west/ Last updated: 2026-08-27T15:23:00.000Z AIxEnergy identifies 31 pipeline data-center and large-load projects in constrained Western water-utility territories. This indicates that diligence should move to the front of site selection. _This post is for paying subscribers only._ ### The AI Boom Is Becoming a Credit Bet URL: https://www.aixenergy.io/who-is-financing-the-ai-buildout-and-where-is-the-risk-going/ Last updated: 2026-08-27T15:22:51.000Z American venture capital did not merely set a six-month record in the first half of 2026; it exceeded every previous full-year investment total. US startups raised $412.7 billion, of which $355.9 billion, or 86.2 percent, went to companies classified by PitchBook as artificial intelligence and machine learning businesses. In 2025, the National Venture Capital Association separately reported $320 billion of total US venture investment, $222 billion of AI deal value, and a 65.4 percent AI share. The rounded 2025 measures do not reconcile exactly, but the direction is unambiguous: AI’s share rose from 50.9 percent in 2024 to 65.4 percent in 2025 and more than 86 percent in the first half of 2026\. Financings of $100 million or more captured 87.5 percent of the first-half total, compared with 67 percent in 2025.1 Article Summary 0:00 /34.272653 1× OpenAI closed $122 billion of committed capital in March 2026 at an $852 billion post-money valuation, a round co-led by SoftBank and Andreessen Horowitz with Amazon and Nvidia as the largest strategic investors.2 Anthropic followed in May 2026 with a $65 billion Series H at a $965 billion post-money valuation.3 The two rounds totaled $187 billion on a reported basis. Anthropic, however, disclosed that its $65 billion Series H included $15 billion of previously committed hyperscaler investments, including $5 billion from Amazon. Subtracting that known overlap produces an adjusted total of $172 billion. On that basis, OpenAI and Anthropic alone represented 48.3 percent of all US AI venture investment in the first half of 2026 and 41.7 percent of the entire US venture market. The record was therefore not merely AI-concentrated; it was concentrated in two frontier-model companies. The largest checks into OpenAI's March round came from Amazon, Nvidia, and SoftBank, companies that supply the cloud capacity, chips, and platform access these labs need to operate and may receive commercial benefits alongside financial return. What has changed is that the two pools increasingly compete for the same rounds and get counted together. Sovereign investors have moved into a similar lane. MGX, an Abu Dhabi technology investment company established with Mubadala and G42 as foundational partners, closed a $49 billion AI fund on July 1, 2026.4 These are strategic actors pursuing compute access and geopolitical positioning (as well as return). Where the checks are too large even for this pool of strategic capital, financing has also moved to public debt and equity markets. Meta completed a $25 billion investment-grade bond sale on April 30, 2026, structured across six tranches with maturities extending as long as 40 years. The sale followed an increase in Meta’s capital-spending outlook, which the company subsequently narrowed in July to between $130 billion and $145 billion for 2026, up from its earlier lower bound of $125 billion.5 Oracle's fiscal 2026 capital expenditures reached $55.7 billion, while free cash flow was negative $23.7 billion; the company said it expected to raise approximately $40 billion of debt and equity in fiscal 2027\. Separately, Oracle disclosed approximately $260 billion of additional lease commitments as of May 31, 2026, substantially all related to data-center arrangements expected to commence over the following several fiscal years and generally run for 15 to 19 years. These are contractual capacity commitments rather than debt or equity already raised, and they remained outside the balance-sheet lease liabilities until commencement. Oracle separately reported $129.5 billion of outstanding indebtedness with maturities extending from 2026 through 2066.6 Meta’s Hyperion campus illustrates how the buildout is moving into project-finance structures. Funds managed by Blue Owl own 80 percent of the joint venture and Meta owns 20 percent. The parties committed to finance approximately $27 billion of total development costs in proportion to their ownership interests, although Blue Owl’s disclosed initial cash contribution was approximately $7 billion. Meta remains the construction manager, property manager and sole tenant. Its leases begin with four-year terms and extension options, while a capped residual-value guarantee protects the venture for the first 16 operating years. A portion of Blue Owl’s financing was funded through private securities issued to PIMCO and other bond investors.7 The retirement-account channel also requires care. An August 2025 executive order directed federal agencies to expand the pathways through which defined-contribution plans could consider private-market investments; it did not itself place private credit in workers' accounts or displace plan sponsors' fiduciary duties.8 Retirement exposure is therefore a possible transmission channel, not evidence that ordinary 401(k) savers already hold material positions in these projects. The most unusual layer is the financing that flows among AI companies, suppliers, cloud providers, and lenders. In June 2026, Apollo and Blackstone launched the Broadcom AI XPV Platform with an initial $35 billion capital solution supporting more than one gigawatt of Anthropic compute capacity. The financing is committed across a multiyear draw schedule, with deployment beginning at Fluidstack-operated sites in mid-2026\. The broader platform is designed to support more than 20 gigawatts of XPU-based computing capacity through 2028.9 This is a private-credit and vendor-financing structure. Researchers have documented how vendor equity stakes let hyperscalers report unrealized gains as earnings. Amazon's equity stake in Anthropic added a $16.8 billion pre-tax unrealized gain to its first-quarter 2026 earnings,10 and Alphabet reported a $36.9 billion gain on equity securities, a broader line that includes its Anthropic stake alongside other private holdings such as Waymo.11 Under GAAP rules for equity securities, these mark-to-market gains are recorded in earnings and disclosed separately from operating results. ## Selected AI Capital Transactions and Financial Exposures Major equity financings, debt issuances, infrastructure commitments, lease exposures, acquisitions and capital-spending measures associated with the AI buildout. __Selected AI capital transactions, commitments and financial exposures__ | Date | Entity / Project | Transaction | Financing channel | Status | Headline amount | Eligible capital\* | Key terms or scale | Primary risk bearer | Principal risk or disclosure issue | | ------------- | ----------------------------- | ----------------------------------------- | ---------------------------------- | ------------------ | --------------------- | ------------------ | ------------------------------------------------------------------------------- | -------------------------------------------------- | -------------------------------------------------------------------- | | Mar. 3, 2025 | Anthropic | Series E | Venture equity | Closed | $3.5B raised | $3.5B | $61.5B post-money valuation | Equity investors | Valuation, cash burn and compute commitments | | Mar. 19, 2025 | AI Infrastructure Partnership | Infrastructure investment platform | Fund equity and prospective debt | Announced | $30B equity target | — | Up to $100B of potential investment including debt | Fund investors and lenders | Deployment pace, leverage and asset concentration | | Mar. 31, 2025 | OpenAI | 2025 funding round | Venture equity | Closed | $40B raised | $40B | $300B post-money valuation | Equity investors | Compute costs, valuation and governance | | Aug. 31, 2025 | Oracle | Uncommenced data-center leases | Long-term lease commitments | Outstanding | $99.8B committed | — | Additional $6.6B subsequently committed; 16-year terms | Oracle and data-center lessors | Start dates, utilization, minimum payments and counterparty exposure | | Sept. 2, 2025 | Anthropic | Series F | Venture equity | Closed | $13B raised | $13B | $183B post-money valuation | Equity investors | Valuation and compute requirements | | Oct. 21, 2025 | Meta / Hyperion | Hyperion data-center joint venture | Project finance and joint venture | Closed | $27B development cost | $27B | $7B Blue Owl contribution; Meta residual-value guarantee for 16 operating years | Investors, lenders and Meta | Construction overruns, lease renewal and residual-value exposure | | Oct. 30, 2025 | Meta | Six-part bond issuance | Corporate debt | Completed | $30B issued | $30B | Maturities from five to 40 years | Bondholders and Meta shareholders | Debt service, refinancing and returns on AI capital spending | | Feb. 1, 2026 | Oracle | Calendar 2026 financing plan | Debt and equity issuance | Superseded plan | $45B–$50B planned | — | $47.5B midpoint | Prospective bondholders and shareholders | Execution, dilution, leverage and customer concentration | | Feb. 12, 2026 | Anthropic | Series G | Venture equity | Closed | $30B raised | $30B | $380B post-money valuation | Equity investors | Valuation and infrastructure commitments | | Feb. 28, 2026 | Oracle | Uncommenced data-center leases | Long-term lease commitments | Outstanding | $261B committed | — | Expected to commence through FY2028; 15–19-year terms | Oracle and data-center lessors | Utilization, commencement, guarantees and cancellation rights | | Mar. 31, 2026 | OpenAI | 2026 funding round | Strategic and institutional equity | Closed | $122B committed | $122B | $852B post-money valuation; includes about $3B through a bank channel | Equity investors | Commercial dependencies, governance and valuation | | Mar. 31, 2026 | OpenAI | Revolving credit facility | Bank credit | Available, undrawn | $4.7B capacity | — | Liquidity backstop rather than capital raised | Banks and OpenAI | Covenants, draw conditions and future leverage | | Apr. 30, 2026 | Meta | Six-part bond issuance | Corporate debt | Completed | $25B issued | $25B | Maturities extending to 40 years | Bondholders and Meta shareholders | Credit metrics, debt service and AI investment returns | | May 13, 2026 | Isomorphic Labs | Series B | Venture equity | Closed | $2.1B raised | $2.1B | AI-enabled drug discovery investment | Equity investors | Scientific execution and commercialization | | May 28, 2026 | Anthropic | Series H | Venture equity | Closed | $65B headline | $50B | $965B post-money valuation; $15B was previously committed hyperscaler capital | Equity investors | Double counting, valuation and strategic dependence | | June 9, 2026 | Broadcom / Anthropic Compute | Initial AI infrastructure financing | Private credit and vendor finance | Committed | $35B committed | $35B | At least 1 GW initially; platform designed for more than 20 GW through 2028 | Private-credit investors, banks and counterparties | Collateral value, chip obsolescence, leases and credit support | | June 10, 2026 | Oracle | FY2026 financing completed | Corporate debt and equity | Completed | $48B raised | $48B | $43B of debt and $5B of equity | Bondholders and shareholders | Leverage, dilution and infrastructure returns | | June 10, 2026 | Oracle | FY2026 capital expenditures | Internal capital expenditure | Actual | $55.7B spent | — | Free cash flow was negative $23.7B | Shareholders and creditors | Utilization, cash conversion and useful asset life | | June 10, 2026 | Oracle | FY2027 financing expectation | Prospective debt and equity | Planned | About $40B | — | Includes a previously announced $20B at-the-market equity program | Future bondholders and shareholders | Market absorption, execution, dilution and leverage | | July 1, 2026 | MGX Fund I | Final fund close | Investment fund commitments | Closed | $49B committed | $49B | Commitments are not equivalent to deployed capital | Limited partners and sovereign sponsor | Deployment concentration and investment returns | | July 7, 2026 | Amazon | U.S.-dollar bond offering | Corporate debt | Marketed | $25B target | — | Final completed amount not yet reflected in the source record | Bondholders and Amazon shareholders | Market absorption, debt service and capital-spending returns | | July 21, 2026 | Aligned Data Centers | Acquisition by AIP, MGX and BlackRock GIP | Infrastructure acquisition | Closed | $40B enterprise value | — | Acquisition value, not new infrastructure capital | Acquirer equity and acquisition lenders | Purchase price, leverage and integration | | July 28, 2026 | Meta / El Paso Data Center | Strategic infrastructure venture | Project finance and joint venture | Pending close | More than $10B | — | At least 1 GW; initial capacity expected in 2028 | Venture investors and Meta | Construction, tenant concentration and undisclosed financing terms | | July 29, 2026 | Meta | FY2026 capital-expenditure guidance | Internal capital expenditure | Guidance | $130B–$145B | — | $137.5B midpoint; $31.1B spent in the second quarter | Shareholders and creditors | Demand realization, margins and infrastructure utilization | **\*Eligible capital** is the amount included in the strict additive financing aggregate after excluding targets, undrawn credit facilities, capital expenditures, lease-exposure stocks, acquisition values, pending transactions and identified overlap. The transactions shown contain approximately **$474.6 billion** of aggregation-eligible capital. Still, venture capital may be having a harder time than the record headline totals suggest. Fintech funding rose 23 percent by dollar volume in the first half of 2026, but deal count fell 25.7 percent year over year, to its lowest level in several years, as capital concentrated into fewer, larger checks for category leaders. In North America, seed and angel investment in the second quarter fell 27 percent year over year, although small rounds often enter commercial databases with a lag. Money is pouring into AI and a small number of adjacent infrastructure plays at the top, while companies outside the dominant AI thesis face a higher burden of proof.12 As The New York Times recently argued, the AI buildout is crowding out housing, manufacturing, energy infrastructure, and other investment.13 AI is bidding aggressively for scarce inputs in particular markets, and record financing totals conceal weaker conditions elsewhere. Economists disagree sharply about how much growth AI investment is actually generating. The Bureau of Economic Analysis reports spending categories such as information-processing equipment and software, not an “AI” contribution line. Estimates therefore depend on which categories an analyst classifies as AI-related, how imported chips and equipment are treated, and whether the calculation includes direct investment alone or broader spillovers.14 The obvious historical comparison is 1999 and 2000, and it is instructive precisely because it does not resolve cleanly in either direction. The dot-com crash showed that genuine technological change, useful infrastructure, and severe overvaluation can coexist. Today's leading hyperscalers and chip suppliers have far stronger revenues, cash flows, and access to investment-grade financing than many companies at the 2000 peak. The Bank for International Settlements' Annual Economic Report 2026 identified heavy debt issuance, circular financing, long-dated lease structures, and opaque private-credit links as potential amplifiers if AI expectations are repriced.15 The Fed's October 2025 minutes show several participants raising the possibility of a disorderly fall in equity prices tied to AI, while Governor Michael Barr has framed disappointing AI outcomes as a scenario in which risk shifts toward the financial sector.16 Senator Elizabeth Warren and colleagues separately asked the Financial Stability Oversight Council to investigate AI-related debt.17 The relevant bubble question is not whether artificial intelligence is real. Transformative technologies and mispriced financial claims frequently coexist. The question is whether demand, margins and utilization will arrive quickly enough to support the capital structures now being built around them. Equity investors bear valuation and dilution risk. Bondholders bear credit and refinancing risk. Lessors bear counterparty and residual-asset risk. Tenants bear long-duration utilization commitments. Private-credit funds bear leverage, collateral and hardware-obsolescence risk. Strategic suppliers and cloud providers bear both investment exposure and commercial concentration. This means an AI repricing would not remain confined to venture portfolios or technology equities. It could move through investment-grade debt, private credit, infrastructure funds, lease markets and the earnings of companies holding strategic equity stakes. The buildout may ultimately prove economically productive. That does not establish that every layer of the financing stack has been priced correctly.18 ## Sign up for AIxEnergy Journal The AIxEnergy Journal provides executive intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, reliability, investment, and policy. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## **Notes and Sources** 1\. PitchBook and National Venture Capital Association, “Q2 2026 PitchBook-NVCA Venture Monitor,” July 2026, https://pitchbook.com/news/reports/q2-2026-pitchbook-nvca-venture-monitor. 2\. OpenAI, “OpenAI Raises $122 Billion to Accelerate the Next Phase of AI,” March 31, 2026, https://openai.com/index/accelerating-the-next-phase-ai/. 3\. Anthropic, “Anthropic Raises $65B in Series H Funding at $965B Post-Money Valuation,” May 28, 2026, https://www.anthropic.com/news/series-h. 4\. MGX, “MGX Is Pleased to Confirm the Final Close of MGX Fund I at $49 Billion in Total Commitments,” July 1, 2026, https://www.mgx.ae/news-insights/mgx-pleased-confirm-final-close-mgx-fund-i-49-billion-total-commitments-significantly. 5\. Meta Platforms, Inc., “Meta Reports First Quarter 2026 Results,” April 29, 2026, https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Reports-First-Quarter-2026-Results/default.aspx; Reuters, “Meta Raises $25 Billion in Bond Sale After Lifting AI Spending Plan,” April 30, 2026, https://www.reuters.com/business/meta-looks-raise-up-25-billion-with-bond-sale-bloomberg-news-reports-2026-04-30/. 6\. Oracle Corporation, “Oracle Announces Record Q4 and FY 2026 Results Driven by Cloud Infrastructure and Cloud Applications,” June 10, 2026, https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Record-Q4-and-FY-2026-Results-Driven-by-Cloud-Infrastructure--Cloud-Applications/default.aspx; Oracle Corporation, Quarterly Report on Form 10-Q for the quarter ended February 28, 2026, https://www.sec.gov/Archives/edgar/data/1341439/000119312526101045/orcl-20260228.htm. 7\. Meta Platforms, Inc., “Meta Announces Joint Venture with Funds Managed by Blue Owl Capital to Develop Hyperion Data Center,” October 21, 2025, https://investor.atmeta.com/investor-news/press-release-details/2025/Meta-Announces-Joint-Venture-with-Funds-Managed-by-Blue-Owl-Capital-to-Develop-Hyperion-Data-Center/default.aspx. 8\. Executive Order, “Democratizing Access to Alternative Assets for 401(k) Investors,” August 7, 2025, https://www.whitehouse.gov/presidential-actions/2025/08/democratizing-access-to-alternative-assets-for-401k-investors/. 9\. Apollo Global Management, “Apollo Leads $35 Billion Capital Solution for Broadcom AI XPV Platform in Partnership with Blackstone and Leading Global Banks,” June 9, 2026, https://ir.apollo.com/news-events/press-releases/detail/629/apollo-leads-35-billion-capital-solution-for-broadcom-ai. 10\. Amazon.com, Inc., “First Quarter 2026 Financial Results,” April 29, 2026, https://www.sec.gov/Archives/edgar/data/1018724/000101872426000012/amzn-20260331xex991.htm. 11\. Alphabet Inc., “Alphabet Announces First Quarter 2026 Results,” April 29, 2026, https://www.sec.gov/Archives/edgar/data/1652044/000165204426000043/googexhibit991q12026.htm. 12\. Gené Teare, “Fintech Funding Surges 23% in H1 2026 as Investors Concentrate Capital in Fewer Deals,” Crunchbase News, July 15, 2026, https://news.crunchbase.com/fintech/funding-rises-deals-slump-h1-2026/; Joanna Glasner, “North American Startup Funding Shattered Records in the First Half,” Crunchbase News, July 7, 2026, https://news.crunchbase.com/venture/na-startup-funding-ma-shattered-records-ai-q2-2026/. 13\. Jennifer M. Harris, “The Generational Force Hollowing Out the Economy,” New York Times, June 29, 2026, https://www.nytimes.com/2026/06/29/opinion/ai-economy-affordability.html. 14\. Federal Reserve Bank of St. Louis, “Tracking AI’s Contribution to GDP Growth,” January 2026, https://www.stlouisfed.org/on-the-economy/2026/jan/tracking-ai-contribution-gdp-growth. 15\. Bank for International Settlements, Annual Economic Report 2026, June 28, 2026, https://www.bis.org/publ/arpdf/ar2026e.htm. 16\. Board of Governors of the Federal Reserve System, “Minutes of the Federal Open Market Committee, October 28–29, 2025,” November 19, 2025, https://www.federalreserve.gov/monetarypolicy/fomcminutes20251029.htm; Michael S. Barr, “Artificial Intelligence and the Labor Market,” speech delivered at the New York Association for Business Economics, New York, February 17, 2026, https://www.federalreserve.gov/newsevents/speech/barr20260217a.htm. 17\. US Senate Committee on Banking, Housing, and Urban Affairs, “Warren, Colleagues Press FSOC to Launch Probe into Financial Stability Risks of AI Debt Bubble,” January 22, 2026, https://www.banking.senate.gov/newsroom/minority/warren-colleagues-press-fsoc-to-launch-probe-into-financial-stability-risks-of-ai-debt-bubble. 18\. Bank for International Settlements, Annual Economic Report 2026. ### The Moratorium Map: How the AI Boom Is Redrawing the Geography of Power and Permission URL: https://www.aixenergy.io/the-moratorium-map-how-the-ai-boom-is-redrawing-the-geography-of-power-and-permission/ Last updated: 2026-08-04T20:59:16.000Z Executive Brief ## Key Takeaways for Decision Makers 1 **Policy access is now an infrastructure constraint.**Land, power, fiber and capital do not guarantee that a data-center project can proceed. \`\`\` 2 **Local decisions have system-level consequences.**AIxEnergy identifies 134 current hard restrictions, including 131 associated with local or tribal jurisdictions. 3 **The central issue is ownership of uncertainty.**Policymakers increasingly want developers to bear more risk when proposed loads are delayed, reduced or canceled. 4 **Electric service is becoming conditional.**Large loads may need to bring supply, accept curtailment, meet performance standards or provide financial protection. 5 **A moratorium is a process, not a permanent classification.**Replacement rules, governing institutions and unresolved approval gates matter more than the label alone. 6 **Policy diligence must precede land control.**An attractive site may remain unfinanceable without a credible path through zoning, grid access and cost allocation. 7 **Legibility may matter more than permissiveness.**Clear requirements, predictable timelines and transparent cost rules create a stronger investment environment. \`\`\` On July 28, 2026, as[ AIxEnergy](https://www.aixenergy.io/) completed the first full release of [The Moratorium Map](https://map.aixenergy.io/?ref=aixenergy.io), Reuters reported that PJM Interconnection was preparing two filings with federal regulators to address a widening gap between electricity demand and available supply. PJM’s latest capacity auction ended approximately 6.8 gigawatts below the system’s reliability requirement despite prices reaching the auction cap. The grid operator forecasts roughly 70 gigawatts of additional large-load demand by 2038\. Its planned filings include a registry identifying the location and electricity requirements of large loads of at least 50 megawatts and a proposed framework under which certain new loads that have not secured sufficient supply could face curtailment during capacity shortages. The framework remains subject to federal review and, in areas involving retail service or direct customer curtailment, state authority and cooperation.¹ This is not a conventional data-center moratorium. It is potentially more consequential. A local moratorium temporarily withholds permission to build. PJM’s proposal would make the quality of access to the operating power system conditional. A data center might possess land, permits, financing and a utility service arrangement while still facing a different reliability position during stressed conditions if it has not brought or contracted for sufficient supply. The logic behind the moratorium is moving from the zoning office toward the grid control room. PJM’s action did not emerge in isolation. On June 18, the Federal Energy Regulatory Commission issued show-cause orders directing all six regional grid operators under its jurisdiction either to justify their existing tariffs or propose reforms for integrating data centers and other large energy users. The orders address transmission-study processes, cost transparency, protection against cost shifting, co-location with generation, flexible transmission service and the study of generating facilities serving nearby large loads.² 💡 Three major actions in forty days—federal, state and regional—reveal the larger pattern. The AI infrastructure buildout has entered the age of permission. On July 14, New York State issued Executive Order 62, which the governor’s administration describes as the nation’s first statewide moratorium on new hyperscale data centers. The order directs the Department of Environmental Conservation to hold in abeyance specified incomplete applications for discretionary state permits associated with qualifying new or expanded data centers capable of consuming 50 megawatts or more while the state prepares a generic environmental impact statement. It does not categorically suspend local approvals, and it excludes certain manufacturing, research, education and medical facilities.³ ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/US_Policy_upscayl_3x_upscayl-standard-4x.png) *Source:* [*AIxEnergy Map*](https://map.aixenergy.io/?ref=aixenergy.io) ## A new control layer around the AI economy The physical expansion of artificial intelligence is occurring at a speed that energy systems and governing institutions were not designed to match. The [International Energy Agency](https://www.iea.org/?ref=aixenergy.io) estimates that global data-center electricity consumption increased 17 percent in 2025 and could roughly double from 485 terawatt-hours in 2025 to 950 terawatt-hours in 2030\. Lawrence Berkeley National Laboratory’s June 2026 U.S. update projects a reference case of 649 terawatt-hours by 2030 and a compounded uncertainty range equivalent to between 9.5 and 15.3 percent of total U.S. electricity consumption.⁴ But the problem is not simply the aggregate quantity of energy. Data centers are unusually large, geographically concentrated and fast to develop relative to much of the infrastructure required to serve them. Developers can propose, resize, phase, postpone or cancel a campus faster than utilities can permit and construct transmission lines, substations and generating resources. The developer retains substantial discretion over whether and when demand materializes. The power system may need to plan and invest before that uncertainty is resolved. That asymmetry is the policy problem beneath the headlines. Who should pay for infrastructure built in anticipation of a load that arrives late or does not arrive at all? What claim should a new data center have on limited generating capacity during a reliability emergency? How much water, land and local infrastructure should a host community commit before a project’s benefits and operating characteristics are sufficiently established? What evidence should regulators require before allowing one customer’s private investment schedule to influence a regional power system? Moratoria are one answer—but only one. Across the world, governments and system operators are assembling a broader permission architecture composed of land-use restrictions, environmental reviews, controlled capacity allocations, connection requirements, performance reporting, special tariffs, financial security and community-investment expectations. [**The Moratorium Map**](https://map.aixenergy.io/?ref=aixenergy.io) documents that emerging architecture. ## What the map shows The dataset contains 234 normalized policy-event records across 230 rendered geographies as of July 28, 2026\. Four records are incorporated into aggregated map popups rather than displayed as separate geographic features. The dataset includes 213 U.S. records and 21 international records. It identifies 134 active or extended hard legal restrictions, together with pending proposals, conditional grid-access rules, reporting obligations, market-monitoring signals and historical or superseded policies. The classifications are intentionally narrower than the language often used in public debate. A proposed restriction is not an enacted restriction. A utility connection standard is not a zoning ban. A reporting rule is not a moratorium. An expired policy should not remain on a current-risk map because it once produced a headline. An area facing power constraints should not be presented as legally closed when no enforceable restriction exists. ### Exhibit 1\. Policy-constraint taxonomy Policy Classification ## Policy-Constraint Taxonomy Current legal restrictions account for more than half of the documented policy landscape, while historical actions, proposals and market signals remain important parts of the broader intelligence picture. Distribution of 234 data-center policy-event records Hard legal restrictions represent 57.3 percent, conditional-access rules 2.1 percent, reporting obligations 0.4 percent, political proposals 10.7 percent, market-monitoring signals 9 percent, and historical or replaced policies 20.5 percent. Hard legal restriction: 134 records, 57.3% Conditional-access rule: 5 records, 2.1% Reporting obligation: 1 record, 0.4% Political proposal: 25 records, 10.7% Market-monitoring signal: 21 records, 9.0% Historical or replaced: 48 records, 20.5% 234 Policy Events - Hard legal restriction 134 57.3% - Historical or replaced 48 20.5% - Political proposal 25 10.7% - Market-monitoring signal 21 9.0% - Conditional-access rule 5 2.1% - Reporting obligation 1 0.4% **Key finding:** Active or extended hard legal restrictions represent 57.3% of all tracked policy events. The remaining 42.7% captures the broader policy lifecycle—including proposals, monitoring signals and earlier measures that have expired or been replaced. Source: AIxEnergy Global Data Center Policy Constraints, version 2.0.0, July 28, 2026\. Shares are calculated across 234 policy-event records. The chart classifies 234 policy-event records into six categories. *Source: AIxEnergy, Global Data Center Policy Constraints, version 2.0.0.* The resulting map is not a list of places that oppose data centers. It is a map of the institutions asserting authority over them. That distinction is central. “Restrictive” and “permissive” are inadequate descriptions of markets in which access depends on project design, location, energy strategy, financial commitments and the ability to satisfy multiple decision-makers. ## The moratorium is a symptom, not the system The United States accounts for more than 90 percent of the records in the current dataset, and the overwhelming majority of documented hard restrictions are local. Of the 134 active or extended hard legal restrictions, 131 are associated with counties, cities, townships, towns, villages or tribal jurisdictions. Only three are classified at a state, utility-authority or broader level. This distribution reflects the fragmented structure of American land-use authority. It also reveals a mismatch between the scale at which the benefits of AI infrastructure are described and the scale at which many of its effects are experienced. The national government sees strategic competition, productivity, technological leadership and investment. A technology company sees computing capacity, latency, market share and speed to deployment. A utility sees a customer that may require new substations, transmission, generation and reserve capacity. A county sees rezoning, construction traffic, noise, backup generation, water demand, tax arrangements and a facility that may employ relatively few permanent workers in proportion to its physical and electrical footprint. A household sees a simpler set of questions: Will this project affect my electricity bill, public services or community? Moratoria tend to emerge where existing institutions have not yet produced a broadly accepted answer. Prince George’s County, Maryland, illustrates the transition. The county established a task force to examine the effects of data centers on energy demand, taxpayers and ratepayers, air, water, woodlands, land use and quality of life. In 2026, county leaders advanced a two-year moratorium intended to create time for a broader review of land use, infrastructure capacity, environmental effects, public safety and the long-term economic implications of development.⁵ The jurisdiction did not simply discover that it opposed technology. It confronted a development model for which its existing rules and analytical capacity were considered inadequate. That is a useful way to interpret many local moratoria. They are institutional pauses used by governments that must make durable land-use and infrastructure decisions using standards often developed for smaller, slower and less electricity-intensive projects. A moratorium buys time. The more important question is what a jurisdiction builds with that time. ## The geography of resistance The documented restrictions are highly concentrated. Ohio and Michigan account for 49 of the 134 current hard restrictions—36.6 percent of the total. The five leading states—Ohio, Michigan, North Carolina, Georgia and Indiana—account for 82, or 61.2 percent. ### Exhibit 2\. Leading states in the current dataset Scroll horizontally to view all columns. | State | Total records | Current hard restrictions | Pending | Monitoring | Historical or replaced | Grade A or B | | -------------- | ------------- | ------------------------- | ------- | ---------- | ---------------------- | ------------ | | Ohio | 35 | 27 | 2 | 1 | 5 | 27 | | Michigan | 30 | 22 | 3 | 1 | 4 | 23 | | North Carolina | 15 | 13 | 1 | 0 | 1 | 13 | | Georgia | 25 | 11 | 0 | 1 | 13 | 11 | | Indiana | 11 | 9 | 0 | 0 | 2 | 9 | | Iowa | 11 | 8 | 1 | 0 | 2 | 8 | | Kansas | 8 | 8 | 0 | 0 | 0 | 8 | | North Dakota | 7 | 4 | 0 | 0 | 3 | 4 | | Tennessee | 7 | 4 | 0 | 0 | 2 | 5 | | Oklahoma | 5 | 4 | 1 | 0 | 0 | 5 | *Source: AIxEnergy, Global Data Center Policy Constraints, version 2.0.0.* These figures should not be converted into a simplistic ranking of state hostility. They are counts of documented policy events, not a statistically normalized measure of restrictiveness. States differ in population, number of local governments, development pressure, media coverage, access to primary documents and the maturity of external tracking. A state with fewer formal moratoria may already possess specialized zoning or utility rules. It may also be earlier in the development cycle or less fully represented in the source record. The concentration nevertheless carries an important signal. The policy frontier is moving beyond established data-center hubs and into communities encountering hyperscale development for the first time. These jurisdictions may welcome investment while remaining uncertain about grid costs, employment benefits, noise, water, land conversion and the reliability of long-term demand projections. The conflict is not necessarily between communities that support technology and those that oppose it. It is often between the speed of private capital and the speed at which public institutions can establish acceptable terms of entry. ## The hidden issue is ownership of uncertainty The public debate is usually framed around electricity consumption: data centers require too much power. However, the more difficult issue is who owns the risk that future demand will differ from the forecast. A utility may need to initiate infrastructure investments years before a campus reaches its announced load. If the project is delayed, downsized or canceled, the physical assets and contractual commitments may remain. Unless the applicable tariff, contract or regulatory order assigns the risk elsewhere, some portion may migrate to other customers. The IEA has highlighted the mismatch between the speed of data-center development and the longer timelines required for energy infrastructure, as well as uncertainty around actual peak demand and the possibility that connection capacity may be reserved before server installations are complete.⁶ This helps explain why the policy debate is shifting from whether to serve data centers to the conditions under which they should be served. PJM’s July proposal would exclude certain incremental large loads from the demand used in future capacity auctions when those loads have not brought or contracted for the new supply required to serve them. The PJM board stated that existing customers should not bear higher capacity costs caused by new loads that arrive without sufficient supply. Implementation would still depend on FERC review and the respective responsibilities of PJM, utilities and state regulators. New York’s order reaches toward the same policy question through different institutions. It directs state agencies to evaluate mechanisms that could include upfront developer contributions, demand-response participation, new clean supply and storage, an insurance pool and protections against infrastructure costs created by project delay, downsizing or cancellation. These are subjects for policy development and regulatory consideration; they are not yet final tariff obligations. FERC’s show-cause orders similarly identify prevention of cost shifting and transparency into transmission costs as central areas for regional grid operators to address. A common policy direction is emerging: the customer controlling the load-development decision may be required to bear more of the financial consequences when the project changes. That direction could reshape study deposits, interconnection agreements, construction advances, milestone requirements, minimum bills, exit charges and the treatment of customer-dedicated supply. It may ultimately matter more than the moratoria themselves. ## Four models of conditional permission The Moratorium Map reveals four emerging models. They overlap, but each controls a different decision point. ### 1\. Pause and rewrite This is the dominant U.S. local-government model. A jurisdiction temporarily suspends approvals while it studies eligible zoning districts, setbacks, noise, water use, transmission, air permits, emergency response, tax treatment or community benefits. The pause may expire, be extended or be replaced by permanent rules. The strategic mistake is to treat the moratorium as the final policy. The consequential work occurs during the pause: who writes the replacement ordinance, what evidence is admitted, how thresholds are defined, whether existing applications are grandfathered, and whether the final process creates a viable conditional pathway or an effective prohibition. Only eight current or extended records in the dataset contain structured future expiration dates. That does not make the remaining restrictions permanent. It demonstrates how often expiration terms are missing, ambiguous or embedded in local legal instruments that require recurring verification. ### Exhibit 3\. Near-term policy-expiration calendar Scroll horizontally to view all columns. | Expiration | Jurisdiction | State | Status | Evidence | Current action | | ---------------------- | --------------------------- | ---------- | -------- | -------- | -------------------------------------------------------------------------------------------------------------- | | Sept. 10, 2026 | Sedgwick County | Kansas | Extended | A | Pause extended while land-use regulations are completed | | Sept. 30, 2026 | DeKalb County | Georgia | Extended | A | Extension during preparation of a data-center text amendment | | Nov. 20, 2026 | Pittsfield Township | Michigan | Extended | A | Temporary moratorium renewed for six months | | Dec. 31, 2026 | Clayton County | Georgia | Extended | A | Pause continued while permanent rules are developed | | Approx. April 13, 2027 | Oakley | California | Extended | B | Extension while the city considers permanent rules | | May 11, 2027 | Marion County | Kansas | Extended | B | One-year extension adopted in May 2026 | | July 14, 2027 | New York Executive Order 62 | New York | Active | A | Specified incomplete applications for discretionary state permits paused during a generic environmental review | | July 7, 2028 | Prince George’s County | Maryland | Extended | A | Two-year pause subject to subsequent county action | *Source: AIxEnergy, Global Data Center Policy Constraints, version 2.0.0\. Dates describe the currently documented terms and may be superseded by subsequent action.* An expiration date is not the date at which risk necessarily disappears. A temporary pause may be replaced by more demanding permanent rules, and an expired moratorium may leave the underlying political conflict unresolved. For developers and investors, the replacement process matters more than the countdown clock. ### 2\. Conditional grid access Ireland provides a clearer model of what follows when the power system itself becomes the gatekeeper. Under the connection policy adopted by Ireland’s Commission for Regulation of Utilities, new data centers must provide generation or storage capacity matching their requested maximum import capacity, either onsite or nearby. They must also meet at least 80 percent of annual demand through additional renewable-electricity projects in Ireland, subject to a six-year development pathway. System operators must consider whether each proposed connection is located in a constrained part of the network. Irish data centers consumed 22 percent of national electricity in 2024 and, based on contracted demand, are projected to reach 31 percent by 2034.⁷ This is neither a general prohibition nor an unrestricted right to connect. It is conditional admission to the system. Singapore uses controlled capacity allocation to pursue a related objective. Its Green Data Centre Roadmap seeks to provide at least 300 megawatts of additional capacity in the near term, with expansion linked to improved energy efficiency and lower-carbon energy pathways.⁸ Under this model, power is not merely purchased. It is allocated in exchange for specified performance. ### 3\. Transparency before performance The European Union is building a third model through information. The Energy Efficiency Directive established data-center reporting requirements, while Delegated Regulation EU 2024/1364 defined common information and performance indicators. The European Commission is using the resulting data to advance a Union-wide rating system and work on minimum performance standards.⁹ Reporting may appear less restrictive than a moratorium. Over time, it can be more structurally significant. Once governments possess comparable information on energy use, water, renewable sourcing and efficiency, they can incorporate those metrics into permitting, procurement, financing, tax treatment and minimum operating standards. The policy sequence is familiar: disclosure, classification, benchmarking and, potentially, obligation. ### 4\. The negotiated infrastructure compact New York’s Executive Order 62 points toward a fourth model: permission linked to a broader package of project commitments. The order directs the state to develop or evaluate frameworks addressing: - community investment and local infrastructure; - labor practices and public transparency; - water withdrawals and environmental review; - large-load interconnection and grid investment; - beneficiary-pays cost allocation; - demand response and system flexibility; and - protection of ratepayers from stranded or underused infrastructure. It also establishes a public environmental-review process and assigns roles to multiple state agencies and institutions. The final form of these policies has not yet been determined. The significance extends beyond conventional permitting. Under this model, a data center is treated as a major infrastructure participant whose entry must be reconciled with the power system, public institutions and the host community. Capital investment and projected tax revenue may therefore no longer be sufficient on their own. A credible project increasingly must answer six questions: *What infrastructure does the project require?* *Who will pay for that infrastructure?* *How will the load behave during constrained or stressed system conditions?* *What protections apply if the project is delayed, downsized or canceled?* *What environmental resources will the project consume?* *What durable value will remain in the host community?* This is the emerging social contract of hyperscale development: permission in exchange for clearer accountability, more transparent risk allocation and a broader public-value proposition. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/Europe_Policy_upscayl_3x_upscayl-standard-4x.png) *Source:* [*AIxEnergy Map*](https://map.aixenergy.io/?ref=aixenergy.io) ## What the Moratorium Map means for capital Policy geography should now enter site selection before—not after—land control and major interconnection spending. A conventional screen emphasizes available power, fiber, land, taxes, water and latency. Those factors remain essential, but they do not establish whether a project has a viable institutional path. > *The investment-grade question is no longer simply, “Does this jurisdiction have a moratorium?”* > > *It is: Who controls each decision, what legal standard applies, what evidence is required, what costs must the developer secure, and what unresolved event could stop the project?* The distinction matters because policy signals are not equally consequential. A pending legislative proposal may matter less than a local zoning rewrite already approaching adoption. An expired moratorium may represent low risk if it produced a clear approval process—or high risk if community opposition and infrastructure questions remain unresolved. A market-monitoring signal is not a legal restriction, but it may still indicate emerging grid scarcity, water constraints or political resistance capable of affecting schedule, financing and valuation. ### The Four Forms of Permission **1\. Land permission** Is the proposed use allowed, conditionally allowed or temporarily suspended? **2\. Grid permission** Is electricity available as firm service, conditional service or only with dedicated supply, storage or demand flexibility? **3\. Economic permission** Are cost-allocation rules, financial-security requirements, minimum payments and cancellation obligations sufficiently defined? **4\. Political permission** Is there a durable local bargain—or merely a legally valid approval vulnerable to reversal, litigation or electoral change? A site that passes only the land-use test is not development-ready. ## What governments should build instead of permanent emergency Moratoria are useful emergency brakes. They are poor operating systems. A durable large-load pathway should require developers to fund appropriate studies, provide milestone security, disclose credible load ramps, protect customers against cancellation risk and identify the supply, storage or flexibility required to support the requested service. It should also create reciprocal obligations for utilities and governments. Utilities should publish study requirements, connection options, capacity constraints and credible schedule ranges. Regulators should define cost-causation principles before individual controversies arise. Local governments should establish zoning criteria, public-participation requirements and community-investment expectations before a site-specific application forces them to invent policy under pressure. Conditional access cannot become arbitrary access. The best regimes will not necessarily be the most permissive. They will be the most legible. Legibility allows serious projects to price their obligations and distinguish themselves from speculative announcements. It also gives communities a basis for determining whether the proposed bargain is credible. ## Using the Moratorium Map The Policy Constraints layer is available to all free and paid [AIxEnergy](https://www.aixenergy.io/) subscribers through the[ AIxEnergy Map](https://map.aixenergy.io/?ref=aixenergy.io)[.](https://map.aixenergy.io/?ref=aixenergy.io) Users can distinguish current hard legal restrictions from conditional-access rules, reporting obligations, political proposals, monitoring signals and expired or replaced actions. Each record includes legal status, policy type, summary, evidence grade, source information and, where available, effective and expiration dates. Related policies are aggregated within a single geographic popup so that an original moratorium, later extension and replacement action are not presented as unrelated places. The map is designed for screening and research—not parcel-level legal interpretation. Its geometries include generalized administrative boundaries and approximate market areas. International coverage remains preliminary relative to the depth of U.S. local-government coverage. ## Disclosure This article represents independent analysis published by [AIxEnergy](https://www.aixenergy.io/aixenergy-convergence-map-technologies-companies-deployments-and-capital-flows/). It does not represent the views or positions of any other individual or organization. ## Notes 1. Laila Kearney and Kavya Balaraman, “[Largest US Power Grid Plans to Cover Widening Electricity Shortages as Data Center Demand Swells](https://www.reuters.com/business/energy/largest-us-power-grid-moves-ahead-with-plan-meet-swelling-demand-2026-07-28/?ref=aixenergy.io),” *Reuters*, July 28, 2026; PJM Interconnection, “[PJM Board Directs Action on Resource Adequacy, Affordability and Large Loads](https://insidelines.pjm.com/pjm-board-directs-action-on-resource-adequacy-affordability-and-large-loads/?ref=aixenergy.io),” July 27, 2026. 2. Federal Energy Regulatory Commission, “[FERC Launches Aggressive Targeted Action to Speed Large Load Integration](https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration?ref=aixenergy.io),” June 18, 2026. 3. New York State, Executive Order No. 62, “[Establishing a Temporary Moratorium on Data Centers in New York While the State Develops Higher Standards for Data Center Development and Benefits Blueprint to Support Localities](https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data-centers-new-york-while-state-develops?ref=aixenergy.io),” July 14, 2026; Office of Governor Kathy Hochul, “[First Statewide Moratorium on New Hyperscale Data Centers Launched by Governor Kathy Hochul](https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul?ref=aixenergy.io),” July 14, 2026. 4. International Energy Agency, [*Key Questions on Energy and AI*](https://www.iea.org/reports/key-questions-on-energy-and-ai?ref=aixenergy.io) (Paris: International Energy Agency, 2026); Sarah Josephine Smith et al., [*United States Data Center Energy Usage Report: 2025 Update*](https://eta.lbl.gov/publications/united-states-data-center-energy-2025?ref=aixenergy.io) (Berkeley, CA: Lawrence Berkeley National Laboratory, June 2026). 5. Prince George’s County Council, “[Qualified Data Center Task Force](https://pgccouncil.us/1051/Qualified-Data-Center-Task-Force?ref=aixenergy.io),” accessed July 28, 2026; Prince George’s County Council, “[Chair Oriadha Announces Two-Year Moratorium to Prevent New Data Centers from Operating in the County](https://pgccouncil.us/CivicAlerts.aspx?AID=2073&ref=aixenergy.io),” June 30, 2026. 6. International Energy Agency, *Key Questions on Energy and AI*, especially the discussion of uncertain peak loads, reserved connection capacity, and the alignment of power-system investment with data-center demand. 7. Commission for Regulation of Utilities, “[The CRU Publishes Its Decision on New Electricity Connection Policy for Data Centres](https://www.cru.ie/about-us/news/the-cru-publishes-its-decision-on-new-electricity-connection-policy-for-data-centres/?ref=aixenergy.io),” December 12, 2025. 8. Infocomm Media Development Authority, “[Singapore Announces Green Data Centre Roadmap for Sustainable Growth](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2024/sg-announces-green-data-centre-roadmap?ref=aixenergy.io),” May 30, 2024. 9. European Commission, Directorate-General for Energy, “[Energy Performance of Data Centres](https://energy.ec.europa.eu/topics/energy-efficiency/energy-efficiency-targets-directive-and-rules/energy-efficiency-directive/energy-performance-data-centres%5Fen?ref=aixenergy.io),” accessed July 28, 2026; European Commission, Directorate-General for Energy, “[Minimum Performance Standards for EU Data Centres](https://energy.ec.europa.eu/resources/preparatory-studies/minimum-performance-standards-eu-data-centres%5Fen?ref=aixenergy.io),” accessed July 28, 2026. ## Bibliography AIxEnergy. *Global Data Center Policy Constraints*. Version 2.0.0\. Updated July 28, 2026\. Proprietary public-source research dataset. Commission for Regulation of Utilities. “[The CRU Publishes Its Decision on New Electricity Connection Policy for Data Centres](https://www.cru.ie/about-us/news/the-cru-publishes-its-decision-on-new-electricity-connection-policy-for-data-centres/?ref=aixenergy.io).” December 12, 2025. European Commission, Directorate-General for Energy. “[Energy Performance of Data Centres](https://energy.ec.europa.eu/topics/energy-efficiency/energy-efficiency-targets-directive-and-rules/energy-efficiency-directive/energy-performance-data-centres%5Fen?ref=aixenergy.io).” Accessed July 28, 2026. ———. “[Minimum Performance Standards for EU Data Centres](https://energy.ec.europa.eu/resources/preparatory-studies/minimum-performance-standards-eu-data-centres%5Fen?ref=aixenergy.io).” Accessed July 28, 2026. Federal Energy Regulatory Commission. “[FERC Launches Aggressive Targeted Action to Speed Large Load Integration](https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration?ref=aixenergy.io).” June 18, 2026. Infocomm Media Development Authority. “[Singapore Announces Green Data Centre Roadmap for Sustainable Growth](https://www.imda.gov.sg/resources/press-releases-factsheets-and-speeches/press-releases/2024/sg-announces-green-data-centre-roadmap?ref=aixenergy.io).” May 30, 2024. International Energy Agency. [*Key Questions on Energy and AI*](https://www.iea.org/reports/key-questions-on-energy-and-ai?ref=aixenergy.io). Paris: International Energy Agency, 2026. Kearney, Laila, and Kavya Balaraman. “[Largest US Power Grid Plans to Cover Widening Electricity Shortages as Data Center Demand Swells](https://www.reuters.com/business/energy/largest-us-power-grid-moves-ahead-with-plan-meet-swelling-demand-2026-07-28/?ref=aixenergy.io).” *Reuters*, July 28, 2026. New York State. Executive Order No. 62\. “[Establishing a Temporary Moratorium on Data Centers in New York While the State Develops Higher Standards for Data Center Development and Benefits Blueprint to Support Localities](https://www.governor.ny.gov/executive-order/no-62-establishing-temporary-moratorium-data-centers-new-york-while-state-develops?ref=aixenergy.io).” July 14, 2026. Office of Governor Kathy Hochul. “[First Statewide Moratorium on New Hyperscale Data Centers Launched by Governor Kathy Hochul](https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul?ref=aixenergy.io).” July 14, 2026. PJM Interconnection. “[PJM Board Directs Action on Resource Adequacy, Affordability and Large Loads](https://insidelines.pjm.com/pjm-board-directs-action-on-resource-adequacy-affordability-and-large-loads/?ref=aixenergy.io).” July 27, 2026. Prince George’s County Council. “[Chair Oriadha Announces Two-Year Moratorium to Prevent New Data Centers from Operating in the County](https://pgccouncil.us/CivicAlerts.aspx?AID=2073&ref=aixenergy.io).” June 30, 2026. ———. “[Qualified Data Center Task Force](https://pgccouncil.us/1051/Qualified-Data-Center-Task-Force?ref=aixenergy.io).” Accessed July 28, 2026. Smith, Sarah Josephine, Alex Hubbard, Alexander Newkirk, Mohan Ganeshalingam, Billie Holecek, Dale A. Sartor, Michael Mills, and Arman Shehabi. [*United States Data Center Energy Usage Report: 2025 Update*](https://eta.lbl.gov/publications/united-states-data-center-energy-2025?ref=aixenergy.io). Berkeley, CA: Lawrence Berkeley National Laboratory, June 2026. ### AI’s Grid Problem Is Not the Prompt. It Is the Commitment URL: https://www.aixenergy.io/beyond-the-prompt-washingtons-ai-power-pledge-tests-who-pays-for-the-grid/ Last updated: 2026-07-29T21:03:24.000Z The White House expanded its Ratepayer Protection Pledge on July 23, announcing that more than 200 additional utilities, data-center developers, cooperatives, states, and other participants had joined the effort. The administration says the pledge now covers 80 percent of electricity delivered to U.S. homes and businesses and 263 million people. Those figures describe the reach claimed by the administration, not an independent audit of customer protection.\[1\] A pledge is not a tariff, a commission order, or an enforceable service agreement. It does not by itself determine which facilities a data-center customer must finance, how shared network costs should be divided, what happens when a project is delayed, or who pays if contracted electricity is never used. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/ChatGPT-Image-Jul-26--2026--12_51_59-AM-1.png) Its importance is nonetheless substantial. The federal government is giving national political force to a basic cost-causation principle: large computing customers should finance the incremental generation and infrastructure required to serve them and should continue paying for capacity procured on their behalf even when their actual use falls short. The original pledge calls for separate rate structures and commitments to pay for power and related infrastructure whether or not the electricity is ultimately consumed.\[1\] That principle moves the AI-energy debate away from one of its most visible but least useful symbols: the electricity assigned to a single prompt. The immediate infrastructure question is not whether one chatbot response consumes a fraction of a watt-hour or several watt-hours. It is whether a 100-megawatt, 500-megawatt, or gigawatt-scale campus can be supplied on schedule—and what happens when utilities begin ordering equipment, reserving capacity, acquiring land, reinforcing transmission, and procuring generation before the project’s final size and operating date are certain. The public consequences of AI are increasingly determined before the first prompt is typed. ## The grid plans for campuses, not prompts Per-query energy estimates remain relevant to model design, efficiency improvement, product comparisons, and emissions accounting. They are poorly suited to deciding whether a particular data-center campus requires a new substation, transmission line, generation resource, water supply, or long-term capacity commitment. Research on AI inference shows why. Energy consumption can vary by orders of magnitude depending on the task, model, hardware, output length, utilization, batching, and operating environment. A production-oriented analysis estimated a median of 0.34 watt-hours for a typical frontier-model query under its assumptions, increasing to 4.32 watt-hours when test-time computation raised token use fifteenfold. The same study found that combined improvements in model design, serving systems, and hardware could plausibly reduce energy per query by eight- to twentyfold.\[2\] These are useful efficiency findings. They do not provide a universal unit for grid planning. Utilities do not reserve transformers or procure capacity in response to individual prompts. They invest against projected campus load, service level, ramp schedule, utilization, location, operating profile, and the probability that the customer will actually arrive. The relevant unit is therefore not the prompt. It is the project commitment. ## Uncertainty is now a planning condition The scale of prospective data-center demand makes this distinction consequential. The International Energy Agency estimated that data centers consumed approximately 415 terawatt-hours of electricity worldwide in 2024, about 1.5 percent of global consumption. Its base case reaches approximately 945 terawatt-hours in 2030\. The global share remains limited relative to total electricity use, but the demand is highly concentrated and can materialize faster than generation, transmission, equipment manufacturing, and permitting.\[3\] For the United States, Lawrence Berkeley National Laboratory’s June 2026 update estimated reference-case data-center electricity consumption of 649 terawatt-hours in 2030\. Its compounded uncertainty range extends from 521 to 843 terawatt-hours, equivalent to approximately 9.5 to 15.3 percent of total U.S. electricity consumption.\[4\] EPRI independently estimated that data centers could account for 9 to 17 percent of U.S. electricity consumption by 2030\. In Virginia—the country’s most concentrated data-center market—EPRI projected a 39 to 57 percent share across its scenarios.\[5\] These projections do not establish that every announced campus will be built. Their breadth is itself the planning fact. Utilities and system operators may need to make physical and financial commitments before the industry’s ultimate load trajectory is known. The uncertainty also cannot be managed through national averages. In PJM’s Dominion zone, summer peak load reached 23,905 megawatts in 2025, 23 percent above 2019\. The 2025–26 winter peak was 45 percent above the 2019–20 level. PJM expects the zone to experience the largest absolute increase in summer peak demand through 2030, largely because of data-center growth.\[6\] A nationally manageable percentage can therefore become a dominant local infrastructure problem. ## The commitment sequence creates the risk The central large-load problem is not simply the amount of electricity requested. It is the sequence in which each party must commit. A developer may request substantial capacity to preserve an option on a site. The utility may then perform studies, reserve engineering resources, acquire property, procure supply, order long-lead equipment, or begin constructing facilities. The customer may not make its final investment decision until it receives a credible service date and price. The utility may be unable to provide that certainty until it undertakes work that creates cancellation exposure. Each party wants evidence before making an irreversible commitment. Each may depend on the other party’s commitment to produce that evidence. Equipment constraints intensify the circularity. Reuters reported in July 2026 that generator step-up transformer lead times had exceeded 160 weeks and that high-voltage circuit-breaker lead times had reached 125 weeks. Some utilities and developers were securing production slots years before expected need.\[7\] Once equipment has been ordered, land acquired, power procured, or construction started, the cost does not disappear if the customer postpones the project, reduces its load, moves to another location, or abandons the development. Someone must absorb it. The customer can pay through deposits, advances, minimum bills, guarantees, or exit charges. Utility shareholders can absorb some costs through disallowance or reduced returns. Other customers can pay through rates. Taxpayers can pay through subsidies or public financing. Suppliers may bear cancellation costs through negotiated terms. The policy question is therefore not whether project risk exists. It is where that risk moves as commitments become less reversible. ## Protection should rise with irreversibility No single customer guarantee can cover every form of development risk. Ratepayer protection requires a portfolio of instruments activated at the point when each exposure is created. At the preliminary-request stage, a study deposit can ensure that the customer reimburses identifiable engineering and administrative work. When the utility begins reserving equipment or manufacturing capacity, the customer’s security should reflect actual supplier cancellation charges, restocking exposure, and other procurement liabilities. Before construction begins, the customer should provide evidence of financing, site control, permitting progress, internal authorization, and a credible development schedule. Construction advances or credit support can then correspond to the portion of investment that becomes irreversible. When the utility is ready to provide service, minimum-demand charges or minimum bills can protect against low utilization of generation, transmission, and capacity procured for the project. If the customer exits early, termination charges can recover reasonably attributable costs that remain after equipment reuse, resale, reassignment, depreciation, insurance, and other mitigation measures. These instruments are complements, not substitutes. A study deposit does not cover a canceled transformer order. A construction contribution does not necessarily cover long-term capacity procurement. A minimum bill protects against underutilization after service begins but may provide little protection if the project disappears before energization. The governing principle should be straightforward:Customer obligations should rise as the utility’s commitments become less reversible. That is more defensible than demanding a large undifferentiated guarantee at the beginning of development. It also avoids allowing a customer to retain unrestricted exit rights after the utility has undertaken costly and project-specific work. ## Cost causation is a principle, not a formula Saying that a data center should pay the costs it causes does not resolve every allocation question. The hardest cases involve infrastructure that serves the project while also providing broader system value. A dedicated substation serving one campus can usually be attributed directly. A transmission reinforcement may be more complicated. It may be accelerated by the data center but also improve reliability, relieve congestion, support future customers, or enable new generation. A utility may oversize equipment because it anticipates additional development. A supply resource may be procured in response to several proposed loads rather than one project. A workable framework should distinguish three classes of cost. Dedicated costs arise from facilities that primarily serve one customer and have limited value without that customer. These costs should generally be assigned directly, subject to normal prudence review. Triggered shared costs arise when a project accelerates or enlarges an investment that also benefits other users. Allocation should reflect both the project’s causal role and the measurable system benefits available to others. Portfolio or anticipatory costs arise from infrastructure built for broader load growth, multiple prospective customers, reliability needs, or economic-development expectations. These costs require system-level planning evidence rather than automatic assignment to the first customer connected. Cost causation can be distorted in both directions. Assigning every shared upgrade to a single large customer can suppress economically valuable development and allow later users to free-ride. Socializing project-specific costs can force existing customers to finance speculative private infrastructure. The correct answer is not always “the data center pays everything.” It is that costs should be classified transparently, assigned according to causation and benefit, and revisited when actual development differs materially from the assumptions used to justify the investment. ## Obligations must be reciprocal Customer security is only one side of the commitment sequence. A customer cannot rationally guarantee a project schedule when the utility cannot substantiate a service date, cost range, or required scope of work. A framework that protects ratepayers by transferring nearly all uncertainty to the customer may discourage credible projects while doing little to improve utility performance. Matched customer obligations should therefore be accompanied by reciprocal utility obligations. Before requiring substantial security, the utility should disclose the basis for its cost estimate, the principal schedule assumptions, the work that will be undertaken, the conditions that could change the estimate, and the portion of the customer’s security that becomes nonrefundable at each stage. Utilities should provide periodic evidence of procurement, engineering, permitting, construction, and interconnection progress. Customers should have defined release or adjustment rights when the utility misses specified milestones for reasons not attributable to the customer. Cost ranges should narrow as studies advance. Service commitments should strengthen as the customer’s financial commitments increase. Material changes in scope should trigger review rather than automatic transfer of additional costs. Reciprocity does not mean that utilities must guarantee outcomes beyond their control. It means that both parties should be required to produce evidence at the point where the other is being asked to take risk. ## Flexibility is a service attribute, not an adjective Flexible and non-firm service may reduce the infrastructure required to connect some large loads. But flexibility has value only when it can be incorporated into planning and operations. A credible flexibility envelope must specify: - the amount of load that can be reduced; - the notice required; - the duration and frequency of curtailment; - the speed of response; - telemetry and control requirements; - testing procedures; - restoration conditions; and - consequences for nonperformance. Recent modeling indicates that standardized flexibility arrangements could expand the range of feasible data-center locations and reduce the need for some transmission reinforcements. The research is promising, but it is based on modeled systems and assumed operating envelopes, not a nationwide record of demonstrated performance.\[8\] A statement that computing workloads are theoretically movable is not equivalent to a dispatchable contractual obligation. The data center may have latency requirements, customer-service commitments, cybersecurity restrictions, backup-power limitations, or internal operating constraints that prevent it from curtailing when the grid needs relief. Flexible service should therefore receive differentiated treatment only when the operating obligation is measurable, enforceable, and reflected in the infrastructure plan. ## The federal process is moving toward enforceable rules The Ratepayer Protection Pledge provides political direction, but implementation will occur through regulatory proceedings, tariffs, utility contracts, and service agreements. On June 18, 2026, the Federal Energy Regulatory Commission issued tailored show-cause orders under section 206 of the Federal Power Act to PJM, MISO, SPP, CAISO, ISO New England, and NYISO. FERC directed the six regional operators and their transmission owners to justify existing tariffs or propose reforms addressing study processes, cost shifting, transparency, co-location, flexible service, and generation located near large loads. The orders also required reports on how adequate generation would be secured.\[9\] The proceedings are significant, but they are not final national rules. FERC expressly recognized that the six regions differ in market structure, geography, operating conditions, and progress on large-load integration. That is appropriate. A single national tariff design would struggle to accommodate vertically integrated utility systems, organized wholesale markets, constrained urban networks, generation-rich regions, water-limited areas, and locations with different infrastructure-development timelines. The durable federal contribution should be a common set of principles and evidence requirements—not an assumption that every system must use the same instrument. ## Environmental impacts also follow the project The same shift from generic claims to project evidence should govern environmental review. Data-center water use, emissions, land requirements, and cooling demand depend on location, grid composition, equipment efficiency, cooling architecture, climate, water source, and operating profile. National averages can indicate scale, but they cannot establish the impact of a particular campus. Design choices matter. A 2025 life-cycle assessment found that advanced cooling configurations reduced greenhouse-gas emissions by 15 to 21 percent, energy demand by 15 to 20 percent, and blue-water consumption by 31 to 52 percent relative to the study’s air-cooled baseline. Those findings are design-specific, not universal guarantees.\[10\] The relevant regulatory question is therefore not whether data centers are generically sustainable or unsustainable. It is what a proposed facility will consume, emit, discharge, and require at its location—and what enforceable design and operating commitments will govern those impacts. ## The emerging standard should be project evidence Before major utility and customer commitments become irreversible, a credible large-load review should require five forms of evidence. 1\. A verified load and milestone profile. The customer should disclose the requested capacity, expected ramp, load factor, operating characteristics, development schedule, site control, financing status, permitting progress, and internal approval milestones. 2\. A cost-allocation map. The utility should identify dedicated, triggered shared, and portfolio costs; explain the causal basis for each classification; disclose expected broader benefits; and specify how allocations will change if the project is delayed, downsized, expanded, or canceled. 3\. An enforceable flexibility envelope. Any claim of flexible, interruptible, or non-firm operation should define quantity, notice, duration, telemetry, testing, restoration, and nonperformance consequences. 4\. A location-specific environmental and community account. The project should disclose expected energy, water, emissions, land, backup-generation, and infrastructure impacts, together with design commitments, mitigation measures, workforce provisions, and applicable community benefits. 5\. A performance and exit plan. The governing agreements should specify what each party owes at every stage, how costs will be mitigated or reassigned, what happens when schedules change, and how stranded or underused investments will be recovered. These disclosures would not eliminate uncertainty. They would make uncertainty governable. They would also allow regulators to distinguish credible projects from speculative capacity claims without imposing a categorical barrier on large-load development. ## Ratepayer protection begins before energization The White House pledge states the right first principle: existing customers should not be required to subsidize the incremental infrastructure needed for private data-center development. But the headline size of a guarantee will not determine whether that principle works. Protection depends on the sequence of commitments, the risks created at each stage, the classification of shared costs, the credibility of project evidence, and the obligations imposed on both customer and utility. The prompt is a useful unit for measuring computational efficiency. It is the wrong unit for governing infrastructure. The grid plans around campuses. Utilities commit capital against forecasts and contracts. Equipment is ordered years in advance. Generation and network capacity may be reserved before the customer makes a final investment decision. When expectations fail, the resulting cost is physical, contractual, and local. That is where AI’s public consequences are determined. Not when the prompt is entered, but when the commitment becomes irreversible. ## Sign up for AIxEnergy Journal The AIxEnergy Journal provides executive intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, reliability, investment, and policy. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## Notes and Sources 1. The White House, “President Trump’s Ratepayer Protection Pledge Secures American AI Dominance, Protects Consumers,” July 23, 2026; and The White House, “Fact Sheet: President Donald J. Trump Advances Energy Affordability with the Ratepayer Protection Pledge,” March 4, 2026. 2. Alexandra Sasha Luccioni, Yacine Jernite, and Emma Strubell, “Power Hungry Processing: Watts Driving the Cost of AI Deployment?,” *Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency* (2024): 85–99, doi:10.1145/3630106.3658542; Felipe Oviedo et al., “Energy Use of AI Inference: Efficiency Pathways and Test-Time Compute,” *Joule* 10 (2026): 102430. 3. International Energy Agency, *Energy and AI* (Paris: IEA, 2025), “Energy Demand from AI.” 4. Sarah Josephine Smith et al., *United States Data Center Energy Usage Report: 2025 Update* (Berkeley, CA: Lawrence Berkeley National Laboratory, June 2026). 5. Electric Power Research Institute, *Powering Intelligence: Updated U.S. Data Center Scenarios* (Palo Alto, CA: EPRI, 2026). 6. U.S. Energy Information Administration, “Commercial Electricity Sales Have Soared in Virginia, Driven by Data Centers,” May 5, 2026. 7. Kavya Balaraman, “U.S. Power Companies Scramble to Secure Equipment as Surging Data Center Demand Strains Supplies,” Reuters, July 9, 2026. 8. Dongjoo Kim, Lin Dong, and Le Xie, “Flexibility-Aware Framework for Efficient Planner-Initiated Siting of Data Center,” *Nature Communications* 17 (2026): article 6512, doi:10.1038/s41467-026-72324-9. 9. Federal Energy Regulatory Commission, “FERC Launches Aggressive Targeted Action to Speed Large Load Integration,” June 18, 2026; dockets EL26-67-000 through EL26-72-000. 10. Husam Alissa et al., “Using Life Cycle Assessment to Drive Innovation for Sustainable Cool Clouds,” *Nature* 641 (2025): 331–38, doi:10.1038/s41586-025-08832-3. ### The Missing Middle of AI Infrastructure URL: https://www.aixenergy.io/the-missing-middle-of-ai-infrastructure/ Last updated: 2026-09-27T14:43:14.000Z The map of American data center development is being redrawn in public hearings, zoning rooms, utility dockets, and state legislatures. As of June 2026, fourteen states had considered or were considering moratoria on data centers. In the first quarter of the year, Data Center Watch counted at least 75 U.S. projects, representing roughly $130 billion in proposed investment, that were blocked or delayed amid local opposition. Its count of active opposition groups had spread to 49 states.1, 2 0:00 /0:33 1× Rich Deming makes a case for 30–50 MW data centers built around flexibility—and the rules regulators would need to change to make them work. Reducing that backlash to NIMBYism misses the harder infrastructure problem underneath. Communities are being asked to absorb facilities whose scale, power requirements, water demands, backup systems, tax arrangements, and land-use footprint can alter the trajectory of a place. A town that hesitates before accepting a 500 MW or 1 GW campus is not necessarily rejecting digital infrastructure. It may be rejecting the terms on which that infrastructure has arrived. That raises a question the data center debate rarely asks: Is there a viable middle scale between small edge facilities and the campuses that now dominate headlines? Consider a 30–50 MW data center. That range is not a proven industry optimum, and it should not be treated as one. A facility of that size is still a major electric load. On a small utility system, it can be enormous. At the wrong node, it can require the same kinds of upstream upgrades that delay much larger projects. But at selected industrial sites with real distribution or subtransmission headroom, it may offer a different development path: large enough to support serious compute, small enough to be replicated across locations, and potentially flexible enough to fit the grid rather than forcing the grid to be rebuilt around it. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/ChatGPT-Image-Jul-12--2026--07_01_30-AM.png) ## The real problem A data center and its utility approach reliability from different directions. The utility plans generation and network capacity to meet the customer’s contracted service obligation. The data center operator designs site infrastructure so that an interruption upstream does not interrupt critical compute. Those are not duplicate systems performing the same job. One protects system adequacy and delivery; the other protects continuity at the facility. That distinction matters because the strongest version of the community-scale argument is not that one of those systems is unnecessary. It is that the grid often treats the facility as a fixed, inflexible load even when the facility has multiple ways to reduce its net withdrawal during stressed hours. Those capabilities may include workload shifting, batteries, thermal storage, fuel cells, or on-site generators. Unless those capabilities are measurable, contractually committed, permitted for the intended duty cycle, and visible to planners, they cannot be relied on as system resources. The same caution applies to data center resilience standards. Uptime Institute’s Tier framework defines performance and topology requirements, but it is technology-neutral; it does not prescribe a single universal fuel or generator configuration. Site operators also make resilience choices based on customer contracts, insurance, workload criticality, fuel logistics, and local regulation. The useful policy question is therefore not, “Why are data centers forced to buy two power systems?” It is, “Which parts of a facility’s flexibility and resilience stack can be converted into dependable grid value without compromising the function they were built to serve?”3 ## What the flexibility evidence shows The case for asking that question is getting stronger. The Nicholas Institute at Duke University examined the 22 largest U.S. balancing authorities, which together represent about 95 percent of national peak load. Its 2025 study estimated 98 GW of “curtailment-enabled headroom” in a scenario limiting average annual load curtailment to 0.5 percent. In that scenario, the average balancing authority experienced some curtailment in 177 hours of the year, but most of those hours retained a substantial share of the new load, and average event duration was about 2.1 hours.4 That result is important, but its limits are equally important. The study was a first-order estimate. It did not fully model transmission constraints or every intertemporal operating constraint on the generation fleet. It does not mean that 98 GW of load can be dropped anywhere on the map without new wires or generation. It means that the power system’s extreme peaks leave a material amount of underused capacity that flexible loads may be able to access, depending on where they connect and how reliably they can respond. The Nicholas Institute’s 2026 follow-up modeled cumulative system-cost savings of roughly $40 billion to $150 billion over the next decade under its flexibility scenarios. A separate Johns Hopkins analysis of PJM found non-firm data center service to be the largest cost lever in its scenarios, reducing total annual system costs by roughly $15–$16 billion. Both are modeled results, not observed savings. But they point in the same direction: the contractual shape of new load can matter almost as much as the magnitude of the load itself.5,6 Commercial arrangements are beginning to test that proposition. In March 2026, Google disclosed that it had integrated 1 GW of data center demand-response capacity into long-term contracts with utility partners. The company describes the capability as shifting or limiting portions of machine-learning workloads and lists agreements with Indiana Michigan Power, Tennessee Valley Authority, Entergy Arkansas, Minnesota Power, and DTE Energy. The mechanism is demand-side flexibility. It is not a fleet of backup generators selling power back to the grid.7 EPRI is trying to make such capability easier to define and plan around. Its Flex MOSAIC framework describes large-load flexibility through performance attributes such as notification time, duration, frequency of use, depth of load adjustment, ramp behavior, and availability. The program also now has a dedicated workstream on distribution-connected data centers, explicitly examining their technical, operational, interconnection, economic, and reliability requirements.8,9 This is where the community-scale concept becomes interesting. The research and early commercial deals prove pieces of the model. ## Three different grid transactions The industry often uses “flexibility” as though it were a single product. It is more useful to separate three transactions. **• First, reduce grid withdrawal.** A data center can: - Shift a training run - Defer batch inference - Move computation to another region - Discharge a battery - Temporarily reduce noncritical facility load From the grid’s perspective, the key fact is the reduction in net demand. This is the domain of demand response, interruptible service, and non-firm service contracts. • **Second, serve the facility behind the meter.** A site can use storage or on-site generation to reduce what it draws from the utility during defined conditions. But a backup generator permitted and configured for emergencies is not automatically an economic peaker. Several factors matter: - Engine classification - Emissions permits - Fuel arrangements - Protection systems - Tariff provisions In 2025, the U.S. Environmental Protection Agency clarified that certain emergency stationary engines may operate for up to 50 hours per year in specified non-emergency grid-reliability circumstances when defined dispatch and recordkeeping conditions are met. Duke Energy’s PowerShare Mandatory 50 program is the agency’s concrete example. That pathway is narrower than routine peak shaving, but it proves that the regulatory boundary is not simply “outage or nothing.”¹⁰ **• Third, export electricity to the grid.** This is a separate engineering and market transaction. A resource that supplies a facility behind the meter is not automatically authorized or technically configured to push power onto a utility feeder. Export may require: - A generator interconnection study - Protection coordination - Synchronization equipment - Anti-islanding controls - Telemetry - Metering - Dispatch rights - Settlement rules - A compensation mechanism Depending on the arrangement, wholesale and retail jurisdiction may also divide responsibilities. The governing distinction is simple: reducing consumption, self-supplying behind the meter, and exporting electricity are related capabilities, but they are not interchangeable. A credible community-scale model has to decide which of those three products it is selling. The strongest design may combine all three, but the contracts cannot treat them as interchangeable. ## The rules are moving Federal regulators have opened part of the door. On June 18, 2026, the Federal Energy Regulatory Commission issued tailored Section 206 show-cause orders to all six RTOs and ISOs under its jurisdiction. The Commission required the grid operators to justify their existing tariff rules or propose reforms. Among the five categories FERC identified were new transmission services for flexible large loads, treatment of co-location and behind-the-meter generation, and processes for studying generation that serves electrically proximate large loads. The action builds on FERC’s December 2025 PJM co-location order and its January 2026 approval of Southwest Power Pool’s High Impact Large Load initiative.11,12 That federal movement does not solve the community-scale problem by itself. A 30–50 MW facility may be connected to distribution or subtransmission facilities and depend heavily on state retail tariffs, utility service regulations, distribution studies, local siting approvals, and state and local air permits. The federal and state questions are therefore complementary. FERC can shape transmission service and wholesale-market treatment. State commissions and utilities still have to design workable retail service and distribution-level pathways. ## The North Carolina test case North Carolina shows how quickly the policy pieces are moving. The governor’s Energy Policy Task Force recommended work on large-load tariffs, bring-your-own-capacity options, load flexibility, and interconnection reform. The report also documents that Duke Energy Carolinas and Duke Energy Progress already have high-load-factor tariffs available from 1 MW, with one-year terms and a minimum billing demand of 75 percent of contract capacity.13 Those existing schedules are not the end of the story. In testimony filed in late June 2026, Duke proposed a new large-load structure for customers of at least 50 MW. According to Canary Media’s review of the filing, the proposal would require a minimum payment of 75 percent and contract terms of 10 or 15 years. The commission could address the proposal in the pending rate case or open a separate docket.14 The next question for North Carolina—and for other states confronting similar growth—is whether customer protection and load flexibility can be designed together. A tariff built only around take-or-pay obligations protects existing customers from speculative load forecasts and stranded infrastructure. That matters. But a tariff that also defines a firm tranche and a measurable flexible tranche could give the utility a resource it can plan around rather than merely a customer it must build around. ## What a community-scale pilot should test The case for a community-scale data center is not strong enough to justify a nationwide template. It is strong enough to justify controlled pilots. A serious pilot would need to test the entire service architecture, not simply offer a discounted interruptible rate. • **Siting discipline.** Projects should be limited to nodes where utility studies show real headroom. Smaller is not automatically easier; a 40 MW load can still overwhelm the wrong feeder, substation, or local supply portfolio. • **A firm and flexible load split.** The service agreement should define what portion of load the utility must serve under all normal conditions and what portion can be curtailed, for how long, with what notice, and how often. Those parameters should be based on system need and facility capability, not a universal percentage copied from another jurisdiction. • **Measurement and enforcement.** Telemetry, automated control, periodic testing, event-performance measurement, and consequences for nonperformance are necessary if planners are going to rely on the flexible tranche. • **A resource-specific resilience plan.** Batteries, fuel cells, engines, thermal storage, and workload shifting have different duration, response, emissions, and fuel-security characteristics. The tariff should value delivered performance while permits and interconnection requirements remain specific to the technology. • **Export as an optional separate service.** Where local network studies show value, a project could test utility-controlled export under a distinct interconnection and compensation agreement. Export should not be assumed to follow automatically from behind-the-meter capability. • **Ratepayer protection.** Minimum bills, credit requirements, exit fees, ramp schedules, customer-funded upgrades, and transparent cost allocation remain necessary. Flexibility is not a substitute for cost causation. • **Community conditions.** Noise, local air emissions, water use, land use, and emergency fuel logistics should be evaluated directly. A smaller facility can reduce some impacts, but “community-scale” should describe a contract with the place, not merely a megawatt threshold. Community-scale data center should not yet be sold as the single answer to the data center backlash. There is no conclusive evidence yet that 30–50 MW is a universal economic sweet spot, that small utilities can absorb such loads without major upgrades, or that existing backup fleets can simply be converted into revenue-producing grid resources. Those claims require site-specific engineering, tariff analysis, permits, and project economics. However, there is enough evidence to support a more consequential proposition. Flexible large loads can unlock meaningful grid headroom. Non-firm service can reduce modeled system costs. Major operators are contracting for demand response at scale. Regulators are rewriting transmission rules. EPRI is developing a common vocabulary for flexibility and studying distribution-connected data centers as a distinct class. EPA has identified a limited pathway for qualifying backup engines to support local reliability. The pieces are emerging. What does not yet exist at scale is the integrated package: a modestly sized data center at a location with real grid headroom; a retail tariff that distinguishes firm from flexible demand; a facility designed to deliver that flexibility; a legally usable behind-the-meter resource stack; and, where it is technically valuable, a separate pathway for controlled export. That is the experiment worth running. The next useful data center may not be the largest one a developer can finance or the smallest one a community will tolerate. It may be the one whose contract tells the truth about what the grid must serve, when the facility can move, and what the site can reliably give back. ## Sign up for AIxEnergy AIxEnergy provides independent intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, reliability, investment, and policy. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## Notes [1\. Mathilda Scott, “Updates on the Cloud: More Moratoriums on Data Centers,” Rockefeller Institute of Government, June 10, 2026.](https://www.rockinst.org/blog/updates-on-the-cloud-more-moratoriums-on-data-centers/?ref=aixenergy.io) [2\. Data Center Watch, “Q1 2026: Data Center Watch Report,” 2026.](https://www.datacenterwatch.org/q1-2026?ref=aixenergy.io) [3\. Uptime Institute, “Tier Classification System,” accessed July 8, 2026.](https://uptimeinstitute.com/tiers?ref=aixenergy.io) [4\. Tyler H. Norris, Timothy Profeta, Dalia Patino-Echeverri, and Adam Cowie-Haskell, Rethinking Load Growth: Assessing the Potential for Integration of Large Flexible Loads in US Power Systems, NI R 25-01 (Durham, NC: Nicholas Institute for Energy, Environment & Sustainability, Duke University, 2025).](https://nicholasinstitute.duke.edu/publications/rethinking-load-growth?ref=aixenergy.io) [5\. Martin T. Ross and Jackson Ewing, Data Centers and Generation Capacity over the Next Decade: Potential Benefits of Flexibility, NI 26-04 (Durham, NC: Nicholas Institute for Energy, Environment & Sustainability, Duke University, 2026).](https://nicholasinstitute.duke.edu/publications/data-centers-and-generation-capacity-over-next-decade?ref=aixenergy.io) [6\. Ralph O’Connor Sustainable Energy Institute, Johns Hopkins University, Data Center Flexibility and Grid Infrastructure Needs in PJM, March 2026.](https://energyinstitute.jhu.edu/wp-content/uploads/2026/03/Data%5FCenter%5FAnalysis%5FPJM-March-20-2026.pdf?ref=aixenergy.io) [7\. Michael Terrell, “A New Milestone for Smart, Affordable Electricity Growth,” Google, March 19, 2026.](https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/demand-response-data-center-milestone/?ref=aixenergy.io) [8\. Electric Power Research Institute, “Flex MOSAIC,” DCFlex, 2026.](https://dcflex.epri.com/flex-mosaic?ref=aixenergy.io) [9\. Electric Power Research Institute, “Workstream 5: Distribution-Connected Data Centers,” DCFlex, 2026.](https://dcflex.epri.com/workstream-5?ref=aixenergy.io) [10\. U.S. Environmental Protection Agency, “Fact Sheet and Frequently Asked Questions: Use of Backup Generators to Maintain the Reliability of the Electric Grid,” May 2025.](https://www.epa.gov/system/files/documents/2025-05/rice-memo-on-duke-energy-regulatory-interpretation-04%5F17%5F25.pdf?ref=aixenergy.io) [11\. Federal Energy Regulatory Commission, “FERC Launches Aggressive Targeted Action to Speed Large Load Integration,” June 18, 2026.](https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration?ref=aixenergy.io) [12\. Federal Energy Regulatory Commission, “FERC to Act on Large Load Interconnection Docket by June 2026,” April 16, 2026.](https://www.ferc.gov/news-events/news/ferc-act-large-load-interconnection-docket-june-2026?ref=aixenergy.io) [13\. North Carolina Energy Policy Task Force, Interim Report, February 2026.](https://governor.nc.gov/documents/files/nc-energy-policy-task-force-2026-report/open?ref=aixenergy.io) [14\. Elizabeth Ouzts, “Duke Energy Proposes Special Rules for Data Centers in North Carolina,” Canary Media, July 6, 2026.](https://www.canarymedia.com/articles/data-centers/duke-energy-proposes-special-rules-for-data-centers-in-north-carolina?ref=aixenergy.io) ## Bibliography [Data Center Watch, “Q1 2026: Data Center Watch Report,” 2026.](https://www.datacenterwatch.org/q1-2026?ref=aixenergy.io) [Electric Power Research Institute, “Flex MOSAIC,” DCFlex, 2026.](https://dcflex.epri.com/flex-mosaic?ref=aixenergy.io) [Electric Power Research Institute, “Workstream 5: Distribution-Connected Data Centers,” DCFlex, 2026.](https://dcflex.epri.com/workstream-5?ref=aixenergy.io) [Elizabeth Ouzts, “Duke Energy Proposes Special Rules for Data Centers in North Carolina,” Canary Media, July 6, 2026.](https://www.canarymedia.com/articles/data-centers/duke-energy-proposes-special-rules-for-data-centers-in-north-carolina?ref=aixenergy.io) [Federal Energy Regulatory Commission, “FERC Launches Aggressive Targeted Action to Speed Large Load Integration,” June 18, 2026.](https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration?ref=aixenergy.io) [Federal Energy Regulatory Commission, “FERC to Act on Large Load Interconnection Docket by June 2026,” April 16, 2026.](https://www.ferc.gov/news-events/news/ferc-act-large-load-interconnection-docket-june-2026?ref=aixenergy.io) [Martin T. Ross and Jackson Ewing, Data Centers and Generation Capacity over the Next Decade: Potential Benefits of Flexibility, NI 26-04 (Durham, NC: Nicholas Institute for Energy, Environment & Sustainability, Duke University, 2026).](https://nicholasinstitute.duke.edu/publications/data-centers-and-generation-capacity-over-next-decade?ref=aixenergy.io) [Mathilda Scott, “Updates on the Cloud: More Moratoriums on Data Centers,” Rockefeller Institute of Government, June 10, 2026.](https://www.rockinst.org/blog/updates-on-the-cloud-more-moratoriums-on-data-centers/?ref=aixenergy.io) [Michael Terrell, “A New Milestone for Smart, Affordable Electricity Growth,” Google, March 19, 2026.](https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/demand-response-data-center-milestone/?ref=aixenergy.io) [North Carolina Energy Policy Task Force, Interim Report, February 2026.](https://governor.nc.gov/documents/files/nc-energy-policy-task-force-2026-report/open?ref=aixenergy.io) [Ralph O’Connor Sustainable Energy Institute, Johns Hopkins University, Data Center Flexibility and Grid Infrastructure Needs in PJM, March 2026.](https://energyinstitute.jhu.edu/wp-content/uploads/2026/03/Data%5FCenter%5FAnalysis%5FPJM-March-20-2026.pdf?ref=aixenergy.io) [Tyler H. Norris, Timothy Profeta, Dalia Patino-Echeverri, and Adam Cowie-Haskell, Rethinking Load Growth: Assessing the Potential for Integration of Large Flexible Loads in US Power Systems, NI R 25-01 (Durham, NC: Nicholas Institute for Energy, Environment & Sustainability, Duke University, 2025).](https://nicholasinstitute.duke.edu/publications/rethinking-load-growth?ref=aixenergy.io) [U.S. Environmental Protection Agency, “Fact Sheet and Frequently Asked Questions: Use of Backup Generators to Maintain the Reliability of the Electric Grid,” May 2025.](https://www.epa.gov/system/files/documents/2025-05/rice-memo-on-duke-energy-regulatory-interpretation-04%5F17%5F25.pdf?ref=aixenergy.io) [Uptime Institute, “Tier Classification System,” accessed July 8, 2026.](https://uptimeinstitute.com/tiers?ref=aixenergy.io) ### The Power Sector's AI Decade - Part 5 URL: https://www.aixenergy.io/the-power-sectors-ai-decade-part-5/ Last updated: 2026-07-26T02:15:47.000Z *This is the fourth of a six-part series at* [*AIxEnergy*](https://www.aixenergy.io/) *and mirrored at Michael Leifman's* [*Substack*](https://michaelleifman.substack.com/) [*Part 1*](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fi/) *introduced three distinctions (vendor capability versus utility adoption, reliability versus safety, task versus role) and a twelve-row sample of the US electricity workforce.* [*Part 2*](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fii/) *developed the four-stage mechanism by which AI capability translates into workforce change in regulated industries and traced it through aviation, freight rail, and maritime.* [*Part 3*](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fiii/) *maps how AI is reshaping the U.S. electric-power workforce role by role, showing that task content is already shifting across the industry while headcount changes will unfold unevenly depending on whether work is routine, reliability-critical, safety-critical, or institutionally protected*. [*Part 4*](https://www.aixenergy.io/the-power-sectors-ai-decade/) *explained that AI and robotics are not only replacing existing power-sector tasks, but also creating new capabilities by letting the grid coordinate more complexity than human operators can manage and sending machines into hazardous places where human bodies should not go.* *This part explains how utilities, regulators, and investors can manage AI adoption by building the workforce, governance, and institutional capacity needed to sustain automation, protect reliability, and avoid hollowing out critical operational capability.* Imagine, if you will, an ordinary substation yard on an ordinary morning, somewhere on the American grid, in the year 2026\. No one is standing in it. A drone lifts from a weatherproof dock, flies its route over the switches and transformers, and settles back onto its perch, and the nearest human is a technician at a desk in another town, watching the images arrive. A four-legged robot the crew has given a friendly name walks the same yard on rounds of its own, reading gauges no person now reads in person. A few counties over, the last of the meter readers have all but vanished from the employment tables, their route work not outsourced and not struck down, simply worn away from within. On a server rack somewhere, a forecasting model that actually worked has been quietly switched off, because the company that built it could not keep on staff the one or two people who knew how to keep it alive. And on a warm evening that past July, in tens of thousands of garages, home batteries woke at the same instant and discharged together, and for two hours behaved like a single power plant that nobody built. Article Summary 0:00 /29.100408 1× None of that is a forecast. Every piece of it has already happened, and the earlier parts of this series gave each one its name, its date, and its source. The reason to gather them into a single morning is to be clear that this is the ordinary present, not the frontier. Now imagine, if you will, that same yard fourteen years on, in 2040, and imagine you are allowed to see it twice. In the first version it is three in the morning and a storm is tearing across the system. In a control room a hundred miles off, a woman watches it come. The job she is doing tonight had no name when she started. She came up through the customer lines, back when there were customer lines staffed by humans, and when that work began to disappear someone saw it going and pointed her toward the work that was arriving, and now she sits above a grid stitched from a million rooftops and batteries and parked cars, reading what the machines surface and deciding what they are not permitted to decide. A conductor comes down on a rural feeder. Before she has finished registering it, the system has already found the fault, self-healed and walled it off, and pulled power around it through three other paths, and the customers on that feeder see their lights blink once and steady. The whole event took just five minutes, and not because anyone hurried, but because nothing waited on a person to notice. In the same area, there is a lineman who got an automatic alert on his phone at the same moment. The map is already on it, the exact span, the likely cause, the section already dead and grounded by the same system that rerouted around it. He drives to the pole, and not to the three wrong poles he might have chased in an earlier decade, and he does the work his hands have always done, the work no machine on that truck can do, on a line made safe before he was on the road. The repair takes him most of the night. No one is in the dark while he does it. Now imagine the same yard, the same storm, but in the other 2040\. The room is just as quiet and just as thinly staffed. The woman who might have been steadying it is not there, because the door she would have come through was never cut; the jobs at the bottom ended and nothing was built to carry anyone up from them. The conductor comes down on the same rural feeder, and the system that should wall it off does something wrong instead, because it is newer and cleverer than anything we run today and no one in the building ever learned it deep enough to know the shape of its mistakes. The power does not reroute. It goes out, and it stays out, and the two operators on shift reach for the phone to call the vendor. And the lineman, a good one, gets a bad ping, a rough location and nothing isolated, so he drives to where the system thinks the fault is and finds nothing, and then works it the slow old way in the rain, with the feeder live and dangerous around him and the customers still dark. What was five minutes in the other 2040 is most of a day here. The report, when it comes, will call the night unforeseeable, and that will not be true. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/ChatGPT-Image-Jul-21--2026--10_37_56-PM.png) Neither of those futures is science fiction, and neither is collapse. Both grids run on more automation than we operate now, and both carry fewer people than 2026 does; the previous installments in this series traced that contraction, and by 2040 it has run its course in either version. The line crew is about the same size in both, because a pair of hands on an energized conductor is still the last thing to go. That much is not the variable. The variable is everything that stood, or failed to stand, in that room at three in the morning. And the uncomfortable truth about the two 2040s is that the second one is what arrives on its own. It is where the grid settles when each company and each regulator does the reasonable thing in front of it: buying a capable system without hiring the smaller number of people who can keep it running, approving a deployment faster than anyone on staff can learn to oversee it, letting the entry-level jobs disappear without opening new ways in. The better grid is the one that had to be built on purpose, against that near-term reasoning, by people willing to spend now for a payoff they would not see for years. That is why what follows is written as strategy rather than forecast, and why it goes actor by actor, because each of them holds one of those choices. ## Utilities 💡 AI workforce strategy must differ by utility type, but every utility ultimately succeeds or fails on whether it builds the specialized people, career pathways, labor plans, and institutional capacity needed to sustain deployed systems over time. Whether AI arrives in your workforce as a contained, one-time decision or as a decade-long institutional problem depends first on which kind of company you are. A generation-only merchant producer, a wires-only delivery company, and a vertically integrated utility that owns both face genuinely different versions of the same transition, and the differences are large enough that a single AI workforce strategy is not much use to any of them. ### Merchant producers: sustain what you deploy The merchant power producer has the most contained version of the transition. Its workforce is thin to begin with, and the roles AI touches first are the cognitive ones close to the market: the traders and scheduling coordinators who bid generation into day-ahead and real-time markets, whose forecasting and bid preparation are already being automated, and a small back office behind them. Plant operators are shrinking too, but mostly through fleet consolidation, not software. The roles that grow, in storage and power-electronics engineering and data analytics, are ones every part of the industry is trying to hire. With no field workforce and few licensing or staffing minimums to work through, the transition here is short and sharp. It reduces to one decision: build the optimization and trading layer that increasingly runs the business, or buy it. The build-or-buy call looks like procurement and is really about workforce. Buy, and you are betting that a vendor keeps the tool current and that a lean internal team can cover the rest. Build, and you have to recruit and hold the small group of data scientists and machine-learning engineers who keep a model running in production. A grid operator did the cautionary version: it switched off a forecasting model that worked because it had never put dedicated staff on keeping the model retrained and validated. A merchant producer is more exposed to that than the operator was, because when the software degrades there is no field operation underneath to absorb the failure. So the real decision is not build-or-buy; it is whether you will staff to sustain what you deploy. Deploy what you cannot sustain and you get the worst outcome available: a system that goes dark after you have already thinned the desk that used to do its job. ### Wires companies: the succession problem The wires company has the harder version, because it owns the workforce most people picture when they picture a utility: line crews, substation and relay technicians, inspectors, cable splicers, vegetation crews. This is skilled physical work in unpredictable conditions, and it stays with people longest. AI thins the cognitive scaffolding around it well before it touches the headcount. At the same time the company runs the roles that contract fastest, customer service and billing, where automated systems already handle the first line, and the one operational role clearly growing, the distribution operator managing a more complex and more distributed grid. So a single payroll holds both extremes: durable field roles whose work is quietly changing, and clerical roles disappearing quickly, on different timelines. For the wires company, the split becomes a succession problem stretched across a decade. The roles the company will need fewer of are held disproportionately by its most heavily unionized workers, who cannot simply be released. The roles it will need more of sit on pay scales that mostly do not exist inside a utility today and require talent it has never had to compete for. Plan from the current org chart and you end up overstaffed where work is contracting, unable to hire where it is growing, and you learn both at the same late moment. Managing that succession is a matter of timing, not headcount. Take the relay technician, who tests and maintains the protective relays that isolate faults on the system. The license, the liability exposure, and the contract all keep that position in place, and they keep it in place while more of the diagnostic work moves into digital relays and automated test equipment that surface the problem and record the result. The person still signs off, because a person must, but the sign-off is increasingly what is left. The role is defended while the work inside it has already moved, and the headcount looks steady right up until the day the redefined job no longer needs as many people to do it. So the moment to open the conversation with labor is now, before any reduction is visible, while there is still something to bargain over besides loss. Wait until the contraction shows up in the numbers and you negotiate from the weak position that has produced the worst transitions in other industries. Start early, over the shape of the changed work and the routes from shrinking roles into growing ones, and you keep room the late mover has already lost. ### Integrated utilities: plan the halves apart The vertically integrated utility has both problems at once, inside one company, one rate case, one workforce plan. Its generation side wants to move quickly on optimization and market-facing AI; its wires side runs on the slow, safety-bound timeline where change is measured in decades. A single blended plan gets both wrong, too aggressive on the durable field roles and too cautious on the contracting back office, because the average sits between two things that are nowhere near each other. The discipline is to plan the two workforces separately even under one roof: two timelines, two sets of growing and shrinking roles, two postures toward labor and toward build-or-buy, brought together only where capital is actually allocated. A real utility does not split cleanly down the middle, and no one should pretend it does. But the two halves are not really connected; averaging them just misreads both. Planning them apart refuses that misleading average; it does not pretend the company is two. ### The bench beneath all three Beneath all three actors is the same question: can the company keep the systems it deploys running once the demonstration is over and the pilot has closed? The forecasting model from earlier in this section is that question in miniature. Someone built a tool that worked, and it went dark not because it failed but because no one was ever dedicated to keeping it retrained and validated as conditions drifted. A demonstration proves a system can work once. Keeping it working means carrying the people who retrain it as conditions drift, catch it when it fails quietly, and validate it against a grid that never stops changing. That standing capacity is nearly invisible from outside and the easiest thing to defer, which is why it gets deferred until a working system goes dark for lack of it. Fund it before you need it. From outside, it is also the clearest signal of which companies can actually run what they buy. ## Regulators and Governments 💡 Regulators and governments determine the pace and unevenness of AI-driven workforce change through reliability, licensing, and safety decisions, so they must coordinate adoption standards, build technical oversight capacity, and move quickly on reliability without weakening safety protections. A state commission that shortens an interconnection study from months to days is trying to get power onto the grid faster, but a downstream effect is that it thins the work of the queue analysts and interconnection engineers. The same holds when a commission clears an operator to lean more heavily on automated dispatch: the aim is a faster and steadier grid, but a downstream effect is that the dispatchers and traders at the desk have less left to do. A regulator’s fastest effect on the workforce does not run through labor law; it runs through reliability decisions made for other reasons, and it arrives sooner than almost any other force in the sector. Each of those decisions is made jurisdiction by jurisdiction, so regulators and states do not stand outside the transition’s unevenness; they are its cause. A tool that is cheap and quick to field under one commission waits behind a longer study or a stricter standard under the next, so the same capability reaches workers on a different schedule depending only on whose territory they work in. FERC, the reliability organization NERC, the seven regional grid operators that run the wholesale markets (only six of them under FERC’s authority), and the state commissions build that patchwork between them, and the pace each sets becomes a condition the rest of the sector has to answer to. In December 2025, FERC found PJM’s tariff unjust and unreasonable and ordered it to write new rules for connecting data centers and other large loads to power plants ([ferc.gov](https://www.ferc.gov/news-events/news/fact-sheet-ferc-directs-nations-largest-grid-operator-create-new-rules-embrace?ref=aixenergy.io)), following in 2026 with show-cause orders to all six grid operators it regulates ([mcguirewoods.com](https://www.mcguirewoods.com/client-resources/alerts/2026/6/ferc-issues-section-206-show-cause-orders-directing-all-six-rtos-isos-to-justify-or-reform-large-load-integration-rules/?ref=aixenergy.io)). These orders are what most people mean by AI and the grid, and they are about load, not labor: how to connect enormous new demand without dumping the cost onto ratepayers. They touch the workforce question at only one point: the study processes they compress to move power faster are the same ones that gave queue analysts their work. ### The slower wave belongs to the safety agencies The line worker, the relay technician, and the reactor operator sit on a slower schedule, and a different set of bodies controls it. The National Labor Relations Board, OSHA, the Nuclear Regulatory Commission, and the state boards that license engineers and operators hold the defenses around the safety-critical roles, and they gate the second and longer wave of this transition, the one that reaches physical work only after the cognitive roles around it have thinned. Their timeline is set by licensing regimes and safety standards that move in years and decades, not in dockets. Where any of these agencies stands today will read as dated within a year, and it is beside the point. What lasts is that they, not the grid regulators, control the speed of the wave that has not yet crested, and how they set it decides whether it arrives gradually or all at once. ### Every commission’s choice about speed Each commission is left a real choice about speed. A commission that moves slowly, studying each deployment at length and approving with caution, does not hold still by doing so; it watches developers, large loads, and capital drift toward neighboring jurisdictions whose processes move faster, and then faces a hard catch-up on a timetable it no longer sets. A commission that moves fast draws that activity in but takes on the harder job of overseeing operations it approved more quickly than its capacity to supervise them grew. Either way the binding hire is the same. These agencies are automating their own analytical work even as they rule on everyone else’s, and the one role clearly growing inside them is the analyst who can judge whether a model is safe to approve and sound enough to keep running, the person a commission needs on staff before the deployments arrive rather than after the strain shows. Neither speed is free. ### A model standard, not a mandate Each commission’s mistake is to treat the pace of adoption as a private choice, because that is exactly what produces the patchwork. The seam between fast and slow jurisdictions is not something any single commission can close from inside its own borders, and it is that seam, more than any one commission’s pace, that decides how unevenly the transition lands on workers. Closing it is a coordination problem, and coordination has two levers here, split by the same federal-state line that runs through everything else. On the federal side, FERC can standardize the adoption pace where it already has authority, across the wholesale markets, transmission, and the interconnection processes it governs, so that the seam does not widen from region to region for no better reason than which operator moved first. What FERC cannot do is reach retail: the Federal Power Act draws a bright line that leaves end-use and distribution matters to the states, which is why the large-load orders themselves left retail cost protection to state commissions. And where FERC has pushed hardest on standardizing, the states have pushed back; NARUC called itself generally disappointed by the diminished state role in FERC’s Order 1920 on transmission planning ([renewableenergyworld.com](https://www.renewableenergyworld.com/power-grid/transmission/state-utility-commissions-are-not-happy-with-fercs-new-transmission-rules/?ref=aixenergy.io)). A standard imposed from above is the version states fight. The state-side lever is not a federal mandate but a model states can choose to take up, and NARUC and NASEO should build it. NARUC, the association of the state commissions, and NASEO, the association of the state energy offices, already build this kind of thing together, from joint working groups to shared tools for implementing FERC orders ([naruc.org](https://pubs.naruc.org/pub/C52FC932-1866-DAAC-99FB-29E2735E0C12?ref=aixenergy.io)). A model standard for the pace and oversight of AI adoption, developed by those two bodies with federal technical expertise of the sort the standing FERC-NARUC Federal-State Current Issues Collaborative already supplies ([ferc.gov](https://www.ferc.gov/news-events/news/ferc-naruc-establish-federal-state-current-issues-collaborative?ref=aixenergy.io)), would hand a commission in any state a template to adopt or leave. A model rule spreads by being useful, not by being ordered, and it is the only form of coordination that narrows the seam without trampling the state authority the rest of the system rests on. What it buys is a transition whose timing is chosen rather than left to accident, so the same worker is not put on a far faster timeline in one state than across the border for no reason anyone decided. ### Reliability can move fast; safety cannot Running fast on reliability is often right. Running fast on safety is not, and the two look enough alike that the pressure to treat them the same never lets up, because a neighboring jurisdiction will always seem to gain by easing a staffing minimum or a licensing standard. A commission that lets that competition speed its safety decisions to match its reliability decisions gives away the protection that separates a managed transition from the grid that went dark at three in the morning, the one trusted past the point anyone on shift understood it. Moving fast on reliability is strategy. Moving fast on safety is how you get the second grid. ## Investors and Capital Allocators 💡 Investors should look past AI announcements and fund utilities with the technical bench, workforce planning, governance, and operating discipline to sustain many practical deployments in production—not just launch a high-profile pilot. Two utilities announce the same thing in the same quarter: a serious AI program, real budget, named use cases, a timeline to production. From the outside the two announcements look nearly identical, and the market tends to price them the same way. Neither announcement says what separates the companies. Three years on, one utility is still running the systems it described and the other switched them off when the people who understood them left. The capacity that produces that difference is close to invisible from where an investor sits, and learning to read it is the whole task. ### Read the institutional signal, not the press release The market sorts these announcements on a rough axis, thoughtful against following the herd off a cliff, and the herd is easy to spot once you know what to look for. Herd signals are procurement theater: a press release, a pilot, a vendor logo, a humanoid-robot photo op. They are cheap to produce, and they are exactly what the consultancy productivity story rewards, because that story is pitched at this audience and it flatters buying over building. Thoughtful signals are institutional and harder to see from outside: real depth on the internal machine-learning and data-engineering bench, a workforce plan built around where the roles are going rather than where they sit, and governance that can carry a capability from pilot into operations without relitigating it at every desk. The institutional signals never appear in the announcement, and all of them decide whether it becomes anything. A utility with a thin technical team and no real workforce plan will announce an ambitious program and put little of it into operation, because it cannot maintain what it deploys and cannot move its people toward the work that is growing. Its AI program is mostly presentation. A utility that has built the bench and done the planning will say less and deliver more. The institutional weight that reads as slowness to someone reviewing operations reads as durability to someone allocating capital, and that inversion is what an investor can use. ### What the signal looks like in practice PJM built the clearest cautionary case. In a paper published in November 2024, PJM described a net-interchange forecasting model, built with the machine-learning method called XGBoost, that came out of a volunteer team carrying the work alongside their regular jobs ([pjm.com](https://www.pjm.com/-/media/DotCom/library/reports-notices/special-reports/2024/20241113-insights-from-pjm-interconnections-exploration-of-artificial-ntelligence.pdf?ref=aixenergy.io)). Net interchange is the net flow of power across PJM’s borders, and the model worked: testing showed it performing well, and PJM planned to put it into production. PJM terminated it anyway, because the team had no dedicated staff to evaluate and maintain it. What ended it was not a build-versus-buy mistake but the absence of anyone to keep it running, a constraint that binds whether a utility builds or buys. The missing staff that killed PJM’s model also, for now, protects the roles a working model would have thinned. A company that cannot deploy the system that would shrink a desk keeps that desk, not by choice but for lack of the capacity to change it. A thin bench can therefore look like a stable workforce and a low-risk operation when it is the opposite, the mark of a company that will neither capture the productivity it advertises nor manage the transition once a competitor forces it. Enel shows what the thoughtful signal looks like in practice. The Italian multinational treats machine learning as part of how it runs the network rather than as a launch, and points it mostly at unglamorous maintenance. Enel’s own account describes applying AI across plant and grid operation and maintenance ([enel.com](https://www.enel.com/learning-hub/innovation/artificial-intelligence?ref=aixenergy.io)); in practice that means sensors on turbines and lines feeding models that flag a failing part, a wearing gearbox bearing for instance, before it takes an asset down, so a crew goes out ahead of the outage rather than after ([aveva.com](https://www.aveva.com/en/our-industrial-life/type/article/why-predictive-maintenance-is-key-to-a-renewables-reliant-grid/?ref=aixenergy.io)). The value is in the accumulation, many small deployments kept running across a footprint that spans dozens of countries rather than one system unveiled and left to rot. Enel is not chasing a single transformational grid brain; it is sustaining many marginal tools, which is the harder thing and the more valuable one, and the thing the herd’s lone pilot cannot match. PJM came back to the same problem from the other direction. Through ordinary automation of its interconnection process and added staff, PJM cut its backlog sharply, by roughly 60 percent on its own account ([insidelines.pjm.com](https://insidelines.pjm.com/pjm-chooses-51-generation-resource-projects-to-address-near-term-electricity-demand-growth/?ref=aixenergy.io)), and layered its April 2025 partnership with Google’s Tapestry project, a self-described moonshot for the grid, on top of that work ([insidelines.pjm.com](https://insidelines.pjm.com/pjm-google-tapestry-join-forces-to-apply-ai-to-enhance-regional-planning-generation-interconnection/?ref=aixenergy.io)). An investor could read the moonshot as the transformational fix and stop there, but the unglamorous automation and staffing had done most of the work, and the partnership sits on capacity PJM built itself. A vendor partnership pays off only for an organization that can direct and integrate it, which is the bench question in another form. That order, capacity first and the moonshot on top of it, is the thing to look for. ### Two habits for allocating capital For anyone allocating capital in this sector, two habits matter most. Weigh the institutional barriers, not just the technical capability: the question is never whether a tool can work in a demonstration but whether this particular company can sustain it in production and move its workforce toward the roles the tool creates. And weigh both the marginal and the transformational, not just the end game. The compounding return tends to come from many maintained, unglamorous deployments rather than a single moonshot; PJM drew that lesson from its own experiments, concluding that smaller, productivity-focused tools can pay off with far less effort and risk and pave the way for the larger systems later ([pjm.com](https://www.pjm.com/-/media/DotCom/library/reports-notices/special-reports/2024/20241113-insights-from-pjm-interconnections-exploration-of-artificial-ntelligence.pdf?ref=aixenergy.io)). A company that can only point at the moonshot is usually the one that has skipped the marginal work, and the institutional work, underneath it. The transformation is real and it is coming, but it accrues to whoever built the capacity to run it, and that capacity is exactly what the announcement never shows. Utilities, regulators, and investors hold real leverage over how fast this transition arrives and how well it goes; each one decides, from some position of strength, the shape of a change that is coming either way. The people who will live inside that shape have far less say in it. What workers can do from the weaker side of the table, and what is owed to the ones no strategy reaches at all, is the other half of this story. ### The Federal AI Land Rush Is an Execution Test URL: https://www.aixenergy.io/the-federal-ai-land-rush-is-an-execution-test-not-a-scale-race/ Last updated: 2026-07-29T19:42:28.000Z *This analysis complements the* [*Payne Institute*](https://payneinstitute.mines.edu/?ref=aixenergy.io)*’s discussion of data center development on sovereign Tribal lands, available here:* [*https://payneinstitute.mines.edu/the-future-of-ai-runs-through-indian-country/*](https://payneinstitute.mines.edu/the-future-of-ai-runs-through-indian-country/?ref=aixenergy.io) In AI infrastructure, a gigawatt can describe several very different things: a long-range development target, capacity covered by a commercial agreement, a financed project phase, a facility under construction, or an operating electrical load. Those states are not interchangeable. That distinction has become central to the federal government's experiment with using public land for AI data centers. Since April 2025, the Department of Energy and the military services have opened federal energy sites and military installations to private data center development. The policy theory is straightforward. The federal government already controls large parcels of land, some near substantial energy infrastructure, and it can coordinate federal review in ways that a private developer assembling a conventional site cannot. But land control is only the beginning of the development chain. 0:00 /0:33 1× Most of the federal projects remain in solicitation, proposal evaluation, conditional selection, or lease negotiation. Portsmouth, Ohio, has advanced further in public view, but even there the distinction between a 10-gigawatt full-buildout plan and the first project phase matters. The real test of the federal-land strategy is therefore not how many nominal gigawatts can be attached to federal parcels. It is whether those parcels can move from announcement to firm power, environmental clearance, financing, construction, commissioning, and operation faster than comparable projects elsewhere. ### DOE Opened a Large Pipeline. Most of It Is Still a Pipeline. The Department of Energy began the current federal-land push on April 3, 2025, when it identified 16 potential sites for data center and energy infrastructure development. DOE said the sites offered advantages that could include existing energy infrastructure and opportunities to accelerate permitting for new generation. The agency set a target of commencing operations at selected sites by the end of 2027\. That was a program goal attached to a request for information, not a finding that the 16 sites were uniformly shovel-ready. AIxEnergy Project Tracker ## Federal AI infrastructure remains mostly a development pipeline Portsmouth is the most advanced visible federal-land project. Most other sites remain in solicitation, selection or lease negotiation. **Executive takeaway:** Headline gigawatts combine different project stages. The meaningful measure is conversion into firm power, financing, construction and operating load. | Project | Scale | Status | Commercial model | Next proof point | | --------------------------------------------------------------------------- | -------------------------------------------------- | ---------------------------------------------- | ----------------------------------------------------------------------- | ----------------------------------------------------------------------- | | Portsmouth AI campus DOE · Ohio · SoftBank and SB Energy | 10 GW 800 MW initial phase reported | Early development Groundbreaking completed | Private data center, generation and transmission investment | First-phase financing, transmission construction and commissioned load | | DOE solicitation portfolio Idaho, Oak Ridge, Paducah and Savannah River | Not disclosed | Solicitation | Privately funded development on DOE land | Developer selection, executed agreements and firm power arrangements | | Initial Air Force portfolio Five installations across TN, AZ, CA, NJ and GA | \>100 MW each Minimum qualifying size | Closed solicitation No public award identified | Commercial leasing with monetary consideration | Developer awards, executed leases and site-level power plans | | Alaska Air Force portfolio JBER, Eielson and Clear Space Force Station | 4,700 acres Capacity not disclosed | Solicitation | Enhanced-use leasing at fair-market value | Developer selection, generation plan and interconnection approval | | Fort Bliss US Army · Texas · Carlyle | 2.5–3 GW Secondary reporting | Conditional selection | Developer-financed facility with cash, in-kind or compute consideration | Final lease, environmental review, firm power and financing | | Dugway Proving Ground US Army · Utah · CyrusOne | \~1 GW Secondary reporting | Conditional selection | Developer-financed facility with potential government compute access | Final lease, environmental clearance, power and financing | | AI Arsenal Separate Defense Department compute program | $29.5B FY2027 request | Budget request | Proposed federal investment in data centers, GPUs and supercomputers | Appropriation, acquisition plan, site selection and contract awards | | Stargate Private-sector comparator · OpenAI and partners | \>10 GW secured $500B initial investment intention | Mixed portfolio | Private corporate and infrastructure investment | Site-level disclosure of construction, commissioning and operating load | **Analytical rule:** Announced, selected, financed, under construction and operating capacity are not equivalent. Solicitation **→** Selection **→** Lease **→** Power and permits **→** Financing **→** Construction **→** Operation In July 2025, DOE selected four locations for the next phase: Idaho National Laboratory, the Oak Ridge Reservation in Tennessee, the Paducah Gaseous Diffusion Plant site in Kentucky, and the Savannah River Site in South Carolina. DOE then issued site-specific solicitations between September and November 2025\. The public documents describe prospective, privately funded developments and, in the cases of Oak Ridge and Paducah, expressly place responsibility for utility interconnection agreements, regulatory requirements, and permitting plans on applicants. Portsmouth is the important exception to the otherwise early-stage DOE portfolio. In March 2026, DOE announced an arrangement under which SoftBank Group and SB Energy plan a 10-gigawatt data center development supported by 10 gigawatts of new generation, including at least 9.2 gigawatts of natural gas generation. The announcement also included a planned $4.2 billion transmission program involving SB Energy and AEP Ohio, $33.3 billion in announced Japanese funding for the 9.2-gigawatt gas-generation component, and a $40 million Community Benefits Agreement. Officials held a groundbreaking ceremony at the site on March 20\. DOE's March 24 account said development of the data center was expected to begin during 2026 with an initial phase on 189 acres. Local public-media reporting described that first phase as an 800-megawatt data center expected, according to a DOE official's estimate, to come online in about two years. The 10 GW number is therefore best understood as the campus's full-buildout plan, not the amount of capacity now operating or a measure of the first construction phase. This makes Portsmouth more important as a test. Its progress can now be measured against concrete development milestones: cleanup and site preparation, power-project financing, transmission construction, the first 800-megawatt phase, commissioning, and actual load. ### Military Land Is Following More Than One Business Model The Department of the Air Force has pursued a largely commercial real-estate model. In October 2025, it issued a request for lease proposals covering AI data center development at Arnold Air Force Base in Tennessee, Davis-Monthan Air Force Base in Arizona, Edwards Air Force Base in California, Joint Base McGuire-Dix-Lakehurst in New Jersey, and Robins Air Force Base in Georgia. The solicitation required qualifying projects to exceed 100 megawatts of new load and involve at least $500 million in capital expenditures; the Department of the Air Force sought monetary lease consideration from selected developers. The opportunity is now marked closed on the service's partnership page, but that page does not identify a project award. The Air Force subsequently expanded the concept to Alaska. After a February request for information, it issued a formal request for lease proposals on April 10, 2026, for approximately 4,700 acres across Joint Base Elmendorf-Richardson, Eielson Air Force Base, and Clear Space Force Station. The announced model is again commercial development through enhanced-use leasing, with the service receiving at least fair-market-value cash consideration. The Alaska opportunity should not be conflated with separate military microreactor initiatives. The data center solicitation itself is an opportunity to develop facilities with associated utility infrastructure and generation; the official materials reviewed for this article do not establish that nuclear microreactors have been selected as the power source for the proposed data centers. The Army's emerging model is different. On March 26, 2026, the Army conditionally selected Carlyle for a proposed development on about 1,384 acres at Fort Bliss, Texas, and CyrusOne for a project on approximately 1,201 acres at Dugway Proving Ground, Utah. The developers are to finance, build, operate, maintain, and eventually decommission the facilities. But the agreements are not final. The Army said the selections begin an exclusive negotiation period for leases and other terms. The Army projects initial operating capability at Fort Bliss in fiscal 2027 and at Dugway in fiscal 2029\. Those are schedules, not completed outcomes. The US Army Corps of Engineers is involved in lease negotiations, technical work, and environmental review. The arrangement is being pursued under 10 USC § 2667, which authorizes leases of qualifying non-excess military property and generally requires consideration of at least fair-market value, which may be paid in cash or in kind. Secondary reporting has described the Fort Bliss plan as a 2.5- to 3 GW development and the Dugway project as approximately 1 gigawatt. Those figures are useful indications of developer ambition, but they should not be treated as final leased or operating capacity while commercial terms remain under negotiation. Reporting also indicates that the Army expects to receive access to a portion of the resulting compute capacity, making the concept more than a conventional cash-rent real-estate transaction. ### The Pentagon's Own Compute Program Is a Separate Transaction The military leasing initiatives also need to be separated from the Pentagon's proposed AI Arsenal spending. The fiscal year 2027 budget justification includes a $29.5 billion mandatory funding request under “Next Generation Technology and Autonomy.” The justification says the funding would support the AI Arsenal initiative, moving from scattered graphics-processing-unit clusters toward a coordinated portfolio that includes hardened, sensitive compartmented information facility-accredited data centers at multiple sites and the procurement and installation of GPUs and AI supercomputers. This is a budget request, not an enacted appropriation. That program is conceptually different from opening federal land to commercial developers. One is a proposed government compute and data-center portfolio. The other uses federal real estate and lease authority to attract privately financed development, sometimes with government access or other consideration built into the transaction. ### Private Development Is Larger, but the Comparison Still Requires Stage Discipline The private AI infrastructure market is unquestionably enormous. But the same project-status discipline that should govern federal claims must also govern private-sector claims. OpenAI announced Stargate in January 2025 with a stated intention to invest $500 billion over four years and secure 10 gigawatts of AI infrastructure in the United States. By September 2025, OpenAI said its portfolio had reached nearly 7 gigawatts of planned capacity and more than $400 billion of investment over the following three years. On April 29, 2026, the company said it had surpassed its initial 10-gigawatt secured-capacity goal. Those disclosures show extraordinary scale. They do not justify treating every gigawatt in the portfolio as simultaneously operating or under construction. OpenAI itself describes a mixed portfolio in which some facilities are operating while other sites are being developed, expanded, secured, or evaluated. That is why adding Portsmouth's 10-gigawatt full-buildout target to a reported Fort Bliss capacity estimate and calling the result a federal commitment is not analytically sound. It combines different sponsors, financing models, project stages, power arrangements, and contractual statuses. Dollar comparisons create a similar problem. The federal-land programs are deliberately structured to mobilize private capital. They cannot be compared cleanly with hyperscaler corporate capital expenditure as though every dollar on one side were public spending and every dollar on the other side funded the same asset class. Morgan Stanley's roughly $2.9 trillion estimate through 2028, for example, is an estimate of global data-center construction cost, not projected spending by the five technology companies cited in the original draft. ### Federal Land Solves the Land Problem. The Harder Problems Remain. The strongest case for federal sites is not that the federal government can outspend Microsoft, Amazon, Alphabet, Meta, Oracle, OpenAI, SoftBank, or private infrastructure investors. It is that federal land can eliminate some costly steps at the beginning of a project. A federal owner can offer a large contiguous site through one process rather than forcing a developer to assemble dozens of parcels. Some energy sites have existing transmission corridors, substations, industrial infrastructure, security perimeters, site characterization, or environmental records that may be useful to a new development. The administration has also directed agencies to identify applicable National Environmental Policy Act categorical exclusions, use FAST-41 permitting coordination where available, and pursue other measures to expedite federal reviews. DOE's own Oak Ridge and Paducah solicitations place responsibility for securing utility interconnection agreements and completing regulatory and permitting plans on applicants. The Army says Fort Bliss and Dugway still require lease negotiations and environmental review. The Air Force's Alaska process is a commercial lease opportunity, not a federal guarantee that a developer has financing, power, construction equipment, network connectivity, customers, or a commissioning date. Federal ownership can create an option on speed. Whether that option becomes actual speed is now an empirical question. The answer will be visible in a much less dramatic set of numbers than the headline gigawatt totals: executed leases, completed environmental reviews, binding power arrangements, interconnection milestones, financed construction phases, energization dates, commissioned megawatts, and operating compute. Portsmouth is the first major federal-land project against which that scorecard can be applied. Fort Bliss and Dugway may become the next tests once final terms are executed. The Air Force and DOE solicitation portfolios remain earlier in the sequence. Washington does not need to win a land-based scale race against the hyperscalers. It needs to prove that land it already controls can move infrastructure through the development chain faster, with transparent cost and risk allocation, than would otherwise be possible. Until that happens, the federal AI land rush should be measured by conversion, not announcements. ### Notes and Sources 1\. US Department of Energy. “DOE Identifies 16 Federal Sites Across the Country for Data Center and AI Infrastructure Development.” April 3, 2025. https://www.energy.gov/articles/doe-identifies-16-federal-sites-across-country-data-center-and-ai-infrastructure 2\. US Department of Energy. “DOE Announces Site Selection for AI Data Center and Energy Infrastructure Development on Federal Lands.” July 24, 2025. https://www.energy.gov/articles/doe-announces-site-selection-ai-data-center-and-energy-infrastructure-development-federal 3\. US Department of Energy. “Site-specific solicitation announcements for Idaho National Laboratory, Savannah River Site, Oak Ridge Reservation, and Paducah.” 2025. https://www.energy.gov/ne/articles/energy-department-seeks-proposals-ai-data-centers-energy-projects-idaho-national 4\. US Department of Energy. “Fact Sheet: Department of Energy Ensuring Affordable Energy Access in Ohio While Powering the Future.” March 20, 2026. https://www.energy.gov/articles/fact-sheet-department-energy-ensuring-affordable-energy-access-ohio-while-powering-future 5\. US Department of Energy Office of Environmental Management. “Partnership Ensures Affordable Energy, Powers AI Future at Portsmouth Site.” March 24, 2026. https://www.energy.gov/em/articles/partnership-ensures-affordable-energy-powers-ai-future-portsmouth-site 6\. Department of the Air Force, Office of Energy Assurance. “Partnership Opportunities.” accessed July 8, 2026. https://www.afcec.af.mil/Energy/Opportunities/ 7\. US Army. “Army reaches conditional agreement with private industry for hyperscaled data centers.” March 26, 2026. https://www.army.mil/article/291360/army\_reaches\_conditional\_agreement\_with\_private\_industry\_for\_hyperscaled\_data\_centers 8\. US House of Representatives, Office of the Law Revision Counsel. “10 USC § 2667: Leases: non-excess property of military departments and Defense Agencies.” current through review date. https://uscode.house.gov/view.xhtml?req=(title:10%20section:2667%20edition:prelim) 9\. Financial Times. “Reporting on proposed Fort Bliss and Dugway project scale and commercial structure.” 2026. https://www.ft.com/content/332c1134-18c4-4e80-a9c3-06aa0c20513e 10\. US Department of Defense Comptroller. “FY2027 Procurement, Defense-Wide Budget Justification.” FY2027. https://comptroller.war.gov/Portals/45/Documents/defbudget/FY2027/budget\_justification/pdfs/02\_Procurement/PROC\_OSD\_PB\_2027.pdf 11\. OpenAI. “The Stargate Project and subsequent site disclosures.” 2025-2026. https://openai.com/index/announcing-the-stargate-project/ 12\. Federal Register. “Executive Order 14318, Accelerating Federal Permitting of Data Center Infrastructure.” July 23, 2025. https://www.federalregister.gov/documents/2025/07/28/2025-14212/accelerating-federal-permitting-of-data-center-infrastructure 13\. The White House. “Executive Order 14347, Restoring the United States Department of War.” September 5, 2025. https://www.whitehouse.gov/presidential-actions/2025/09/restoring-the-united-states-department-of-war/ 14\. Morgan Stanley. “AI Is Now a Macro Variable. Are You Positioned?.” March 9, 2026. https://www.morganstanley.com/insights/articles/ai-market-trends-institute-2026 ### IEA Confirms Data Centers Are Reshaping Global Electricity Demand URL: https://www.aixenergy.io/iea-confirms-data-centers-are-reshaping-global-electricity-demand/ Last updated: 2026-07-26T05:08:11.000Z The IEA expects global electricity demand to grow 3.6% in 2026 and 3.8% in 2027. In the United States, data centers are the main growth driver, while renewables expand, gas generation rises, and emissions increase modestly in 2026. _This post is for subscribers only._ ### China’s Tungsten Controls Expose a Semiconductor Chokepoint URL: https://www.aixenergy.io/chinas-tungsten-controls-expose-a-semiconductor-chokepoint/ Last updated: 2026-07-14T04:27:13.000Z by [Morgan Bazilian](https://www.linkedin.com/in/morganbazilian/?ref=aixenergy.io), [Jahara "FRANKY" Matisek ](https://www.linkedin.com/in/jahara-franky-matisek-phd-9294a9a/?ref=aixenergy.io)and [Brandon Owens](https://www.linkedin.com/in/owensbrandon/?ref=aixenergy.io) Semiconductor manufacturing depends on some of the world’s most sophisticated equipment, but very dependent on a material most investors and policymakers have never heard of. This summer, that material is tungsten hexafluoride, or WF6: a toxic, corrosive, odorless specialty gas used to deposit tungsten in semiconductor manufacturing. 3 0:00 /0:33 1× By June, a stark story had spread through the semiconductor trade coverage. Kanto Denka Kogyo and Central Glass, two major Japanese specialty-chemical producers, were said to be running out of tungsten feedstock and preparing to stop WF6 production around July 1\. Later reports hardened the warning into a declaration that both companies had permanently exited production, removing roughly a quarter of global capacity. 7,8 However, on June 25, Central Glass said it had secured the raw materials needed to meet customer orders and was continuing to supply its semiconductor specialty gases. Kanto Denka’s May 25 financial briefing reported that WF6 sales had increased on higher volumes and price adjustments, forecast substantially higher specialty-gas sales in the fiscal year ending March 2027, and showed continuing WF6 activity at its Chinese affiliate. 1,2 That correction does not make the supply problem disappear. As it turns out, materials can remain legal to export, and a gas producer can remain in operation, while feedstock restrictions, price spikes, inventory risk, and long qualification cycles still create a serious bottleneck. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/07/ChatGPT-Image-Jul-11--2026--02_49_54-PM.png) ## Why WF6 Matters WF6 is used in deposition processes that form tungsten films and features inside logic and memory devices. Tungsten has long been used where copper is unsuitable, including contacts, gates, vias, and 3D NAND word lines. Semiconductor-grade material is extraordinarily pure: commercial products are offered at 99.9995 percent purity, with individual metallic contaminants specified at parts-per-billion-by-weight levels. 3,4,10 Semiconductor fabs qualify materials against process recipes, tool conditions, defect levels, electrical performance, yield, and long-term reliability. A new source of WF6 is therefore not interchangeable with an existing source in the way one barrel of a commodity grade might substitute for another. Trade reports put new-capacity buildout and customer qualification cycles at roughly 18 to 24 months. The response time of a qualified supply chain is much slower than the response time of a spot market. 8 While tungsten remains deeply embedded in qualified manufacturing flows, and replacing it quickly is difficult, molybdenum is already moving into specific metallization applications. Lam Research said in February 2025 that its molybdenum atomic-layer-deposition platform was in volume production, including early adoption in 3D NAND and advanced logic. Kioxia has demonstrated a fluorine-free molybdenum word-line process for 3D flash memory, and Applied Materials said in June 2026 that a molybdenum etch system had been validated in high-volume manufacturing. Substitution is real, but cannot be treated as an emergency drop-in replacement. 10,11,12 ## Supply On February 4, 2025, China’s Ministry of Commerce and General Administration of Customs imposed export controls on selected tungsten-related items under Decision No. 10 of 2025\. The covered list included ammonium paratungstate, tungsten oxide, certain tungsten carbide products, specified solid tungsten products and alloys, and related production technologies. The regime requires exporters to obtain permission. 5 China then tightened controls on dual-use trade involving Japan in early 2026\. Chinese customs data cited by Fastmarkets show that exports of high-purity tungsten powder to Japan fell to zero for three consecutive months from February through April 2026\. Yet WF6 itself was not included on the cited dual-use control list. China exported 210.9 metric tons of WF6 from January through May 2026, including 28.02 metric tons to Japan. 7 The US Geological Survey estimates that China mined 67,000 metric tons of tungsten in 2025 out of a world total of 85,000 metric tons, or about 79 percent. In 2024, China produced 67,000 metric tons out of 82,000, or about 82 percent. 6 The United States has had no commercial tungsten mine production since 2015, although USGS reports that seven US companies can convert concentrates, ammonium paratungstate, tungsten oxide, or scrap into tungsten powders, carbide powders, metal, or chemicals. Outside China, mining is beginning to diversify. Almonty Industries began mining at the Sangdong project in South Korea in March 2026, with a second-phase expansion expected in 2027\. But new mine output and qualified electronic-grade specialty gas are separated by conversion, purification, gas production, logistics, and fab qualification.6,9 ### Prices Reuters reported in late April that ammonium paratungstate prices in Rotterdam had moved above $3,000 per metric ton unit, more than 200 percent higher than at the start of 2026\. Fastmarkets assessed Chinese export APT at $2,600 to $3,200 per metric ton unit on July 1, roughly 600 percent above its level a year earlier. 7,9 WF6 pricing is less transparent. There is no widely used public global benchmark for semiconductor-grade material, and reported prices can mix grades, contract terms, and shipment composition. Chinese customs data cited by TrendForce put the average April 2026 WF6 export unit value at $149.79 per kilogram, 28.33 percent higher than a year earlier and 203.83 percent higher than in March. That month-to-month jump should not be read as a clean spot-price index, but it illustrates the volatility entering the chain. Separately, The Elec reported in October 2025 that several suppliers had notified South Korean chipmakers of WF6 price increases of 70 to 90 percent for 2026. 8,13 Semiconductor buyers can absorb higher material costs for a time because specialty gases represent a small share of the value of a finished advanced chip. But once inventory buffers shrink, the relevant question is whether alternate gas supply has already been qualified. ### AI Demand WF6 is used across logic, DRAM, NAND, and 3D NAND manufacturing. AI infrastructure increases demand for advanced logic and high-bandwidth memory and contributes to strong capital spending across semiconductor manufacturing. But the tungsten squeeze reflects the collision of broad semiconductor demand, more demanding device structures, constrained upstream tungsten flows, and a specialty-gas market with limited transparency and slow qualification cycles. 3,4,7 Higher layer counts in 3D NAND make word-line metallization more demanding, but a 200-layer device does not imply a simple one-for-one count of 200 separate WF6 deposition cycles. Kioxia’s work on molybdenum substitution focuses on the electrical resistance, embeddability, fluorine residue, and leakage challenges that intensify as 3D flash structures scale. 11 This matters because if the problem is described as a single gas suddenly disappearing, the answer appears to be emergency procurement. If the problem is understood as a layered dependency, the response has to include feedstock mapping, conversion capacity, supplier qualification, inventory policy, recycling, and alternative process technologies. The WF₆ episode exposes four linked weaknesses in the semiconductor supply chain: ◆ **Diversification can be superficial.** A fab may source from suppliers headquartered in Japan, South Korea, Europe, or the United States and still depend on the same upstream Chinese feedstock. Resilience analysis must trace the full supply chain, not stop at the direct supplier. ◆ **Unqualified capacity is not real capacity.** An alternate supplier may exist on paper, but if it has not been qualified for the relevant manufacturing process, it cannot be used during a crisis. In semiconductors, qualified process capacity is strategic capacity. ◆ **Mining alone does not create resilience.** The United States and allied countries can expand tungsten production and still remain vulnerable if conversion, high-purity chemical production, packaging, analytical capability, and semiconductor qualification remain concentrated elsewhere. ◆ **Substitution is application-specific.** Molybdenum is already displacing tungsten in selected applications because device scaling is creating technical reasons to change metals. Substitution therefore requires careful attention to process conditions, performance, and qualification. ### Conclusion Central Glass says it is continuing supply. Kanto Denka has reported higher WF6 sales and is forecasting growth in specialty gases. Chinese WF6 exports have continued. The question is how long an opaque, concentrated supply chain can absorb feedstock restrictions, extraordinary price movements, and slow qualification cycles, while semiconductor investment remains strong. 1,2,7 AI buildout depends on industrial chains several tiers upstream: specialty gases, photoresists, precursor chemicals, high-purity metals, substrates, packaging materials, and the equipment needed to make all of them. Their markets are small relative to the systems they enable. The semiconductor industry has spent decades reducing defect rates and increasing efficiency by narrowing materials specifications and qualifying tightly controlled process flows. Those practices are essential to yield. They also make rapid supplier substitution difficult. Resilience, what we have previously termed “pre-logistics” in the military realm, has to be built before the shortages arise. ## Sign up for AIxEnergy AIxEnergy provides independent intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, reliability, investment, and policy. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### Notes and Sources **1\.** Central Glass Co., Ltd., “Regarding the Supply Status of Specialty Gases for Semiconductors,” June 25, 2026\. https://www.cgc-jp.com/news/20260625180000.html **2\.** Kanto Denka Kogyo Co., Ltd., “Financial Results Briefing for the Fiscal Year Ended March 31, 2026,” May 25, 2026, esp. pp. 6, 13, 18, and 20\. https://finance-frontend-pc-dist.west.edge.storage-yahoo.jp/disclosure/20260525/20260525546165.pdf **3\.** EMD Electronics, “Tungsten Hexafluoride Megaclass”; SK ecoplant materials, WF6 product specifications. https://www.emdgroup.com/en/expertise/semiconductors/offering/tungsten-hexafluoride-megaclass.html ; https://www.sk-materials.com/new/kor/html/products/wf6.asp **4\.** SK ecoplant materials, WF6 product information and impurity specifications. https://www.sk-materials.com/new/kor/html/products/wf6.asp **5\.** International Energy Agency, “Decision to Implement Export Controls on Tungsten, Tellurium, Bismuth, Molybdenum and Indium Related Items,” last updated May 5, 2025\. https://www.iea.org/policies/26795-decision-to-implement-export-controls-on-tungsten-tellurium-bismuth-molybdenum-and-indium-related-items **6\.** US Geological Survey, Mineral Commodity Summaries 2026: Tungsten, February 2026\. https://pubs.usgs.gov/periodicals/mcs2026/mcs2026-tungsten.pdf **7\.** Fastmarkets, “AI Specialty Gas Tightness Drives China’s Best-Performing A-Share in H1,” July 7, 2026\. https://www.fastmarkets.com/insights/ai-specialty-gas-tightness-drives-chinas-best-performing-a-share-in-h1/ **8\.** TrendForce, “Key Semiconductor Gas WF6 Prices Reportedly Surge Over 200% as Supply Tightens Ahead of Japan Output Cuts,” June 12, 2026\. https://www.trendforce.com/news/2026/06/12/news-key-semiconductor-gas-wf%E2%82%86-prices-reportedly-surge-over-200-in-china-as-supply-tightens-ahead-of-japan-output-cuts/ **9\.** Reuters, “Tungsten Breaks Records as China Export Curbs, Military Demand Boost Investment,” April 29, 2026; Reuters, “Tungsten Rises to Record Highs as Export Curbs Turn Up Supply Heat,” January 29, 2026\. https://www.reuters.com/world/asia-pacific/tungsten-breaks-records-china-export-curbs-military-demand-boost-investment-2026-04-29/ ; https://www.reuters.com/world/americas/tungsten-rises-record-highs-export-curbs-turn-up-supply-heat-2026-01-29/ **10\.** Lam Research, “Lam Research Ushers in New Era of Semiconductor Metallization with ALTUS Halo for Molybdenum Atomic Layer Deposition,” February 19, 2025\. https://investor.lamresearch.com/2025-02-19-Lam-Research-Ushers-in-New-Era-of-Semiconductor-Metallization-with-ALTUS-R-Halo-for-Molybdenum-Atomic-Layer-Deposition **11\.** Kioxia, “Fluorine-Free Word Line Molybdenum Process for Enhancing Scalability and Reliability in 3D Flash Memory,” September 24, 2024\. https://www.kioxia.com/en-jp/rd/technology/topics/topics-71.html **12\.** Applied Materials, “Applied Materials Unveils Deposition and Selective Etch Systems to Advance 3D Chip Scaling,” June 15, 2026\. https://ir.appliedmaterials.com/news-releases/news-release-details/applied-materials-unveils-deposition-and-selective-etch-systems/ **13\.** The Elec, “Chipmakers Samsung and SK Hynix Face WF6 Gas Price Hike,” October 28, 2025\. https://www.thelec.net/news/articleView.html?idxno=5468 ### Moore’s Law Is Moving Downstream URL: https://www.aixenergy.io/moores-law-is-moving-downstream/ Last updated: 2026-07-29T19:36:57.000Z IBM’s NanoStack research shows that semiconductor scaling is not over. But for AI, cheaper computation can shift the binding constraints outward—to power delivery, cooling, grid equipment, and the institutions that decide who pays for expansion. _This post is for subscribers only._ ### July 2026 Monthly Digest URL: https://www.aixenergy.io/july-2026-monthly-digest/ Last updated: 2026-07-11T22:00:38.000Z The central problem in the AI-energy system is no longer forecasting demand. It is governing commitments made before demand can be known with confidence. _This post is for paying subscribers only._ ### Why Utilities Need an AI Governance Architecture URL: https://www.aixenergy.io/utilities-need-an-ai-governance-architecture-not-another-framework/ Last updated: 2026-07-29T19:35:53.000Z How NIST, ISO, OWASP, Microsoft, Google, and OECD fit together, and what the electric industry must build for itself _This post is for subscribers only._ ### The Electricity Customer Bill of Rights Part I: A Framework for Infrastructure-Scale Electric Demand URL: https://www.aixenergy.io/the-electricity-customer-bill-of-rights-part-i-a-constitutional-framework-for-infrastructure-scale-electric-demand/ Last updated: 2026-07-29T19:38:48.000Z The AI economy needs electricity, but customers should not subsidize its growth. This series proposes an Electricity Customer Bill of Rights to protect affordability, reliability, transparency, communities, and fair cost allocation while enabling responsible infrastructure-scale demand. _This post is for paying subscribers only._ ### America’s Data Center Reckoning URL: https://www.aixenergy.io/americas-data-center-reckoning/ Last updated: 2026-07-29T19:39:18.000Z A wave of data center moratoriums is spreading across the United States as local governments and state legislatures try to catch up with the infrastructure demands of artificial intelligence. _This post is for paying subscribers only._ ### How Artificial Intelligence Data Centers Are Reshaping the Regulatory Architecture of U.S. Electricity URL: https://www.aixenergy.io/how-artificial-intelligence-data-centers-are-reshaping-the-regulatory-architecture-of-u-s-electricity/ Last updated: 2026-07-15T21:23:12.000Z American utilities spent most of the past two decades managing a grid whose total electricity demand, adjusted for weather and GDP, was essentially flat. Load forecasters grew accustomed to small positive numbers, and the system's institutions—interconnection queues, capacity markets, integrated resource plans—were calibrated accordingly. That era is over. The proximate cause is artificial intelligence. Interview with Brandon Owens 0:00 /238.146735 1× The more important point is not simply that demand is growing again. It is that the character of demand is changing. AI data centers combine the electrical scale of heavy industry, the mobility of digital commerce, the capital structure of hyperscale technology firms, and the operational logic of software systems. They are not ordinary factories with server racks inside them. A steel mill, refinery, or LNG terminal is tied to a physical production process, a local supply chain, and a relatively stable industrial use case. Some AI workloads, by contrast, may be divisible across time, geography, or computing environments. Some training runs can be deferred. Some inference can be routed across regions. Some facilities can draw on backup or dedicated generation. Some projects can relocate before capital is fully committed. This hybrid character makes AI data centers difficult to govern through regulatory categories developed for passive retail loads and conventional industrial customers. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/06/ChatGPT-Image-Jun-21--2026--08_25_52-AM.png) Lawrence Berkeley National Laboratory's 2024 report—the first congressionally mandated update to its widely cited 2016 study—documents the scale of the break. U.S. data centers consumed approximately 176 terawatt-hours (TWh) in 2023, or 4.4 percent of national electricity consumption, up from 58 TWh a decade earlier. The LBNL authors project that figure will reach between 325 and 580 TWh by 2028, representing 6.7 to 12 percent of total national supply—a range so wide that it admits the honest uncertainty embedded in the forecasts (Shehabi et al., 2024). Much of the recent acceleration is attributable to hyperscale and GPU-intensive AI workloads, although cloud growth, enterprise computing, cryptocurrency, and broader digital demand also contribute. National aggregates understate geographic intensity. Northern Virginia hosts what is now the largest concentration of data center capacity in the world—by some estimates more than 4,900 megawatts of operational IT load as of early 2025, constituting roughly 13 percent of all reported global data center capacity and 25 percent of capacity in the Americas (JLARC, 2024; Northern Virginia Technology Council, 2026). PJM Interconnection, which operates the grid serving that region, projects that roughly 30 of the 32 gigawatts of peak load growth it anticipates between 2024 and 2030 will come from data centers—a proportion the grid operator's own board chair described as an "onrush of demand" (PJM, 2025a; PJM Inside Lines, 2025). MISO projects its peak load will climb from 121 GW in 2025 to 163 GW by 2035, with data centers potentially comprising a quarter of its total demand by 2040 (Data Center Dynamics, 2025). FERC's March 2026 State of the Markets report confirmed that more than 50 GW of data center capacity was operating nationwide at year-end 2025, representing 24 percent compound annual growth since 2020 (FERC, 2026). What makes the current moment difficult for regulators is the combination of scale, speed, and category confusion. Data centers are retail customers, but they are also transmission-planning drivers, capacity-market drivers, potential demand-response resources, candidates for co-located generation, and sources of dynamic reliability behavior. Institutional frameworks built around slow, incremental load growth—the interconnection queue, the three-year-forward capacity market, the integrated resource plan with its five-year planning horizon—are being asked to absorb multi-gigawatt demand additions that can materialize, or evaporate, within months. The result is a proliferating set of regulatory responses that are, in many cases, working at cross-purposes. The article that follows maps those responses, identifies their common pressures and contradictions, and asks whether the current architecture is adequate to the task. ## WHY AI DATA CENTERS ARE DIFFERENT The regulatory challenge begins with the load itself. Large industrial customers are not new to the power sector. Utilities have long served steel mills, aluminum smelters, chemical plants, refineries, mines, and manufacturing complexes. These customers can be very large, and some have historically provided interruptible load or demand response. But AI data centers differ from conventional large loads in several important respects. 1. First, their development cycle is unusually fast relative to the infrastructure required to serve them. A data center developer can identify a site, submit an inquiry, and begin commercial negotiation far more quickly than a utility can plan, permit, procure, and build the transmission, substation, transformer, and generation resources that may be needed to serve hundreds of megawatts of new load. 2. Second, their load forecasts are unusually uncertain. A proposed data center campus may appear in utility or RTO forecasts before it has secured final permits, financing, equipment, power supply, or customer commitments. Conversely, once a hyperscaler commits to a region, load can ramp quickly and in large increments. This creates a forecasting problem that is qualitatively different from ordinary economic load growth. 3. Third, the operational characteristics of AI data centers are not fully captured by conventional planning models. Their steady-state demand may appear as a high-load-factor block of consumption, but their response to voltage disturbances, protection-system operations, uninterruptible power supply transfers, backup-generation starts, and software-driven workload management may be more dynamic than that of a conventional commercial load. 4. Fourth, data centers have more strategic options than many earlier industrial loads. They can sign long-term power purchase agreements, co-locate with generation, install backup generation and batteries, build private substations, participate in demand response pilots, or shift some workloads across time and geography. These options do not eliminate the need for reliable grid service, but they change the bargaining relationship between customer and utility. The core regulatory question is therefore not whether data centers should be encouraged or discouraged. The question is how to classify, price, verify, and operate a new class of load that does not fit neatly into inherited categories. ## FEDERAL REGULATION ### FERC AND THE EMERGING FEDERAL FRAMEWORK FOR LARGE LOADS The Federal Energy Regulatory Commission sits at the center of U.S. electricity regulation where interstate transmission and wholesale power markets are concerned. Its authority under the Federal Power Act is well established for transmission service, wholesale rates, regional market rules, and generator interconnection. What has historically been less clear is how that authority applies when the transmission system is being accessed not by a generator, but by a new class of infrastructure-scale load. For most of the industry's history, that distinction created few institutional challenges. Large industrial customers connected through utility service territories, state-regulated tariffs, and local interconnection arrangements. Their effects on transmission systems, capacity markets, and reliability planning could generally be absorbed through existing utility forecasting and planning processes. The emergence of AI data centers, advanced manufacturing facilities, and other gigawatt-scale loads has exposed the limits of that framework. A single AI campus may request hundreds or even thousands of megawatts. Regional clusters of such facilities can materially alter transmission plans, capacity-market outcomes, generation investment decisions, and long-term reliability assessments. Increasingly, these projects are arriving on timelines measured in months rather than years, reflecting the pace of the digital economy rather than the pace of traditional utility infrastructure development. The clearest early manifestation of this challenge emerged in PJM, where developers proposed co-location arrangements that would place large data center loads at or near generating facilities. These arrangements promised faster access to power in regions facing transmission constraints, generation shortages, and interconnection backlogs. They also exposed significant gaps in existing tariffs regarding transmission rights, cost allocation, reliability obligations, and wholesale-market participation. FERC initially addressed these issues through a series of PJM-specific proceedings, including the rejection of the proposed Susquehanna-Amazon co-location arrangement and subsequent orders directing PJM to develop new tariff provisions governing co-located loads. Those proceedings revealed a broader reality: large loads located near generation may reduce certain network requirements, but they do not necessarily eliminate dependence on the transmission system for backup service, balancing, ancillary services, reliability support, market settlement, or emergency operations. Physical proximity to generation does not eliminate public-grid obligations. Behind the meter is not always behind the grid. The significance of these disputes expanded dramatically on June 18, 2026, when FERC issued coordinated Section 206 show-cause orders directing all six RTOs and ISOs under its jurisdiction—PJM, MISO, SPP, NYISO, ISO New England, and CAISO—and their transmission-owning utilities to either justify or reform their tariff frameworks for large-load integration. At the same time, FERC directed each region to explain how it would ensure adequate generation resources exist to serve both existing customers and the next wave of large electrical loads. The scope of the proceeding marks an important shift in federal policy. For decades, large-load growth was treated primarily as a local utility planning issue. FERC is now treating large-load integration as a national transmission, market-design, and reliability challenge. The Commission explicitly tied the proceeding to the Department of Energy's October 2025 large-load initiative and repeatedly framed the effort around improving "speed-to-power" for strategically important industries including artificial intelligence, advanced manufacturing, and digital infrastructure. The Commission identified five broad areas where existing tariffs may no longer be adequate: (1) Transmission service applications and study processes; (2) Cost allocation and transparency; (3) Co-location and behind-the-meter generation; (4) Flexible transmission service; and (5) Planning for generation serving large electrical loads. Taken together, these categories represent more than a collection of technical tariff issues. They reflect an attempt to redesign the interface between the electric grid and a new generation of infrastructure-intensive industries. FERC's emerging large-load agenda should therefore be understood alongside Orders 2023 and 1920\. Order 2023 addressed generator interconnection through cluster studies, readiness requirements, and queue reform. Order 1920 addressed long-term regional transmission planning and cost allocation. The June 2026 proceeding introduces what may become the third major pillar of AI-era grid governance: the rules governing how large loads connect to the transmission system, how quickly they can be studied, what obligations they must satisfy, how flexibility is valued, and who pays for the infrastructure required to serve them. Particularly notable is FERC's emphasis on flexible transmission service. Historically, transmission service has largely been structured around a binary choice between firm and non-firm service. The Commission is now explicitly exploring whether large loads capable of curtailment, staged energization, demand flexibility, storage integration, or operational coordination should be offered alternative service structures. This reflects an emerging recognition that some AI infrastructure may function not only as load, but as a controllable grid resource. The jurisdictional issue is therefore not simply whether FERC should regulate data centers. The issue is that a single AI-scale project can simultaneously become a retail customer, a transmission-planning driver, a capacity-market driver, a reliability concern, a cost-allocation issue, and a wholesale-market participant through co-located generation or flexible service arrangements. State commissions retain authority over retail tariffs, utility cost recovery, and siting approvals. FERC retains authority over interstate transmission and wholesale markets. RTOs and ISOs administer regional tariffs, planning processes, and resource adequacy frameworks. NERC establishes reliability standards and assessments. Yet no single institution governs the entire transaction. That fragmentation was manageable when load growth was gradual and predictable. It becomes increasingly fragile when hundreds or thousands of megawatts can enter planning forecasts before a project's commercial viability has been fully demonstrated. The large-load question is therefore as much a federalism challenge as it is an interconnection challenge. The emerging federal objective is not to displace state authority. It is to establish a coherent framework governing transmission access, reliability obligations, cost responsibility, operational flexibility, and resource adequacy for a new class of infrastructure-scale load. In doing so, FERC has effectively acknowledged that artificial intelligence is no longer merely a technology-sector issue. It has become a power-system governance issue. The June 2026 proceeding may ultimately be remembered not as a tariff case, but as the moment federal regulators formally recognized that the future of the AI economy depends as much on electricity market design as on software, semiconductors, and compute infrastructure. ### NERC AND THE RELIABILITY PROGRAM NERC's 2025 Long-Term Reliability Assessment, released in January 2026, is the starkest warning the organization has issued in at least two decades. Summer peak demand across the North American bulk power system is forecast to grow by 224 GW over the next ten years—69 percent higher than the 132 GW projected in the previous year's LTRA—while winter peak demand is projected to surge by 246 GW (NERC, 2025). To put those numbers in context: the entire current generating capacity of PJM is approximately 182 GW. Thirteen of NERC's twenty-three assessment areas now face elevated or high resource adequacy risk within the five-year planning horizon, including the MISO, PJM, ERCOT, and Northwest regions (NERC, 2025; Microgrid Knowledge, 2026). NERC's director of reliability assessments, John Moura, put it plainly: "The system is changing faster than the infrastructure needed to support it" (Utility Dive, January 30, 2026). What distinguishes NERC's 2025 assessment from prior warnings is the precision of its attribution. Artificial intelligence data centers and the broader digital economy account for most of the projected load growth—though the report is careful to note that large industrial facilities, electrification, cryptocurrency mining, and demographic change all contribute (NERC, 2025). The assessment also identified a new reliability phenomenon: a series of events in 2024 and 2025 in which 1,000 MW or more of data center load dropped unexpectedly from the bulk power system simultaneously, triggering frequency and voltage disturbances that required emergency management. A July 2024 voltage fluctuation in Northern Virginia simultaneously disconnected sixty data centers, producing a 1,500 MW power surplus that operators had to absorb in real time (Belfer Center, 2026). NERC issued a Level 3 alert on this problem in May 2026\. In April 2025, FERC directed NERC to investigate these load-loss events formally, and the resulting Large Load Task Force is the first systematic federal effort to characterize data center load dynamics for planning purposes (White & Case, 2025b). Peer-reviewed engineering literature corroborates the concern. Kwon, Mukherjee, and Adetola (2025), writing from Pacific Northwest National Laboratory, classify AI data centers as "Large Dynamic Digital Loads" whose training and inference cycles generate sharp load ramps—on the order of minutes—that stress transfer capabilities and introduce transient stability risks not anticipated in conventional planning (Kwon et al., 2025). A 2026 study in Energies similarly documents that the combination of high power density, variable GPU utilization, and intensive cooling requirements reshapes system energy use in ways that standard load models were not designed to capture (Chen et al., 2026). The translation from engineering paper to regulatory standard is slow; NERC's standards development process operates on multi-year cycles even for priorities. For reliability planners, the central question is not only how much electricity data centers consume, but how they behave during disturbances. A large computational load that trips, transfers to UPS, starts backup generation, or rapidly sheds noncritical workload during a voltage event is not equivalent to a static block of demand. If many facilities respond similarly to the same disturbance, the resulting load change can become material to frequency, voltage, and balancing operations. The relevant planning questions therefore include ride-through behavior, protection settings, telemetry, commissioning data, backup-generation logic, and whether operators have sufficient visibility into large-load response during contingencies. In that sense, the reliability challenge is shifting from serving large load to observing, modeling, and coordinating large load. ## REGIONAL RESPONSES ### PJM: The Pressure Cooker PJM is not merely one regional example among many. It is the first large-scale stress test of how AI data center growth interacts with capacity markets, transmission planning, co-location, state retail regulation, and cost allocation. The region contains the country’s most concentrated data center load, relies on a forward capacity market, and spans multiple states with different regulatory priorities. If the United States is going to learn how data center load reshapes electricity governance, PJM therefore provides the clearest early case study. No regional entity has absorbed the data center shock more visibly than PJM Interconnection. The mechanism through which that shock transmits to consumers is PJM's capacity market, which holds three-year-forward auctions to procure the reserves needed to meet peak demand. For decades these auctions were, from a public perspective, unremarkable. That ended abruptly. The 2025/2026 capacity auction—held in 2024 for the delivery year beginning June 2025—cleared at a price approximately 833 percent above the prior year's level, driven by the surge in forecast data center load in the PJM footprint (AAF, 2026; Introl, 2026). Subsequent auctions have cleared at or near the FERC-approved price cap: $329.17 per megawatt-day for 2026/2027 and $333.44 per megawatt-day for 2027/2028 (PJM, 2025b). The 2027/2028 auction—held in December 2025—totaled $16.4 billion and, critically, fell approximately 6,625 MW short of PJM's 20 percent installed reserve margin target, marking the first time in the grid operator's history that it failed to procure sufficient capacity for reliability (Utility Dive, December 2025). Monitoring Analytics, PJM's independent market monitor, has been unsparing in its diagnosis. Data center load—both existing facilities and projected builds—accounted for $6.5 billion, or 40 percent, of the 2027/2028 auction's total cost; across PJM's last three base capacity auctions, data center-related charges totaled $21.3 billion, or 45 percent of the $47.2 billion in aggregate clearing costs (Monitoring Analytics, 2026). "Data center load growth is the primary reason for recent and expected capacity market conditions," the market monitor wrote in January 2026—a conclusion it has now repeated in multiple quarterly reports. By May 2026, wholesale power costs in PJM for the first quarter had risen 75.5 percent year-over-year, from $77.78 to $136.53 per megawatt-hour, with data center growth again identified as the primary driver (The Register, May 2026). Load forecasting has become a contested issue in its own right. The problem is circular: PJM must include data center load in its forecast to set the capacity auction clearing price, but data center developers routinely file requests for load adjustments without firm commitments—meaning phantom facilities inflate the forecast, which drives up the auction price, which raises costs for all ratepayers whether or not those facilities are ever built. PJM's market monitor characterized data center forecasts as having "extreme uncertainty." In January 2026, PJM revised its near-term forecast downward by applying stricter vetting criteria—reducing the 2028 peak projection by 4.4 GW, or 2.6 percent—while simultaneously increasing its long-term annual growth rate projection to 3.6 percent per year through 2036 (Utility Dive, January 15, 2026; PJM Inside Lines, 2026). The problem is that data center development is easier to announce than to verify. A proposed campus may have site control or a utility inquiry long before it has an executed electric service agreement, posted collateral, land-use approvals, major equipment orders, construction financing, or a realistic energization date. Yet once the load enters a regional forecast, it can affect capacity procurement, reserve margins, transmission planning, and ultimately customer bills. This creates a pathway through which speculative load can impose real system costs before it becomes real demand. The logical response is milestone-based load verification. RTOs, utilities, and state commissions should not treat all proposed data center load as equally certain. Instead, load should enter planning and capacity-market assumptions according to verified project maturity: site control, zoning approval, executed service agreement, posted security, interconnection-study completion, major equipment procurement, construction start, and scheduled energization. The purpose is not to exclude uncertain growth. It is to weight uncertainty explicitly. A 300 MW inquiry should not have the same planning status as a 300 MW project with permits, financing, executed service contracts, and equipment on order. Forecasting, in this context, becomes less an exercise in predicting economic growth than a process for verifying infrastructure commitments. PJM's own institutional response has been halting. In August 2025, the Board launched a Critical Issue Fast Path (CIFP) for Large Load Additions, proposing a Non-Capacity Backed Load service that would allow new data centers to receive grid connection in exchange for agreeing to curtail during declared emergencies—effectively a firm load-shedding obligation in return for expedited access (PJM Inside Lines, 2025). Stakeholders failed to reach consensus on these rules by November 2025, and FERC's December order superseded some of what PJM had been developing. The grid operator simultaneously began processing more than 170,000 MW of queued generation projects and introduced a multi-year collaboration with Google and Tapestry to deploy AI-enhanced tools for interconnection study processing (PJM, 2025b; PJM Inside Lines, 2026). ### MISO and SPP: Fast-Track Experiments MISO has experienced the sharpest regional acceleration: data center capacity in its fifteen-state Midwest and South footprint grew at a 43 percent compound annual rate between 2020 and 2025, and the grid operator expects 8 to 14 GW of new data center load in 2026 and 2027 alone (FERC, 2026; Data Center Dynamics, 2025). By 2030 it projects data centers will constitute a fifth of its electricity demand; by 2040, potentially a quarter. Its ten-year peak load forecast has risen from 121 GW in 2025 to a projected 163 GW in 2035. MISO's primary institutional response has been its Expedited Resource Addition Study process, an accelerated interconnection pathway designed specifically to address resource adequacy deficits. The second ERAS cycle, announced in December 2025, targeted 6.1 GW of new capacity, with battery storage and natural gas projects constituting the bulk of the queue (DWGP, 2025). Importantly, MISO's ERAS is structured to serve resource adequacy gaps of individual load-serving entities, not simply to accelerate any and all project requests—a design choice that limits the risk of queue inflation while prioritizing the most pressing reliability needs. SPP filed its own Expedited Resource Addition Study with FERC in May 2025, citing what it characterized as an emerging resource adequacy crisis in its fourteen-state central U.S. region—where annual load rose approximately 21 percent between 2020 and 2024 (S&P Global, 2025; K&L Gates, 2025). FERC approved a complementary Priority Process in November 2025, permitting existing generating facility owners to increase injection capability by up to 20 percent outside the normal interconnection queue. Both MISO and SPP are effectively using expedited interconnection as a supply-side substitute for the demand-side management tools they lack authority to impose directly on data center customers. ### ERCOT: A Different Legal Terrain The Electric Reliability Council of Texas operates outside FERC's jurisdiction—an artifact of Texas's decision to maintain an essentially intrastate grid—and has consequently had more freedom to act quickly, though its tools differ. ERCOT's peak demand could rise from 98 GW in 2026 to over 111 GW by 2032, predominantly on data center and cryptocurrency growth, and its T&D development pipeline has quintupled from under 500 miles in 2024 to over 2,400 miles in planned or under-construction transmission (ERCOT, 2025; S&P Global, 2025). Texas recorded the largest year-on-year increase in wholesale electricity sales of any regional system in 2025, at 5.2 percent (FERC, 2026). At the legislative level, Texas enacted SB 6 in 2025, directing the Public Utility Commission of Texas to develop both mandatory and voluntary demand management programs for large loads. The mandatory component requires curtailment protocols for new large loads of 75 MW or more that interconnect after December 31, 2025, during firm load-shed events; the voluntary Large Load Demand Management Service allows ERCOT to competitively procure demand reductions from such loads ahead of anticipated emergency conditions (Yes Energy, 2026). ERCOT also established a fast-track large load interconnection process and launched its Real-Time Co-optimization plus Batteries initiative in December 2025\. Texas's approach—imposing curtailability as a condition of rapid interconnection—represents the most developed attempt by any U.S. grid operator to directly restructure data center demand behavior rather than simply managing its consequences. ## STATE-LEVEL RESPONSES: PUBLIC UTILITY COMMISSIONS AND LARGE LOAD TARIFFS ### Virginia: The Epicenter Virginia's regulatory response is the most fully developed in the country, partly because it has had the most time to absorb a problem that arrived first and hardest at its doorstep. Northern Virginia is widely regarded as the world’s dominant data center market—constituting some 13 percent of reported global data center capacity, a position built over two decades on cheap land, abundant fiber infrastructure, proximity to federal government clients, and generous state tax incentives (JLARC, 2024). Virginia's data center sales and use tax exemption, available through 2035, catalyzed an estimated $24 billion in qualified investment in fiscal 2023 alone. The consequence for ratepayers has been a 183 percent projected increase in electricity demand in Dominion Energy's service territory by 2040, and a $7.6 billion planned transmission expansion whose costs, by Dominion's own admission, will initially fall 55 percent on residential customers (JLARC, 2024; Virginia Mercury, April 2025). JLARC's December 2024 report was blunt about the trajectory. Although data centers currently pay accurately for their load under existing tariff structures, future infrastructure costs—new transmission, new generation—will be socialized across all ratepayers unless allocation rules change. JLARC projected that data center growth could add $444 per year to a typical Dominion residential bill by 2040 if nothing changes (JLARC, 2024). That finding drove the Dominion rate case that dominated Virginia energy politics through 2025. In November 2025, the Virginia State Corporation Commission approved Dominion's proposed GS-5 rate class—a new tariff applicable to customers demanding 25 MW or more at a monthly load factor exceeding 75 percent, effective January 1, 2027 (SCC, PUR-2025-00058, 2025). The class imposes minimum demand charges of 85 percent of contracted distribution and transmission demand and 60 percent of generation demand, with 14-year service contracts. It also rejected Dominion's full base-rate request: the company sought $822 million in new 2026 revenue; the Commission approved $565.7 million (SCC, 2025). Simultaneously, Virginia SB 253, introduced in February 2026, would authorize the SCC to shift distribution costs and PJM capacity auction costs more directly onto the GS-5 class through 2033—a proposal its sponsor described as the only energy bill in the session likely to reduce residential rates in the near term (Virginia Mercury, February 2026). The SCC's November order was widely praised as a meaningful first step but criticized by environmental and consumer advocates as insufficient. The Piedmont Environmental Council's analysis suggested that the approved tariff would still leave 61 percent of data center-driven grid upgrade costs on residential ratepayers after the 14-year contract term—the period during which those costs are allocated to the GS-5 class (Inside Climate News, 2026). The SCC itself acknowledged the limits of what it was doing, noting that it expected to revisit the tariff design in 2027 "with two more years' experience." ### Other State Responses Virginia's approach has not emerged in isolation. American Electric Power subsidiaries have proposed large load tariffs across multiple state jurisdictions: Indiana, West Virginia, and Kentucky require 12-year contracts, collateral posting, and minimum demand charges. AEP Ohio's tariff requires demand charges equal to 85 percent of the 11-month rolling average bill and an 8-year minimum contract, with initial-year costs for a 100 MW facility running close to $10 million; the practical result has been a roughly 50 percent reduction in new connection requests in AEP Ohio's territory, which Enverus analysts characterized as "clearing the queue" of speculative projects (EUCI, 2025). Ohio's Public Utilities Commission is still adjudicating competing settlement proposals from data center operators who pushed back on those terms. Georgia and Indiana show how quickly the tariff conversation is spreading beyond the earliest data center hubs. Georgia regulators have focused on ensuring that large new loads bear the generation and infrastructure costs required to serve them, while attempting to limit near-term impacts on ordinary retail customers. Indiana’s approach, reflected in the Indiana Michigan Power settlement, is more collaborative: it creates a framework for very large customers while requiring them to cover network upgrades and report material reductions in committed load. These examples suggest that state commissions are converging around a common toolkit: minimum bills, long-term service commitments, collateral, exit fees, direct upgrade funding, reporting obligations, and periodic review of actual load realization. A 2025 EUCI survey of eleven large load tariffs across the country found consistent patterns: average minimum bills of 80 percent of contracted demand, average contract lengths of fourteen years, security deposits averaging seven years of minimum bills, and exit fees averaging five years of minimum bills (EUCI, 2025). Trigger thresholds ranged widely—from 5 MW to 500 MW—with an average of 114 MW. NARUC has convened state commissioners through its Center for Partnerships and Innovation to share methodology and track developments, reflecting a growing recognition that what happens in Virginia and Ohio will influence regulatory design far beyond those states. The important point is not that all states should adopt the same tariff. They should not. Grid conditions, market structures, economic development priorities, and utility cost-recovery mechanisms differ. The common principle is that large-load tariffs are becoming instruments of forecast discipline. They do not merely recover costs; they test whether a proposed load is real enough to justify public infrastructure commitments. ## COST ALLOCATION: THE PUBLIC LEGITIMACY PROBLEM The cost-allocation problem is where the technical debate becomes a political one. Data centers can bring jobs, tax base, digital infrastructure, and strategic economic value. But the public legitimacy of the AI infrastructure buildout will depend on whether ordinary customers believe they are being asked to finance the grid requirements of some of the world’s largest technology companies. The costs at issue are not limited to the energy consumed by the data center. Energy charges are often the easiest part of the problem. The harder questions involve transmission upgrades, distribution substations, transformers, generation procurement, capacity-market obligations, backup service, and stranded assets if a project is delayed, downsized, or canceled. A tariff that recovers volumetric energy costs may still fail to recover the fixed system costs created by reserved capacity. This is why minimum demand charges, take-or-pay contracts, collateral requirements, and exit fees have become central to large-load tariff design. They are not simply punitive devices. Properly designed, they align commercial commitment with infrastructure commitment. If a customer wants the grid to reserve hundreds of megawatts of capacity, the customer should carry a corresponding obligation if the project fails to materialize. The policy balance is delicate. Excessive charges could drive data centers toward self-supply, neighboring jurisdictions, or behind-the-meter arrangements that reduce regulatory visibility. Insufficient charges could socialize costs across residential and small commercial customers. The central design question is therefore not whether data centers should pay more, but how to make their payment obligation match the cost, timing, uncertainty, and reliability implications of the service they require. ## UTILITY-LEVEL RESPONSES AND THE RISE OF CUSTOMER-DRIVEN INFRASTRUCTURE At the most granular level, investor-owned utilities are navigating the data center surge through a combination of tariff design, long-range resource planning, and—increasingly—experimentation with demand flexibility. At least thirty-six utilities had adopted large load tariffs of varying design by late 2025, including Dominion Energy, Wisconsin Electric Power, and Arizona Public Service (EUCI, 2025). The proliferation reflects a common set of concerns: the risk of building infrastructure for facilities that never materialize, the specter of stranded assets, and the equity question of who pays for grid upgrades that primarily serve a narrow commercial class. The stranded-asset risk is not theoretical. AEP subsidiaries in Indiana, West Virginia, and Kentucky collectively acquired approximately 750 MW of generating capacity to serve data centers that were ultimately not built; those utilities are now seeking to sell that capacity into PJM wholesale markets (EUCI, 2025). The episode illustrates the danger in the current model, where utilities are asked to make irreversible capital commitments against highly uncertain load forecasts—sometimes within compressed timelines set by data center developers eager to begin construction. Dominion Energy, the utility most directly exposed to the Northern Virginia boom, has used its integrated resource planning process as well as the GS-5 tariff to manage the transition. Its 2024 IRP projects approximately 27 GW of new generation by 2039, including 21 GW of renewable energy and small modular reactors and roughly 6 GW of natural gas (Belfer Center, 2026). The IRP's scale—and the pace of transmission expansion it implies—gives some sense of what the data center era means in practical planning terms. One important frontier, still underdeveloped in formal regulation, is data center demand flexibility. The Electric Power Research Institute's DCFlex initiative—a partnership involving Google, Meta, Microsoft, Duke Energy, PJM, and more than forty other organizations—is testing demand response, GPU workload shifting, and the use of uninterruptible power supplies as dispatchable grid resources (ITIF, 2025). Lawrence Berkeley National Laboratory's recent synthesis of the large load flexibility literature suggests that incorporating even modest data center flexibility into integrated resource planning—rather than treating data centers as fixed, inflexible loads—could generate net present value savings of $300 to $400 million over the 2025 to 2050 period (LBL, cited in Knowledge Problem, 2026). The 2025 Nicholas Institute report "Rethinking Load Growth," produced at Duke University, estimated that the U.S. grid could reliably absorb 76 to 126 GW of new flexible demand with no additional capacity expansion, provided those loads accepted curtailment for as little as 0.25 to 1 percent of annual hours (ITIF, 2025). These estimates are large enough to warrant regulatory attention, even if their precise magnitude remains uncertain. On the generation side, the nuclear co-location pipeline has grown with unusual speed. IEA data from April 2026 placed conditional offtake agreements between data center operators and small modular reactor projects at 45 GW, up from 25 GW at the end of 2024 (IEA, 2026). Whether that pipeline matures into operating capacity is an open question—SMR technology remains commercially unproven at scale, and permitting timelines are long—but the trend suggests that AI demand may accelerate nuclear development in ways that energy policy alone could not have achieved. In the interim, many developers have turned to onsite natural gas generation to circumvent interconnection delays, a workaround that the IEA flags as raising its own reliability concerns given the rapid and large demand swings that AI training workloads impose (IEA, 2026). A parallel development is the emergence of what might be called a shadow grid: a layer of customer-driven energy infrastructure built around the needs of hyperscale computing. It includes co-located generation, dedicated power purchase agreements, private substations, battery-backed campuses, backup generation fleets, microgrid concepts, and proposals for nuclear or gas-fired supply tied to specific data center loads. The term should not be read to mean that data centers are separating from the public grid. In most cases, they remain dependent on the grid for backup, balancing, transmission access, and market settlement. But the shadow grid concept captures an important change in bargaining structure. The data center is no longer merely a passive customer waiting at the end of a utility line. It increasingly arrives with its own generation strategy, flexibility options, carbon commitments, and infrastructure demands. ## REGULATORY RESPONSE The regulatory pattern that emerges from FERC, NERC, PJM, ERCOT, Virginia, Ohio, and the utility tariff cases is not yet a coherent architecture. It is a set of improvisations. But those improvisations point toward a more durable framework for governing AI-era load. Executive Exhibit 2 ## Regulatory and Operational Response Matrix for AI-Scale Load Growth AI data centers are no longer treated simply as large retail customers. They now affect transmission access, resource adequacy, capacity markets, reliability standards, state tariff design, utility capital planning, and customer-risk allocation. **Strategic implication:** the emerging governance model is shifting from passive service obligation to verified deliverability. Large loads must demonstrate project maturity, carry the costs they cause, provide measurable flexibility where claimed, and give system operators enough visibility to protect reliability. | Actor / Tier | Primary Authority | Key Actions & Proposals (2024–2026) | Cost Allocation | Load Forecasting | Reliability Measures | | ---------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | FERC Federal | FPA §§ 201, 205, 206; interstate transmission and wholesale markets. | Order 2023 generator interconnection reform; Order 1920 long-term regional transmission planning; October 2025 DOE-directed large-load initiative; June 2026 Section 206 show-cause orders directing all six RTOs/ISOs and transmission owners to justify or reform large-load tariffs within 60 days. Reform categories include study processes, cost allocation, co-location and behind-the-meter generation, flexible transmission service, and planning for generation serving large loads. | Moves the federal debate toward cost causation, transparency, participant responsibility for network upgrades, and tariff structures that reduce cross-subsidization between large new loads and existing customers. | Requires organized markets to confront whether current load forecasts, study queues, readiness requirements, and project-verification methods are adequate for multi-hundred-megawatt and gigawatt-scale loads. | Requires 30-day resource adequacy reports from each RTO/ISO explaining how adequate generation will be available to serve existing customers and new large loads; elevates flexible service, co-location rules, operational visibility, and generation adequacy into national market-design issues. | | NERC Electric Reliability Organization | ERO jurisdiction; mandatory reliability standards for the bulk power system. | 2024 and 2025 Long-Term Reliability Assessments; Large Load Task Force; Level 3 alert on unexpected data center load-loss events in May 2026; expanding attention to large dynamic digital loads, ride-through behavior, protection settings, and disturbance response. | No direct retail or transmission-cost allocation authority, but reserve-margin requirements, reliability standards, and operational requirements can influence capacity needs, infrastructure investment, and customer-service obligations. | Data Center Load Information Survey deployed with PJM; advanced analytics and data collection on load volatility, tripping risk, and data center operating profiles. | Focuses on unexpected large-load loss, frequency and voltage stability, transient response, telemetry, ride-through performance, and the operational risks created when multiple large digital loads respond similarly to grid disturbances. | | PJM Regional transmission organization | FERC-regulated wholesale market and transmission tariff authority across a thirteen-state region and the District of Columbia. | PJM co-location proceedings became the central early test case for large-load governance; Critical Issue Fast Path for Large Load Additions; Non-Capacity Backed Load service concept; expedited process for shovel-ready generation; stricter data center vetting in January 2026 load forecast; Google/Tapestry collaboration on AI-assisted interconnection studies. | Independent market monitor identified $21.3 billion in data center-related capacity costs across three auctions; ongoing debate over cost causation, capacity-market treatment, price caps, co-location charges, and whether speculative load should impose costs on existing customers. | Revised 2028 peak forecast downward by 4.4 GW using stricter data center vetting while increasing the long-term annual growth rate to 3.6 percent; roughly 30 GW of projected 2024–2030 load growth attributed to data centers. | December 2025 capacity auction fell 6,625 MW below the 20 percent reserve-margin target; reliability backstop mechanism under reform; NCBL concept would require curtailability during declared emergencies; PJM must now respond to FERC's national large-load orders. | | MISO Independent system operator | FERC-regulated wholesale market and transmission operator across a fifteen-state Midwest and South footprint. | Expedited Resource Addition Study process with two fast-track interconnection cycles in 2025; 6.1 GW second cycle in December 2025; expects 8–14 GW of new data center load in 2026–2027; now subject to FERC's June 2026 show-cause requirement for large-load tariff adequacy. | Load-responsible entities bear upgrade costs under tariff; ERAS caps additions at LRE resource adequacy deficits, reducing speculative queue risk and linking expedited additions to demonstrated reliability need. | Peak forecast rises from 121 GW in 2025 to 163 GW in 2035; data centers projected to reach as much as 25 percent of demand by 2040; large-load forecasting increasingly tied to resource adequacy and transmission deliverability. | ERAS targets resource adequacy gaps; second cycle led by battery storage and gas; FERC-approved complement to the normal queue; June 2026 FERC orders require MISO to explain how supply will meet both existing and new large-load demand. | | SPP Regional transmission organization | FERC-regulated wholesale market and transmission operator across a fourteen-state central U.S. region. | Expedited Resource Addition Study filed in May 2025; FERC-approved Priority Process in November 2025 for capacity uprates at existing facilities of up to 20 percent of output; early model for pairing large-load growth with accelerated supply additions. | Load-responsible-entity-based cost allocation; Priority Process limits cost exposure by focusing on incremental capability at existing resources and resource adequacy needs. | Annual load rose roughly 21 percent from 2020 to 2024; large-load and industrial demand growth increasingly shape SPP's transmission planning, interconnection priorities, and resource adequacy outlook. | Resource adequacy concerns prompted expedited processes; coordination with load-responsible entities; June 2026 FERC orders require SPP to address generation adequacy, study processes, and large-load tariff treatment as part of the national proceeding. | | ERCOT Texas ISO | Intrastate market structure outside most FERC jurisdiction; overseen by the Public Utility Commission of Texas. | Fast-track large-load interconnection; SB 6 demand-management programs; mandatory curtailment protocols for new large loads of 75 MW or more; voluntary Large Load Demand Management Service; Real-Time Co-optimization plus Batteries launched in December 2025; transmission and distribution pipeline expanded to more than 2,400 miles. | Market-based structure combined with mandatory curtailment obligations and competitive procurement of demand reductions; large-load customers increasingly expected to contribute flexibility or carry service obligations in exchange for faster access. | Peak demand forecast rises from 98 GW in 2026 to more than 111 GW by 2032; data centers and cryptocurrency loads heavily influence forecasts; PUCT data collection underway. | Mandatory large-load curtailment during firm load-shed events; voluntary demand-management procurement; RTC+B improves operational dispatch; Texas provides a parallel model for large-load governance outside FERC's organized-market jurisdiction. | | Virginia SCC State public utility commission | State retail regulation; Dominion Energy rate-setting; investor-owned utility oversight. | GS-5 rate class approved in November 2025, effective January 2027, for customers of 25 MW or more with at least 75 percent load factor; 14-year contracts; minimum demand charges; SB 253 proposes further assignment of energy costs to data centers. | 85 percent minimum demand charge for transmission and distribution; 60 percent minimum for generation; 14-year contracts; collateral; exit fees; JLARC found data centers could add $444 per year to a typical Dominion residential bill by 2040 without reform. | JLARC December 2024 demand analysis; SCC directed biennial GS-5 review beginning in 2027; tariff design increasingly used to separate verified projects from speculative inquiries. | Collateral, exit fees, and long-term commitments reduce stranded-asset risk; $7.6 billion Dominion transmission expansion under review; Virginia remains the leading state-level test case for aligning data center growth, ratepayer protection, and grid reliability. | | Ohio PUC / AEP Ohio State-utility tariff model | State PUC regulation of AEP Ohio tariff; FERC authority for wholesale and transmission matters. | Large-load tariff with demand charges equal to 85 percent of the 11-month rolling average demand; 8-year minimum contract; up to $100,000 load study fee; competing settlement proposals pending before the PUC. | First-year costs of roughly $10 million for a 100 MW facility; connection requests reduced by about 50 percent; AEP subsidiaries in Indiana, West Virginia, and Kentucky absorbed 750 MW for data centers that did not materialize and are now seeking to sell that capacity into PJM. | Tariff functions as a forecasting filter by requiring financial commitment before speculative projects enter utility and regional planning assumptions. | De-risking mechanism deters phantom projects, reduces stranded-asset exposure, and helps utilities align infrastructure commitments with credible customer demand. | | IOUs Industry-wide | State PUC-regulated retail service; FERC authority for wholesale markets and interstate transmission. | At least 36 utilities adopted large-load tariffs by late 2025; DCFlex initiative with EPRI, Google, Meta, Microsoft, Duke, PJM, and others; growing use of staged energization, dedicated substations, customer-funded upgrades, co-location agreements, and behind-the-meter generation strategies. | Average large-load tariff terms include approximately 80 percent minimum bills, 14-year contracts, 7-year security deposits, and 5-year exit fees; utilities increasingly seek direct funding, collateral, and take-or-pay commitments for infrastructure built to serve large loads. | IRPs revised to incorporate AI load scenarios; bottom-up project pipelines increasingly combined with milestone-based verification; LBL finds $300–400 million in NPV savings from integrating data center flexibility into IRP; Nicholas Institute estimates 76–126 GW of flexible load absorption potential. | UPS and batteries explored as grid resources through DCFlex; SMR co-location pipeline; onsite gas generation raises reliability and emissions questions; load curtailment, staged service, telemetry, and operational visibility are becoming core elements of utility large-load governance. | ## CONCLUSION The regulatory landscape mapped here is no longer just a collection of improvised responses. It is beginning to harden into an architecture. FERC’s June 2026 Section 206 orders did not solve the large-load problem, but they changed its institutional status. What had been treated as a series of local utility disputes, state tariff proceedings, PJM co-location fights, and reliability warnings is now formally a national market-design question. Every major organized market must explain whether its tariffs can accommodate AI-scale loads, how those loads will be studied, who will pay for the infrastructure they require, how flexible service will be defined, and where the generation needed to serve them will come from. That is progress. It is not yet coherence. Three tensions remain unresolved. The first is jurisdictional. The federal-state divide in electricity regulation was built for a world in which interstate transmission and wholesale markets were largely federal matters, while retail service, distribution, and utility cost recovery were primarily state matters. AI data centers blur that line. They are retail customers, but they can also reshape transmission plans, capacity-market prices, co-location arrangements, generator-interconnection priorities, resource adequacy obligations, and reliability standards. FERC’s June 2026 orders mark the clearest federal effort yet to define the transmission-system consequences of very large loads, but they do not eliminate state authority over retail tariffs, siting, or utility revenue recovery. Nor do they make RTOs, utilities, and state commissions see the same project through the same lens. A single data center campus may still be treated as a customer by one institution, a forecast input by another, a transmission driver by a third, and a reliability risk by a fourth. Until those roles are better aligned, uncertainty will persist over who studies, builds, pays for, verifies, and operates the infrastructure needed to serve AI-scale demand. The second tension is distributional. The public legitimacy of the AI infrastructure buildout will depend on whether ordinary customers believe they are being asked to finance the grid requirements of some of the largest technology companies in the world. PJM’s capacity-market data illustrate the scale of the issue: across three consecutive auctions, data center loads, both existing and forecast, accounted for 45 percent of total clearing costs, or $21.3 billion in aggregate charges. PJM serves roughly 67 million people, most of whom do not own or operate data centers. In Virginia, JLARC projected that data center growth could add $444 per year to a typical Dominion residential bill by 2040 absent structural reform. Large-load tariffs such as Virginia’s GS-5 class and Ohio’s data center tariff are early attempts to address that inequity. They are also incomplete. Their effectiveness will depend on contract terms, collateral requirements, exit fees, enforcement, and whether projected loads actually materialize. The third tension is evidentiary. Load forecasting for data centers is difficult in ways conventional industrial forecasting is not. A steel mill, once announced, is tied to a physical process, a local supply chain, and a specific site. A data center campus may be delayed, scaled back, relocated, or canceled depending on GPU supply chains, hyperscaler capital allocation, permitting constraints, water availability, power prices, fiber access, and the changing economics of AI training and inference. PJM’s January 2026 downward revision of its near-term forecast, removing 4.4 GW from the 2028 peak, showed both the scale of uncertainty and the direction of reform. Tighter vetting, collateral requirements, minimum demand charges, staged energization, withdrawal penalties, and milestone-based verification all point toward the same institutional move: load forecasting is becoming less a projection exercise and more a credibility test. Demand flexibility offers one of the most important bridges across these tensions, but it should not be mistaken for a substitute for infrastructure. If data centers can be designed and contracted to curtail, shift workloads, discharge storage, stage energization, or rely on dispatchable backup resources during defined grid conditions, they may become operational assets rather than fixed blocks of demand. The DCFlex initiative, ERCOT’s large-load curtailment framework, PJM’s Non-Capacity Backed Load concept, and FERC’s new focus on flexible transmission service all point in this direction. But flexibility becomes a reliability resource only when it is measurable, enforceable, compensated appropriately, and integrated into planning and operations. A data center’s theoretical ability to reduce load during system stress has little value unless operators can see it, model it, dispatch it, and rely on it. The broader conclusion is that U.S. electricity institutions are moving from managing the data center surge toward governing it. Managing means reacting to price spikes, reliability alerts, disputed forecasts, tariff filings, co-location disputes, and queue pressure. Governing means establishing rules in advance: which loads count, when they count, what evidence they must provide, who pays for the infrastructure they trigger, what obligations attach to different forms of service, what operational information system operators receive, and how federal and state institutions coordinate. The emerging regulatory test is credible deliverability. Can power be delivered to a specific place, on a specific timeline, under a credible commercial structure, with the necessary generation, transmission, substations, transformers, reliability services, customer commitments, flexibility arrangements, and cost-allocation rules in place? That question is becoming the bridge between AI’s infrastructure velocity and the grid’s institutional machinery. The path forward is neither unconditional accommodation nor reflexive resistance. AI infrastructure is economically and strategically important. Data centers can become constructive grid participants if their obligations are defined clearly. But the public grid cannot be planned around opaque, speculative, or cost-shifting demand. The emerging bargain should be straightforward: faster and more certain service for verified loads; cost responsibility for infrastructure caused; operational credit for measurable flexibility; and transparency sufficient to protect reliability and customers. The data center boom is therefore more than a demand-growth event. It is a test of whether U.S. electricity regulation can adapt to a new class of programmable, mobile, infrastructure-scale load. The jurisdictions that succeed will not be those that simply say yes or no to data centers. They will be those that learn how to verify load, price risk, allocate costs, require useful flexibility, and coordinate across the institutional boundaries that AI has already begun to blur. ## Sign up for AIxEnergy AIxEnergy provides independent intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, reliability, investment, and policy. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## REFERENCES American Action Forum (AAF). (2026, January 8). *Virginia’s new data center electricity rate class*. [https://www.americanactionforum.org/insight/virginias-new-data-center-electricity-rate-class/](https://www.americanactionforum.org/insight/virginias-new-data-center-electricity-rate-class/?ref=aixenergy.io) Baker Botts. (2025, December 19). *FERC issues order providing guidance for “co-locating” power plants with data centers within PJM*. [https://www.bakerbotts.com/thought-leadership/publications/2025/december/ferc-issues-order-providing-guidance-for-co-locating-power-plants-with-data-centers-within-pjm](https://www.bakerbotts.com/thought-leadership/publications/2025/december/ferc-issues-order-providing-guidance-for-co-locating-power-plants-with-data-centers-within-pjm?ref=aixenergy.io) Belfer Center for Science and International Affairs, Harvard Kennedy School. (2026, February 10). *AI, data centers, and the U.S. electric grid: A watershed moment*. [https://www.belfercenter.org/research-analysis/ai-data-centers-us-electric-grid](https://www.belfercenter.org/research-analysis/ai-data-centers-us-electric-grid?ref=aixenergy.io) Citizens Utility Board (CUB). (2025, December 17). *CUB statement: Capacity market prices reflect need for data center reform*. [https://www.citizensutilityboard.org/blog/2025/12/17/cub-statement-capacity-market-prices-reflect-need-for-data-center-reform/](https://www.citizensutilityboard.org/blog/2025/12/17/cub-statement-capacity-market-prices-reflect-need-for-data-center-reform/?ref=aixenergy.io) Data Center Dynamics. (2026, April 22). *ERCOT and MISO forecast huge increases in peak load, driven by data center demand*. [https://www.datacenterdynamics.com/en/news/ercot-and-miso-forecast-huge-increases-in-peak-load-driven-by-data-center-demand/](https://www.datacenterdynamics.com/en/news/ercot-and-miso-forecast-huge-increases-in-peak-load-driven-by-data-center-demand/?ref=aixenergy.io) Day Pitney. (2025, December 19). *Alert: Federal Energy Regulatory Commission orders major PJM tariff reforms for co-located load and behind-the-meter generation*. [https://www.daypitney.com/alert-federal-energy-regulatory-commission-orders-major-pjm-tariff-reforms](https://www.daypitney.com/alert-federal-energy-regulatory-commission-orders-major-pjm-tariff-reforms?ref=aixenergy.io) Duncan, Weinberg, Genzer & Pembroke, P.C. (DWGP). (2025, December 19). *MISO and SPP move forward with fast-track interconnection study processes*. [https://dwgp.com/firm-announcements/miso-and-spp-move-forward-with-fast-track-interconnection-study-processes](https://dwgp.com/firm-announcements/miso-and-spp-move-forward-with-fast-track-interconnection-study-processes?ref=aixenergy.io) Engineering News-Record (ENR). (2025, December 15). *Energy sector debates new US rules to power up more “large load” data centers*. [https://www.enr.com/articles/62219-energy-sector-debates-new-us-rules-to-power-up-more-large-load-data-centers](https://www.enr.com/articles/62219-energy-sector-debates-new-us-rules-to-power-up-more-large-load-data-centers?ref=aixenergy.io) EUCI. (2025, December 16). *Utilities adopt large load tariffs to cope with the costs and power demands of data centers*. [https://www.euci.com/utilities-adopt-large-load-tariffs-to-cope-with-the-costs-and-power-demands-of-data-centers/](https://www.euci.com/utilities-adopt-large-load-tariffs-to-cope-with-the-costs-and-power-demands-of-data-centers/?ref=aixenergy.io) Federal Energy Regulatory Commission (FERC). (2025a, February 20). *Order on large loads co-located at generating facilities* \[Docket Nos. EL25-49-000, AD24-11-000, and EL25-20-000\]. Federal Energy Regulatory Commission. Federal Energy Regulatory Commission (FERC). (2025b, October 27). *Notice inviting comments: Interconnection of large loads to the interstate transmission system* \[Docket No. RM26-4-000\]. Federal Energy Regulatory Commission. Federal Energy Regulatory Commission (FERC). (2025c, December 18). *Order directing PJM tariff revisions for co-located loads and behind-the-meter generation* \[Docket Nos. EL25-49-000, EL25-49-001, AD24-11-000, and EL25-20-000\]. Federal Energy Regulatory Commission. [https://www.ferc.gov/news-events/news/ferc-directs-nations-largest-grid-operator-create-new-rules-embrace-innovation-and](https://www.ferc.gov/news-events/news/ferc-directs-nations-largest-grid-operator-create-new-rules-embrace-innovation-and?ref=aixenergy.io) Federal Energy Regulatory Commission (FERC). (2026, March 19). *2025 state of the markets report*. Federal Energy Regulatory Commission. [https://www.utilitydive.com/news/data-centers-miso-ferc-market-report/815831/](https://www.utilitydive.com/news/data-centers-miso-ferc-market-report/815831/?ref=aixenergy.io) Information Technology and Innovation Foundation (ITIF). (2025, November 24). *The United States needs data centers, and data centers need energy, but that is not necessarily a problem*. [https://itif.org/publications/2025/11/24/united-states-needs-data-centers-data-centers-need-energy-but-that-is-not-necessarily-a-problem/](https://itif.org/publications/2025/11/24/united-states-needs-data-centers-data-centers-need-energy-but-that-is-not-necessarily-a-problem/?ref=aixenergy.io) Inside Climate News. (2026, January 7). *Virginia regulators approve new Dominion rates, assign more costs to data centers*. [https://insideclimatenews.org/news/07012026/virginia-regulators-approve-new-dominion-rates/](https://insideclimatenews.org/news/07012026/virginia-regulators-approve-new-dominion-rates/?ref=aixenergy.io) International Energy Agency (IEA). (2026, April 16). *Data centre electricity use surged in 2025, even with tightening bottlenecks driving a scramble for solutions*. [https://www.iea.org/news/data-centre-electricity-use-surged-in-2025-even-with-tightening-bottlenecks-driving-a-scramble-for-solutions](https://www.iea.org/news/data-centre-electricity-use-surged-in-2025-even-with-tightening-bottlenecks-driving-a-scramble-for-solutions?ref=aixenergy.io) Introl. (2026, February 6). *PJM’s 6 GW capacity shortfall*. [https://introl.com/blog/pjm-grid-crisis-2027-data-center-capacity-shortfall](https://introl.com/blog/pjm-grid-crisis-2027-data-center-capacity-shortfall?ref=aixenergy.io) Joint Legislative Audit and Review Commission (JLARC). (2024, December). *Data centers in Virginia*. Commonwealth of Virginia. [https://jlarc.virginia.gov/landing-2024-data-centers-in-virginia.asp](https://jlarc.virginia.gov/landing-2024-data-centers-in-virginia.asp?ref=aixenergy.io) K&L Gates. (2025, June 6). *Regional grid operators attempt to tackle resource adequacy by fast-tracking generator interconnection*. [https://www.klgates.com/Regional-Grid-Operators-Attempt-to-Tackle-Resource-Adequacy-by-Fast-Tracking-Generator-Interconnection-6-6-2025](https://www.klgates.com/Regional-Grid-Operators-Attempt-to-Tackle-Resource-Adequacy-by-Fast-Tracking-Generator-Interconnection-6-6-2025?ref=aixenergy.io) Knowledge Problem. (2026). *Data centers, flexibility, and the architecture of the grid*. Substack. Kwon, K.-B., Mukherjee, S., & Adetola, V. (2025). *Operational risks in grid integration of large data center loads: Characteristics, stability assessments, and sensitivity studies*. Pacific Northwest National Laboratory. arXiv:2510.05437. Mayer Brown. (2025, November 26). *FERC large load interconnection preliminary rulemaking: Key takeaways for data center developers, other large load projects, and investors*. [https://www.mayerbrown.com/en/insights/publications/2025/11/ferc-large-load-interconnection-preliminary-rulemaking](https://www.mayerbrown.com/en/insights/publications/2025/11/ferc-large-load-interconnection-preliminary-rulemaking?ref=aixenergy.io) Microgrid Knowledge. (2026). *NERC warns of 2028 capacity shortfalls amid exploding AI and data center loads*. [https://www.microgridknowledge.com/grid/article/55355943/nerc-warns-of-2028-capacity-shortfalls-amid-exploding-ai-and-data-center-loads](https://www.microgridknowledge.com/grid/article/55355943/nerc-warns-of-2028-capacity-shortfalls-amid-exploding-ai-and-data-center-loads?ref=aixenergy.io) Monitoring Analytics. (2026, January 5). *Analysis of the 2027/2028 RPM base residual auction: Part A*. Monitoring Analytics, LLC. National Association of Regulatory Utility Commissioners (NARUC). (2025). *Load growth resources for regulators*. [https://www.naruc.org/core-sectors/electricity-energy/load-growth/](https://www.naruc.org/core-sectors/electricity-energy/load-growth/?ref=aixenergy.io) North American Electric Reliability Corporation (NERC). (2024, December). *2024 long-term reliability assessment* (Corrected July 2025). [https://www.nerc.com/globalassets/our-work/assessments/2024-ltra\_corrected\_july\_2025.pdf](https://www.nerc.com/globalassets/our-work/assessments/2024-ltra%5Fcorrected%5Fjuly%5F2025.pdf?ref=aixenergy.io) North American Electric Reliability Corporation (NERC). (2026, January 29). *2025 long-term reliability assessment*. [https://www.nerc.com/globalassets/our-work/assessments/nerc\_ltra\_2025.pdf](https://www.nerc.com/globalassets/our-work/assessments/nerc%5Fltra%5F2025.pdf?ref=aixenergy.io) Northern Virginia Technology Council (NVTC). (2026). *2026 data center report*. [https://www.nvtc.org/communities/data-center-and-cloud/report/](https://www.nvtc.org/communities/data-center-and-cloud/report/?ref=aixenergy.io) PJM Inside Lines. (2025, September 17). *PJM kicks off initiative to balance reliability with large load growth*. [https://insidelines.pjm.com/pjm-kicks-off-initiative-to-balance-reliability-with-large-load-growth/](https://insidelines.pjm.com/pjm-kicks-off-initiative-to-balance-reliability-with-large-load-growth/?ref=aixenergy.io) PJM Inside Lines. (2026, January). *PJM’s updated 20-year forecast continues to see significant long-term load growth*. [https://insidelines.pjm.com/pjms-updated-20-year-forecast-continues-to-see-significant-long-term-load-growth/](https://insidelines.pjm.com/pjms-updated-20-year-forecast-continues-to-see-significant-long-term-load-growth/?ref=aixenergy.io) PJM Interconnection. (2025a). *2025 long-term load forecast report*. PJM Interconnection LLC. [https://www.pjm.com/-/media/DotCom/library/reports-notices/load-forecast/2025-load-report.pdf](https://www.pjm.com/-/media/DotCom/library/reports-notices/load-forecast/2025-load-report.pdf?ref=aixenergy.io) PJM Interconnection. (2025b, December 17). *2027/2028 base residual auction results* \[Press release\]. [https://www.pjm.com/-/media/DotCom/about-pjm/newsroom/2025-releases/20251217-pjm-auction-procures-134479-mw-of-generation-resources.pdf](https://www.pjm.com/-/media/DotCom/about-pjm/newsroom/2025-releases/20251217-pjm-auction-procures-134479-mw-of-generation-resources.pdf?ref=aixenergy.io) PJM Interconnection. (2025c, October). *Critical issue fast path for large load additions: Problem statement*. [https://www.pjm.com/-/media/DotCom/committees-groups/cifp-lla/2025/20251014/20251014-informational-only---critical-issue-fast-path---large-load-additions---problem-statement---updated---clean.pdf](https://www.pjm.com/-/media/DotCom/committees-groups/cifp-lla/2025/20251014/20251014-informational-only---critical-issue-fast-path---large-load-additions---problem-statement---updated---clean.pdf?ref=aixenergy.io) S&P Global Market Intelligence. (2025, October 16). *Grid congestion remains key issue as data center load growth stresses system*. [https://www.spglobal.com/market-intelligence/en/news-insights/research/2025/10/grid-congestion-remains-key-issue-as-data-center-load-growth-stresses-system](https://www.spglobal.com/market-intelligence/en/news-insights/research/2025/10/grid-congestion-remains-key-issue-as-data-center-load-growth-stresses-system?ref=aixenergy.io) Shehabi, A., Smith, S. J., Hubbard, A., Newkirk, A., Lei, N., Siddik, M. A. B., Holecek, B., Koomey, J., Masanet, E., & Sartor, D. (2024). *2024 United States data center energy usage report* (LBNL-2001637). Lawrence Berkeley National Laboratory. [https://eta-publications.lbl.gov/sites/default/files/2024-12/lbnl-2024-united-states-data-center-energy-usage-report\_1.pdf](https://eta-publications.lbl.gov/sites/default/files/2024-12/lbnl-2024-united-states-data-center-energy-usage-report%5F1.pdf?ref=aixenergy.io) Sheng, Y., Zhang, C., Zhu, Z., Xu, H., Wen, J., Wang, R., Yang, J., Wang, Q., & Bu, S. (2026). Power for AI data centers: Energy demand, grid impacts, challenges and perspectives. *Energies, 19*(3), 722\. [https://doi.org/10.3390/en19030722](https://doi.org/10.3390/en19030722?ref=aixenergy.io) The Register. (2026, May 15). *Datacenters slurping juice help drive 75% jump in PJM power prices*. [https://www.theregister.com/on-prem/2026/05/15/datacenters-slurping-juice-help-drive-75-jump-in-pjm-power-prices/](https://www.theregister.com/on-prem/2026/05/15/datacenters-slurping-juice-help-drive-75-jump-in-pjm-power-prices/?ref=aixenergy.io) Utility Dive. (2025, December 18). *PJM capacity prices hit record high as grid operator falls short of reliability target*. [https://www.utilitydive.com/news/pjm-interconnection-capacity-auction-data-center/808264/](https://www.utilitydive.com/news/pjm-interconnection-capacity-auction-data-center/808264/?ref=aixenergy.io) Utility Dive. (2026, January 15). *PJM trims near-term load forecast on stricter data center vetting, economic outlook*. [https://www.utilitydive.com/news/pjm-interconnection-load-forecast-data-centers/809717/](https://www.utilitydive.com/news/pjm-interconnection-load-forecast-data-centers/809717/?ref=aixenergy.io) Utility Dive. (2026, January 30). *NERC forecasts peak demand to rise 24% on new data center loads*. [https://www.utilitydive.com/news/nerc-10-year-peak-demand-forecast-jumps-24-on-new-data-center-loads/810955/](https://www.utilitydive.com/news/nerc-10-year-peak-demand-forecast-jumps-24-on-new-data-center-loads/810955/?ref=aixenergy.io) Utility Dive. (2026). *Sudden data center load losses prompt NERC alert, recommendations*. [https://www.utilitydive.com/news/data-center-load-disruptions-nerc-alert-recommendations/818036/](https://www.utilitydive.com/news/data-center-load-disruptions-nerc-alert-recommendations/818036/?ref=aixenergy.io) Virginia Mercury. (2025, April 25). *Will special rate classes protect Va. residents from the costs of serving data centers?* [https://virginiamercury.com/2025/04/25/will-special-rate-classes-protect-va-residents-from-the-costs-of-serving-data-centers/](https://virginiamercury.com/2025/04/25/will-special-rate-classes-protect-va-residents-from-the-costs-of-serving-data-centers/?ref=aixenergy.io) Virginia Mercury. (2026, February 10). *Bill would put more energy costs on data centers, slash residential customers’ rates*. [https://virginiamercury.com/2026/02/10/bill-would-put-more-energy-costs-on-data-centers-slash-residential-customerss-rates/](https://virginiamercury.com/2026/02/10/bill-would-put-more-energy-costs-on-data-centers-slash-residential-customerss-rates/?ref=aixenergy.io) Virginia State Corporation Commission (SCC). (2025, November 25). *Final order: Dominion Energy Virginia biennial review 2025* \[Case No. PUR-2025-00058\]. [https://www.scc.virginia.gov/about-the-scc/newsreleases/release/scc-issues-order-on-dev-biennial-review-2025](https://www.scc.virginia.gov/about-the-scc/newsreleases/release/scc-issues-order-on-dev-biennial-review-2025?ref=aixenergy.io) White & Case LLP. (2025a, April 24). *NERC tees up plan to assess grid risks associated with data centers*. [https://www.whitecase.com/insight-alert/nerc-tees-plan-assess-grid-risks-associated-data-centers](https://www.whitecase.com/insight-alert/nerc-tees-plan-assess-grid-risks-associated-data-centers?ref=aixenergy.io) White & Case LLP. (2025b, October 27). *DOE directs FERC to accelerate interconnection of data centers*. [https://www.whitecase.com/insight-alert/doe-directs-ferc-accelerate-interconnection-data-centers](https://www.whitecase.com/insight-alert/doe-directs-ferc-accelerate-interconnection-data-centers?ref=aixenergy.io) Yes Energy. (2026, February 18). *How ISOs and RTOs are addressing large load growth in 2025*. [https://www.yesenergy.com/blog/how-isos-and-rtos-are-addressing-large-load-growth-in-2025](https://www.yesenergy.com/blog/how-isos-and-rtos-are-addressing-large-load-growth-in-2025?ref=aixenergy.io) ### FERC Just Fired the Starting Gun on the Next Phase of the AI Economy URL: https://www.aixenergy.io/ferc-just-fired-the-starting-gun-on-the-next-phase-of-the-ai-economy/ Last updated: 2026-07-29T19:39:28.000Z FERC’s June 2026 action reframes AI infrastructure as a grid-governance challenge: power access, flexible service, cost allocation, and resource adequacy now shape where the AI economy can grow. _This post is for paying subscribers only._ ### The AI-Energy Decision Stack URL: https://www.aixenergy.io/the-ai-energy-decision-stack-how-leaders-make-infrastructure-choices-before-the-evidence-is-complete/ Last updated: 2026-07-29T19:39:38.000Z Infrastructure planning once focused on forecasting demand. AI changes the challenge. The critical question is no longer how much demand will emerge, but which infrastructure commitments should be made before demand is fully known—and who bears the risk if assumptions prove wrong. _This post is for paying subscribers only._ ### IEA's World Investment Report Confirms Energy and AI Are Now the Same Story URL: https://www.aixenergy.io/ieas-world-investment-report-confirms-energy-and-ai-are-now-the-same-story/ Last updated: 2026-07-13T17:55:04.000Z Every year, the [International Energy Agency's *World Energy Investment*](https://iea.blob.core.windows.net/assets/4fda38df-523c-46f5-ae75-49481abdc8fc/WorldEnergyInvestment2026.pdf?ref=aixenergy.io) report serves as one of the most important scorecards in the global economy. Markets fluctuate, technologies rise and fall, governments change priorities, and headlines shift from one crisis to the next. Investment, however, reveals where governments, corporations, utilities, investors, and developers are actually placing their bets. Capital allocation is ultimately the most honest expression of belief. Article Summary 0:00 /49.715601 1× [WorldEnergyInvestment2026WorldEnergyInvestment2026.pdf34 MBdownload-circle](https://www.aixenergy.io/content/files/2026/06/WorldEnergyInvestment2026.pdf "Download") The 2026 edition arrives at a particularly consequential moment. Global energy investment is expected to reach approximately $3.4 trillion this year, representing a 5% increase despite geopolitical instability, elevated interest rates, persistent supply-chain constraints, and growing uncertainty across global markets. Roughly $2.2 trillion will flow toward renewables, nuclear power, grids, storage, electrification, efficiency, and other low-emissions technologies, while approximately $1.2 trillion will be invested in oil, natural gas, and coal. AIxEnergy Briefing Box ## Top 5 Takeaways from the IEA’s *World Energy Investment 2026* Report 1. 1 ### The Age of Electricity Has Arrived Nearly 60% of global energy investment now flows into electricity generation, grids, storage, and electrification. The energy system is no longer organized around fuels; it is increasingly organized around electrons. 2. 2 ### Energy Security Is Driving Investment Decisions The strongest case for electrification is no longer just climate. It is resilience, affordability, and reduced exposure to volatile fuel imports. The IEA estimates that renewables, efficiency, electrification, and nuclear avoided approximately $260 billion in fossil fuel imports in 2025 alone. 3. 3 ### Artificial Intelligence Has Become an Energy Industry Data-center-related energy investment exceeded $100 billion globally in 2025, while gas turbine orders reached a 25-year high. AI is no longer simply creating electricity demand; it is reshaping generation, transmission, storage, and utility planning. 4. 4 ### The Grid Is Now the Critical Bottleneck The challenge is no longer producing electricity. It is moving it, storing it, and integrating it. Transmission, transformers, interconnection queues, permitting, and workforce constraints are becoming limiting factors for economic growth and AI deployment. 5. 5 ### The Future Belongs to Infrastructure Intelligence The biggest trend in the report is convergence. Energy systems are merging with digital systems. The rise of AI-driven infrastructure, the emergence of the Shadow Grid, and the need for a Cognitive Grid point toward a future where power and computation become increasingly inseparable. Those figures alone would make the report noteworthy. Yet the real significance of the report lies not in the magnitude of investment, but in what the pattern of investment reveals about the future. For much of the last decade, discussions about energy revolved around climate change, decarbonization targets, renewable deployment, and emissions trajectories. Those themes remain important, but the 2026 report documents a broader transformation: the global economy is increasingly reorganizing itself around electricity, infrastructure resilience, digital systems, and artificial intelligence. The transition now underway is not simply a shift from fossil fuels to clean energy. It is a shift from an economy organized around fuels to one organized around infrastructure. Viewed through this lens, the report becomes much more than a catalog of investment statistics. It becomes an early map of a new economic architecture in which electricity, computation, and infrastructure increasingly converge. ## The Arrival of the Age of Electricity The most important conclusion in the report is surprisingly easy to miss. Electricity-related investments now account for nearly 60% of all global energy investment. Spending on electricity generation, transmission, distribution, storage, and related infrastructure has reached approximately $1.6 trillion annually and approaches $2 trillion when end-use electrification is included. This statistic represents a fundamental restructuring of the global energy system. For most of modern history, energy systems were organized around fuels. Coal powered industry. Oil powered transportation. Natural gas heated buildings and fueled industrial processes. Electricity served as a delivery mechanism, but fuels remained the foundation of economic activity. That relationship is reversing. Transportation is increasingly electrified. Buildings are increasingly electrified. Industrial processes are increasingly electrified. Data centers are entirely dependent upon electricity. Future fuels such as hydrogen derive their value from electricity. Even energy security strategies increasingly revolve around expanding electrical infrastructure rather than securing access to imported fuels. The report's investment data demonstrates that this shift is no longer theoretical. Renewable power investment now reaches approximately $665 billion annually. Solar alone accounts for roughly $365 billion, equivalent to nearly one billion dollars invested every day. Wind investment remains substantial at roughly $200 billion annually, while hydropower contributes approximately $75 billion. Together, renewable technologies account for roughly 70% of global power-sector investment. At the same time, nuclear energy is experiencing a resurgence that would have seemed improbable only a few years ago. Global nuclear investment now exceeds $80 billion annually, while seventy-eight gigawatts of capacity are under construction across fifteen countries. More than forty nations now maintain active policies supporting new nuclear development. The significance of these numbers extends beyond individual technologies. Collectively, they reveal that electricity has become the dominant destination for global energy capital. The world's largest economies are increasingly investing not only in fuels themselves, but in the infrastructure required to generate, move, store, and manage electricity. This distinction matters because it changes the nature of competition. In the twentieth century, economic power often depended upon access to fuel resources. In the twenty-first century, economic power may increasingly depend upon access to electrical infrastructure. ## Energy Security Has Become a Dominant Investment Driver One of the most striking themes throughout the report is the reemergence of energy security as a dominant force shaping investment decisions alongside cost, climate, reliability, and affordability. The IEA devotes considerable attention to the geopolitical disruptions that have unfolded during the past year. Damage to energy facilities, attacks on shipping infrastructure, uncertainty surrounding key maritime corridors, and concerns about future supply disruptions have materially altered how governments and investors evaluate risk. The result has been a shift in emphasis. For much of the previous decade, energy investment decisions were frequently framed through the lens of carbon reduction. Increasingly, they are also being framed through the lens of resilience, reliability, and strategic autonomy. This change helps explain several trends that might otherwise appear contradictory. Natural gas investment is projected to reach approximately $330 billion in 2026, the highest level in a decade. Coal investment is expected to rise to roughly $180 billion, the highest level since 2012\. Nuclear investment continues to accelerate. Grid investment is expanding rapidly. Domestic manufacturing of energy technologies is becoming a strategic objective across multiple regions. Viewed through a narrow climate lens, these developments can appear inconsistent. Viewed through a security lens, they become more understandable. Countries are not simply abandoning decarbonization. They are expanding their definition of energy security and, in many cases, attempting to balance emissions goals against resilience, affordability, and exposure to geopolitical risk. The report highlights one of the strongest examples of this shift. Investments in renewables, electrification, efficiency, and nuclear energy avoided approximately $260 billion in fossil fuel import costs across major regions during 2025\. China alone accounted for roughly $110 billion of those savings. This finding may ultimately prove more important than many emissions-related statistics. The strongest argument for electrification is increasingly economic and strategic as well as environmental. Countries that generate more energy domestically through electricity reduce exposure to volatile fuel markets, geopolitical disruptions, and import dependencies. The energy transition is therefore becoming inseparable from national security policy. ## Artificial Intelligence Has Entered the Energy Investment Equation No section of the report better illustrates the changing nature of energy demand than the discussion of artificial intelligence and data centers. Only a few years ago, AI barely registered within most energy forecasts. Today it has become a meaningful driver of investment decisions across multiple segments of the energy sector. The report estimates that energy-sector investment associated with data-center development exceeded $100 billion globally during 2025\. To put that figure in perspective, it exceeds total energy investment across Africa. The implications extend far beyond electricity demand forecasts. The IEA documents a surge in gas-fired generation orders to approximately 130 GW during 2025, the highest level in twenty-five years. A substantial portion of this increase is directly linked to growing data-center demand. Nearly $28 billion in gas turbine orders are associated with onsite generation serving computational infrastructure. These developments reveal an important reality that many energy discussions still fail to recognize. Data centers are no longer simply customers of the grid. They are becoming infrastructure actors. Technology companies increasingly influence generation development, transmission planning, storage deployment, advanced nuclear investment, geothermal commercialization, and utility resource decisions. The report notes that technology companies now account for approximately 40% of corporate power purchase agreements globally, underscoring the degree to which computational infrastructure is reshaping energy markets. Historically, economic growth drove energy demand. Increasingly, energy availability may determine where digital economic growth occurs. This inversion represents one of the most important developments identified in the report. ## The Grid Has Become a Binding Constraint For decades, energy debates focused primarily on generation technologies. Policymakers argued about the optimal mix of coal, gas, nuclear, wind, and solar. Analysts built scenarios around competing generation portfolios. Investors sought opportunities within specific technology categories. The report suggests that the central challenge has shifted. Generation is no longer the only binding constraint; integration, interconnection, grid equipment, permitting, and system operation are becoming equally decisive. Global grid investment now approaches $550 billion annually. Battery storage investment exceeds $100 billion. Both categories are growing rapidly because the constraints facing modern energy systems increasingly reside outside generation itself. Transmission corridors have become scarce. Transformers have become strategic assets. Interconnection queues have become barriers to investment. Permitting timelines have become economic bottlenecks. Workforce shortages have become deployment constraints. The challenge facing energy systems is no longer merely producing electrons. It is transporting them, storing them, managing them, and integrating them into increasingly complex networks. This shift carries profound implications for investors and policymakers. Regions that can accelerate transmission deployment, streamline permitting, modernize planning processes, and expand grid infrastructure will likely capture disproportionate shares of future economic growth. Regions that fail to address these constraints risk becoming increasingly unattractive locations for industrial development and computational infrastructure. ## Capital Markets Will Determine the Pace of the Transition Energy transitions are often described as technological transformations. In practice, they are capital allocation transformations. One of the report's most important findings is that financing conditions increasingly influence deployment outcomes. Interest rates, risk premiums, financing costs, and policy certainty often matter as much as technology costs. This reality is particularly important because the emerging energy system is fundamentally capital intensive. Renewable generation requires substantial upfront investment. Transmission projects require enormous capital commitments years before revenue generation. Nuclear projects require long-term financial certainty. Battery storage requires significant initial expenditures. Artificial intelligence infrastructure follows a similar pattern. Data centers, advanced semiconductors, dedicated generation assets, cooling systems, and transmission upgrades all require large upfront investments supported by expectations of future growth. As a result, the AI economy and the energy economy increasingly share common financial characteristics. Both depend upon long-duration capital, predictable regulatory environments, and confidence in future demand. The report highlights another important challenge. Financing costs remain dramatically higher across many developing economies than in advanced economies and China. This disparity represents one of the most significant barriers to global energy development. While technology costs continue to decline, financing costs frequently determine whether projects proceed at all. The countries that succeed in attracting affordable capital may ultimately gain a larger advantage than those that merely possess superior technology. ## Critical Minerals and the New Resource Competition The report also highlights a growing tension within the global energy transition. For much of modern history, energy security centered on access to fuels. The emerging energy system increasingly depends upon access to materials. Copper, lithium, nickel, graphite, rare earth elements, semiconductors, advanced steel, and power electronics now occupy a strategic position within global supply chains. The report notes that investment in critical minerals declined during 2025 after several years of rapid growth. Lithium investment experienced particularly sharp declines following price corrections. Copper investment, however, continued to rise. This distinction is important. Copper sits at the center of nearly every major trend identified in the report. Electrification requires copper. Data centers require copper. Transmission expansion requires copper. Electric vehicles require copper. Renewable energy deployment requires copper. The Age of Electricity is, in many respects, also the Age of Copper. The report further highlights China's extraordinary position within global manufacturing supply chains. China accounts for approximately 75% of clean-energy manufacturing investment, roughly 80% of lithium-ion battery production supply-chain capacity, and approximately 95% of photovoltaic wafer production capacity. These figures reflect more than industrial success. They represent a strategic concentration of manufacturing capability that will influence energy markets, supply chains, and geopolitical relationships for years to come. ## Innovation Remains the Ultimate Force Multiplier Perhaps the most encouraging finding in the report concerns innovation. The energy transition is often portrayed as a story of deployment. The data suggest it is equally a story of technological progress. The report notes that without the cost reductions achieved through innovation over the last decade, delivering the same amount of energy additions expected in 2026 would require nearly twice as much investment. Solar costs have fallen dramatically. Battery costs have fallen dramatically. Power electronics have improved dramatically. Operational efficiency has increased dramatically. Innovation has quietly become one of the most important drivers of energy investment. This lesson matters because the next decade will likely depend upon continued innovation. Artificial intelligence, advanced nuclear systems, enhanced geothermal technologies, next-generation storage, advanced materials, and digital grid technologies all have the potential to reshape investment patterns. The report also highlights growing public and private investment in energy-related research and development. China now accounts for more than one-third of public energy R&D spending globally and more than 40% of corporate energy-related R&D. Innovation leadership increasingly influences industrial leadership. The countries that develop technologies often become the countries that manufacture technologies. The countries that manufacture technologies frequently become the countries that dominate supply chains. ## The Emergence of the Shadow Grid The report does not use the term Shadow Grid. That is AIxEnergy's analytical lens for describing one of the most important developments implied by the data. The Shadow Grid consists of infrastructure developed outside traditional utility planning frameworks to support mission-critical loads. This includes dedicated generation, microgrids, private transmission assets, behind-the-meter energy systems, dedicated renewable portfolios, long-duration storage facilities, and potentially small modular reactors developed specifically to serve large computational and industrial loads. The rise of onsite generation supporting data centers represents one of the clearest early examples of this phenomenon. The IEA's data on captive data-center generation and gas turbine orders provide a strong empirical signal that some customers are no longer waiting for traditional infrastructure processes to catch up with demand. Historically, customers adapted to available infrastructure. Increasingly, infrastructure is being developed around specific customers. When interconnection timelines become too long, alternatives emerge. When transmission expansion cannot keep pace with demand, alternatives emerge. When reliability becomes uncertain, alternatives emerge. The Shadow Grid should not be viewed as a temporary response. It represents a structural adaptation to the growing mismatch between infrastructure development timelines and digital economic growth. ## The Rise of the Cognitive Grid The report also does not use the term Cognitive Grid. That is AIxEnergy's interpretive framework for understanding the intelligence layer required to operate the kind of electricity system the IEA describes. The report focuses primarily on physical assets, yet the future system it describes will require far more than generation, transmission, and storage. It will require intelligence. The twentieth-century grid was mechanical. The late twentieth-century grid became digital. The twenty-first-century grid is becoming cognitive. Managing millions of distributed assets, increasingly variable generation portfolios, growing electrification loads, electric vehicles, batteries, microgrids, and data centers exceeds the capabilities of traditional operating models. Artificial intelligence increasingly becomes a practical necessity rather than a novelty. Forecasting, planning, asset management, maintenance optimization, outage response, market operations, and system balancing will all depend upon advanced computational capabilities. The same technologies driving electricity demand are increasingly becoming essential for operating electricity systems. Electricity powers computation, and computation increasingly optimizes electricity. This feedback loop may become one of the defining characteristics of the next phase of economic development. ## From the Age of Fuels to the Age of Infrastructure Intelligence The IEA's *World Energy Investment 2026* report will be remembered for its headline figures: $3.4 trillion in annual investment, nearly $2.2 trillion flowing toward clean energy and electrification, unprecedented investment in grids and storage, renewed momentum for nuclear energy, and the emergence of artificial intelligence as a material force in energy markets. Yet the deeper significance of the report lies beneath the numbers. It documents the emergence of a world increasingly organized around electricity rather than fuels, infrastructure rather than commodities, resilience rather than efficiency alone, and intelligence rather than hardware alone. The most important investment trend is not solar, nuclear, batteries, gas, or even data centers. The most important trend is convergence. Energy systems are converging with digital systems. Electricity systems are converging with computational systems. Infrastructure investment is converging with artificial intelligence investment. The result is the emergence of a new economic architecture in which power, data, and computation become increasingly inseparable. The report captures the early stages of that transformation. The Shadow Grid represents the physical adaptation to growing infrastructure constraints. The Cognitive Grid represents the intelligence layer required to manage growing infrastructure complexity. Together they provide a framework for understanding why traditional energy models are becoming increasingly inadequate. The next decade will not simply determine how the world generates electricity. It will determine how the world organizes economic activity, technological leadership, and national competitiveness. The capital flows documented in this year's report suggest that this transformation is already underway. The Age of Electricity is not a forecast. The investment decisions have already been made. What remains uncertain is which countries, companies, utilities, and investors will recognize soon enough that the future of energy and the future of artificial intelligence are becoming the same story. ## Sign up for AIxEnergy AIxEnergy provides independent intelligence on AI infrastructure, data centers, electricity demand, electric grids, utilities, power markets, reliability, investment, and policy. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### The Utility-AI Leadership Edge Part III: How CFOs Can Fund the AI-Driven Grid Buildout Without Leaving Customers Behind URL: https://www.aixenergy.io/the-utility-ai-leadership-edge-part-iii-how-cfos-can-fund-the-ai-driven-grid-buildout-without-leaving-customers-behind/ Last updated: 2026-07-13T01:39:17.000Z *This is Part III of The Utility-AI Leadership Edge, a four-part series on how utility leaders can use AI to improve resilience, reliability, and growth. Parts* [*I* ](https://www.aixenergy.io/the-utility-ai-leadership-edge-part-i-how-utility-coos-can-harness-ai-to-deliver-resilience-reliability-and-growth/)*and* [*II*](https://www.aixenergy.io/the-utility-cios-mandate-in-the-age-of-ai-managing-risk-while-unlocking-grid-scale-opportunity/) *covered the COO and CIO mandates; Part III turns to the CFO’s role in funding AI-driven grid growth while protecting affordability and fair cost allocation. Part IV will examine the CEO’s role in aligning the enterprise for transformation.* For much of the past century, the utility CFO’s job was relatively predictable. Demand growth was steady, capital planning followed well-established patterns, and rate cases were largely built around replacing aging infrastructure, maintaining reliability, and generating reasonable shareholder returns. The financial challenge was one of optimization rather than transformation. That era is ending, and the change is arriving faster than many utility planning cycles were designed to absorb. 0:00 /0:36 1× Today, utility CFOs are standing at the center of one of the largest infrastructure investment cycles in modern history. Electrification is accelerating, manufacturing is reshoring, artificial intelligence is driving an unprecedented wave of hyperscale data center development, distributed energy resources are reshaping grid architecture, and extreme weather events are increasing resilience requirements. At the same time, regulators, legislators, and customers remain intensely focused on affordability. The result is a fundamental shift in the CFO mandate: the challenge is no longer simply raising capital, but raising capital at unprecedented scale while ensuring that residential and small commercial customers do not become the unintended financiers of growth driven by a relatively small number of large-load customers. The utilities that succeed over the next decade will not be those that spend the most. They will be those that fund growth most intelligently, with a clear line of sight between capital deployed, customers served, risks allocated, and outcomes delivered. ## **The Scale of the Challenge Has Changed** Across North America, utilities are receiving load requests that would have been unimaginable just a few years ago. Single data center campuses now routinely seek hundreds of megawatts of capacity, and in some regions utilities are evaluating gigawatt-scale requests from clusters of hyperscale operators, AI computing facilities, advanced manufacturing plants, and electrification projects. Historically, large industrial loads represented meaningful but manageable additions to the system. Today’s requests often rival the demand of entire cities, and they arrive in concentrated geographies where the grid may already be constrained by transformer availability, interconnection backlogs, limited transmission capacity, workforce limitations, or local permitting friction. The traditional utility finance model assumes that infrastructure investments are broadly beneficial across the customer base over long periods of time. Large-load growth challenges that assumption because the costs can be immediate, location-specific, and driven by the requirements of one customer or a small group of customers. When a utility must invest hundreds of millions of dollars in substations, transmission upgrades, transformers, switchgear, and distribution infrastructure to serve a specific load pocket, the critical question is no longer theoretical. The question is who should pay, and the answer will increasingly define the quality of utility regulatory relationships over the next decade. The scale of the issue is now visible in national data. The U.S. Department of Energy reported that data centers consumed roughly 4.4 percent of total U.S. electricity in 2023 and could consume between 6.7 and 12 percent by 2028, while total data center electricity use rose from 58 TWh in 2014 to 176 TWh in 2023.¹ That growth is not occurring in a vacuum. The Energy Information Administration projects that U.S. electricity consumption will reach record highs in 2026 and 2027, with commercial electricity consumption surpassing residential consumption for the first time on record in 2026.² For CFOs, this means load growth is no longer simply a sales opportunity or a capital plan input. It is a commercial, operational, regulatory, and customer trust challenge that has to be managed with discipline from the first interconnection conversation. ## **The Affordability Imperative** Utility leaders understand that economic development matters. Data centers create jobs, attract investment, and support broader regional growth, while manufacturing expansion strengthens local economies and electrification supports public policy objectives and long-term sustainability goals. Yet residential customers are increasingly concerned about affordability. Electricity bills have become more visible, more political, and more contested at the same time that utilities are asking regulators to approve large capital plans for generation, transmission, distribution modernization, resilience, and grid hardening. Many regulators are asking difficult but necessary questions. Should residential customers absorb infrastructure costs required to serve large hyperscale facilities, should small businesses subsidize grid upgrades primarily benefiting a handful of large-load customers, and how should utilities allocate risk when future demand forecasts remain uncertain? These are not merely regulatory questions. They are financial strategy questions, and the CFO must become the architect of equitable cost allocation models that encourage growth while protecting customer trust. That trust is already being tested. The Wall Street Journal has reported on the growing dispute between utilities and technology companies over who should pay for the power infrastructure required by the data center buildout, while the Washington Post has highlighted concerns that residential customers are seeing bills rise even as large-load customers may benefit from negotiated structures or direct access to high-voltage infrastructure.³ The concern is not that utilities are building for growth. The concern is that customers may come to believe they are paying for someone else’s growth without receiving a clear, measurable, and fairly allocated benefit. This is where the CFO’s role becomes more than financial stewardship. It becomes customer stewardship, because affordability is not only a rate case issue; it is the foundation of permission to build. ## **The Growing Importance of CIAC** One of the most powerful tools available to utilities is Contributions in Aid of Construction, commonly known as CIAC. Historically, CIAC has been used to require customers to fund portions of infrastructure that primarily benefit their own service requirements. While not new, its importance is increasing dramatically as utilities evaluate massive load requests. In a world of gigawatt-scale interconnections, CIAC becomes a practical mechanism for balancing growth and fairness. Rather than placing all upgrade costs into rate base, utilities can require large customers to contribute toward dedicated infrastructure investments. This may include new substations, feeder expansions, transmission upgrades, switching equipment, interconnection facilities, and other assets required primarily to serve the customer’s load. Properly structured CIAC arrangements can achieve multiple objectives simultaneously. They protect residential customers from bearing disproportionate costs, create clearer economic signals for large-load developers, reduce stranded asset risk if projects fail to materialize, and improve regulatory support by demonstrating prudent stewardship of customer dollars. The most effective CFOs are no longer viewing CIAC as a transactional tool. They are treating it as a strategic instrument within broader infrastructure financing frameworks, especially where load requests are large, concentrated, uncertain, or tied to customers whose business models may evolve faster than utility assets depreciate. This matters because AI infrastructure forecasts remain highly uncertain. The International Energy Agency projects that global data center electricity consumption could roughly double by 2030 in its base case, but it also emphasizes that outcomes depend heavily on AI adoption, hardware efficiency, software efficiency, deployment constraints, and energy-sector bottlenecks.⁴ For utility finance, uncertainty does not justify inaction. It justifies better risk allocation, clearer commercial terms, and stronger alignment between the customer creating the need and the customer paying for the solution. ## **Beyond Traditional Rate Base Funding** The magnitude of investment required over the next decade means utilities cannot rely solely on traditional financing approaches. Rate base investment will remain central, but it cannot become the default answer for every infrastructure need created by every large-load customer. CFOs are increasingly exploring diversified funding strategies that preserve balance sheet strength while accelerating infrastructure deployment. These strategies include CIAC, direct assignment of dedicated facilities, minimum demand commitments, collateral requirements, exit fees, long-term service agreements, public-private partnerships, federal funding, state economic development support, and customer-backed infrastructure contributions. One opportunity lies in economic development partnerships. State and local governments often have strong incentives to attract major investments and may contribute funding for enabling infrastructure when the broader economic benefits are clear. Federal programs also remain relevant as policymakers recognize the importance of grid modernization, resilience, domestic manufacturing, and energy security. Grants, tax incentives, loan programs, and public-private partnerships can significantly reduce the amount of capital ultimately recovered through rates. Utilities are also beginning to explore infrastructure partnerships with large-load customers themselves. Rather than viewing customers solely as ratepayers, some organizations are creating collaborative investment structures where risks and benefits are shared more directly. This shift reflects a broader reality. The grid is increasingly becoming a platform for economic growth, and platform investments often require platform financing models. For CFOs, the opportunity is not only to find more capital. It is to design financing structures that make growth durable, fair, and executable across engineering, operations, commercial teams, regulators, and customers. ## **Large-Load Tariffs Are Becoming a Core CFO Tool** Large-load tariffs are quickly becoming one of the most important regulatory tools in the utility finance toolkit. In 2025, state regulators approved 29 large-load tariffs, and additional tariffs have continued to emerge as utilities and commissions respond to the scale of data center and advanced manufacturing demand.⁵ This trend is significant because large-load tariffs are not simply pricing mechanisms. They are risk-allocation frameworks that can help determine how infrastructure costs, demand uncertainty, and customer-specific grid impacts are handled before capital is deployed. The strongest tariffs include minimum contract terms, minimum bills, collateral requirements, demand ramp schedules, exit provisions, curtailment or flexibility provisions, and direct assignment of upgrade costs where appropriate. These elements matter because they help ensure that existing customers are not left paying for infrastructure built around speculative or delayed projects. The weaker versions create a new customer class without solving the underlying problem. A tariff that does not address stranded asset risk, timing uncertainty, or cost causation may offer the appearance of fairness while still leaving residential and small commercial customers exposed. For CFOs, tariff design should therefore be treated as a capital markets issue as much as a regulatory issue. Investors, rating agencies, regulators, and customers will increasingly want to know how much of a utility’s capital plan depends on large-load growth and what protections exist if that load does not arrive as forecast. Good tariff design also requires commercial discipline. Sales, engineering, operations, regulatory, and finance teams have to show up together with a shared understanding of what the customer needs, what the system can support, what risks are being created, and what commitments are required before the utility builds. ## **AI Creates Both Demand and Opportunity** Artificial intelligence is creating one of the largest load growth opportunities in utility history. At the same time, AI can help CFOs manage the financial consequences of that growth. Historically, infrastructure planning often relied upon deterministic forecasts and long planning cycles. Today’s environment requires far greater adaptability because customer development pipelines, technology requirements, chip efficiency, energy intensity, generation availability, and interconnection timelines are all moving at once. Advanced forecasting tools can improve load growth predictions by incorporating customer development pipelines, regional economic indicators, weather patterns, distributed energy resources, electrification trends, market signals, and known infrastructure constraints. Digital twins and scenario modeling can help CFOs evaluate multiple investment pathways before committing capital. AI-driven planning can also identify opportunities to defer investments through operational optimization, distributed energy resources, demand response programs, flexible load management, storage, and non-wires alternatives. In many cases, the lowest-cost infrastructure investment is the one that never has to be built. This does not eliminate the need for capital deployment. It improves capital efficiency, and for CFOs managing billions of dollars in future investment decisions, that distinction matters enormously. The same point is emerging in the broader AI and energy conversation. AIxEnergy’s recent analysis has emphasized that the next bottleneck is not just power availability, but proof: institutions need to verify what their models claim before using them to guide consequential infrastructure decisions.⁶ For utilities, that principle is especially important. Forecasts must become more transparent, assumptions must become more auditable, and capital plans must be supported by evidence that regulators and customers can understand. The real value of AI is not that it makes planning sound more sophisticated. The real value is that it can help utilities make better decisions earlier, reduce avoidable spend, improve customer outcomes, and strengthen the operating rhythm between the field, the control room, the finance organization, and the executive team. ## **Flexibility Should Be Treated as a Financial Asset** The most underpriced resource in utility finance is flexibility. Utilities have traditionally treated demand as something to be served, but the next phase of grid modernization requires demand to become something that can also be shaped. This is especially relevant for data centers. Not all computing workloads are equally time-sensitive, and some AI training, batch processing, and non-latency-critical workloads can shift in time, reduce consumption during peak events, or respond to grid signals without undermining service quality. Research and field demonstrations are beginning to show what this could mean in practice. A 2025 field demonstration in Phoenix found that software-based orchestration of AI workloads reduced cluster power usage by 25 percent for three hours during peak grid events while maintaining quality-of-service guarantees.⁷ For CFOs, this changes the investment equation. A flexible 500 MW customer is not financially equivalent to an inflexible 500 MW customer, because flexibility can reduce peak capacity requirements, defer network upgrades, improve asset utilization, and lower the cost of serving load. Tariffs and contracts should reflect that distinction. Large-load customers should contribute appropriately toward the infrastructure they require, but they should also be able to earn credit for verifiable grid value when they provide flexibility, storage, controllable demand, clean capacity, or other system benefits. This is the next step beyond cost allocation. The mature framework is cost causation plus grid value creation, with measurement and verification strong enough to stand up in front of customers, regulators, and investors. ## **Protecting Residential and Small Commercial Customers** Perhaps the most important responsibility facing utility CFOs is preserving trust. Customers generally support economic growth, reliability improvements, and investments that strengthen communities. What they do not support is the perception that they are paying for someone else’s growth. That perception can quickly turn necessary grid investment into a legitimacy problem, especially when household budgets are already under pressure. Utilities must therefore establish transparent frameworks that clearly connect costs to beneficiaries. This requires thoughtful rate design, rigorous cost allocation methodologies, proactive stakeholder engagement, and clear documentation of when a proposed investment creates broad system value rather than narrow customer-specific value. Large-load customers should contribute appropriately toward infrastructure requirements they create. Risk-sharing mechanisms should protect existing customers from speculative development activity, and regulators should receive clear evidence demonstrating that infrastructure investments generate broad system benefits before costs are socialized. Most importantly, CFOs must communicate these principles consistently and transparently. Affordability is not merely a regulatory requirement; it is a strategic asset. Utilities that maintain customer trust will find it easier to secure regulatory support, attract investment capital, and pursue long-term growth opportunities. Utilities that lose that trust will find that even technically sound capital plans become harder to approve, finance, and execute. Customer trust is also built through execution. Utilities have to deliver what they promise, explain what they are doing, and show that growth is being managed with the same discipline they expect from their best industrial and commercial customers. ## **The Emerging CFO Playbook** The utility CFO of the future will look different from the CFO of the past. Financial stewardship remains essential, but it is no longer sufficient. Tomorrow’s CFO must understand grid architecture, load forecasting, distributed energy resources, regulatory strategy, economic development, AI-enabled planning, customer affordability, and capital market expectations. They must become translators between investors seeking growth, regulators seeking fairness, operators seeking reliability, and customers seeking affordability. The first priority is to build a beneficiary-based capital framework. Every major investment should be mapped by who benefits, who creates the need, who bears the risk, and whether the investment produces broad system value or primarily customer-specific value. The second priority is to treat CIAC as strategy rather than administration. CIAC should be integrated into large-load policy, interconnection review, collateral requirements, and capital planning before engineering decisions are effectively locked in. The third priority is to use large-load tariffs to price risk honestly. Minimum bills, long-term contracts, demand commitments, ramp schedules, exit fees, and security requirements should become standard tools for high-impact load additions. The fourth priority is to require proof before socialization. Before costs are moved into rate base, utilities should demonstrate credible load certainty, broad system benefit, and alternatives analysis that includes non-wires options and flexibility. The fifth priority is to integrate AI into finance and planning in practical ways. AI should improve forecasting, scenario analysis, asset utilization, outage prediction, interconnection screening, and capital prioritization, not simply add another layer of complexity to already difficult planning processes. The sixth priority is to build cross-functional operating discipline around large-load growth. Finance, commercial, engineering, operations, regulatory, supply chain, and customer teams need a shared process for evaluating projects, pricing risk, setting expectations, and moving from opportunity to execution without losing accountability. Most importantly, CFOs must recognize that the challenge is not choosing between growth and affordability. The challenge is designing financial structures that enable both. ## **A Defining Decade** The utility industry is entering a defining decade. Few periods in modern utility history have presented this level of opportunity, and few have carried this level of responsibility. The investments made over the next ten years will shape economic development, grid reliability, customer affordability, and national competitiveness for generations. Utilities will need to build more infrastructure, but they will also need to build smarter financing models. They will leverage tools such as CIAC, public-private partnerships, advanced forecasting, AI-driven planning, large-load tariffs, flexible load programs, and innovative capital structures to ensure that growth remains sustainable. They will also need to demonstrate that the benefits and risks of this growth are being allocated fairly. A utility’s most valuable asset is not its transmission system, its substations, or its generation portfolio. Its most valuable asset is the trust of the customers and communities it serves. The CFO’s job is to protect that trust while funding the future. That is the new mandate, and it will define the next era of utility leadership. ## Sign up for AIxEnergy AIxEnergy helps leaders understand the physical economy of artificial intelligence by connecting energy, infrastructure, markets, policy, and governance into clear strategic insight. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## **Notes** 1. U.S. Department of Energy, “DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers,” December 20, 2024\. The DOE summary reports that U.S. data center electricity use rose from 58 TWh in 2014 to 176 TWh in 2023 and could reach 325–580 TWh by 2028\. ([The Department of Energy's Energy.gov](https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data-centers?ref=aixenergy.io)) 2. Reuters, “US Power Use to Beat Record Highs in 2026 and 2027 as AI Use Surges, EIA Says,” June 9, 2026\. The report cites EIA projections that U.S. power consumption will reach 4,271 billion kWh in 2026 and 4,397 billion kWh in 2027\. ([Reuters](https://www.reuters.com/business/energy/us-power-use-beat-record-highs-2026-2027-ai-use-surges-eia-says-2026-06-09/?utm%5Fsource=chatgpt.com)) 3. *Wall Street Journal*, “Who Pays? AI Boom Sparks Fight Over Soaring Power Costs,” July 29, 2025; *Washington Post*, “Why Your Power Bill Is Spiking Faster Than a Nearby Data Center’s,” January 15, 2026\. ([Wall Street Journal](https://www.wsj.com/business/energy-oil/ai-data-center-power-costs-bbfcd862?utm%5Fsource=chatgpt.com)) 4. International Energy Agency, “Energy Demand from AI,” in *Energy and AI*, accessed June 15, 2026\. The IEA projects that global data center electricity consumption could reach roughly 945 TWh by 2030 in its base case. ([IEA](https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai?ref=aixenergy.io)) 5. Meris Lutz, “Large Load Tariffs Proliferate as States Take More Active Role in Data Center Regulation,” *Utility Dive*, March 31, 2026\. The article reports that state regulators approved 29 large-load tariffs in 2025\. ([Utility Dive](https://www.utilitydive.com/news/large-load-tariffs-proliferate-as-states-take-more-active-role-in-data-cent/816184/?utm%5Fsource=chatgpt.com)) 6. AIxEnergy, “This Week,” accessed June 15, 2026\. The tag page includes “The Grid’s Next Bottleneck Is Proof,” “The Pause Button,” and related AI-and-grid analyses published in June 2026\. ([AIxEnergy](https://www.aixenergy.io/tag/thisweek/)) 7. Philip Colangelo et al., “Turning AI Data Centers into Grid-Interactive Assets: Results from a Field Demonstration in Phoenix, Arizona,” arXiv, July 2025\. The paper reports a 25 percent reduction in cluster power usage for three hours during peak grid events while maintaining quality-of-service guarantees. ([arXiv](https://arxiv.org/abs/2507.00909?utm%5Fsource=chatgpt.com)) ### The Grid’s Next Bottleneck Is Proof URL: https://www.aixenergy.io/the-grids-next-bottleneck-is-proof/ Last updated: 2026-07-29T19:39:52.000Z AI is making analysis cheap just as data centers are forcing harder electricity decisions. The winners will be the institutions that can verify what their models claim. _This post is for paying subscribers only._ ### The Geography of Intelligence URL: https://www.aixenergy.io/intelligence-geography/ Last updated: 2026-07-29T19:40:03.000Z As AI Pushes U.S. Electricity Demand to Record Highs, a New Computing Architecture Is Emerging _This post is for paying subscribers only._ ### The Power Sector's AI Decade URL: https://www.aixenergy.io/the-power-sectors-ai-decade/ Last updated: 2026-07-13T01:39:02.000Z *This is the fourth of a five-part series at* [*AIxEnergy*](https://www.aixenergy.io/) *and mirrored at Michael Leifman's* [*Substack*](https://michaelleifman.substack.com/) [*Part 1*](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fi/) *introduced three distinctions (vendor capability versus utility adoption, reliability versus safety, task versus role) and a twelve-row sample of the US electricity workforce.* [*Part 2*](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fii/) *developed the four-stage mechanism by which AI capability translates into workforce change in regulated industries and traced it through aviation, freight rail, and maritime.* [*Part 3*](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fiii/) *maps how AI is reshaping the U.S. electric-power workforce role by role, showing that task content is already shifting across the industry while headcount changes will unfold unevenly depending on whether work is routine, reliability-critical, safety-critical, or institutionally protected*. *This article explains that AI and robotics are not only replacing existing power-sector tasks, but also creating new capabilities by letting the grid coordinate more complexity than human operators can manage and sending machines into hazardous places where human bodies should not go.* Years ago, in a training on power plant fundamentals at GE, I came to understand how much hazard is packed into the ordinary machinery of making electricity. A generating station is full of places hostile to the human body, from the energized switchgear where an arc flash can reach temperatures above [35,000 degrees Fahrenheit, nearly four times as hot as the surface of the sun](https://www.osha.gov/electrical/flash-hazards?ref=aixenergy.io) in a fraction of a second, hot enough to vaporize the copper and aluminum it travels through, to the confined and superheated interiors of boilers and the high structures that were never meant for a person to stand on. The lesson that stayed with me was that across much of the power system the binding constraint on the work is not the skill of the worker but the vulnerability of the human body, which is why so much of it is hedged with safe working distances, protective equipment, and procedures written because the failure mode is measured in milliseconds. Article Summary 0:00 /66.821224 1× The boiler is one of those places. In the fall of 2012, a Grove City College junior named Jake Loosararian toured the Scrubgrass Generating Plant in Kennerdell, Pennsylvania, and the plant manager told him that a worker had [died the previous year while inspecting a boiler for defects](https://www.aipowerpittsburgh.com/builders/gecko-robotics-why-pittsburgh?ref=aixenergy.io). Loosararian and his classmates built a forty-pound robot with an ultrasonic scanner that could climb the boiler walls and hunt for the cracks and thinning that precede a failure. That senior project became [Gecko Robotics](https://www.cnbc.com/2025/07/10/gecko-robotics-ceo-built-billion-dollar-business-by-going-through-hell.html?ref=aixenergy.io), valued at $1.25 billion by June 2025, whose magnetic-wheeled robots now scale the walls of boilers, tanks, and other structures across the power sector, carrying sensors and, the company says, capturing far more of the surface than a rope crew ever could. The first three parts of this series traced power-sector work that AI and robotics can now do in place of people. This part takes up the other half of the story, the half that does not register as a subtracted job, because it is work people could not do well, safely, or at all. Some of it was out of reach because no feasible number of people could do it fast enough. The rest was out of reach because a human body could not be where the work had to happen, or could not survive being there. Two kinds of capability are arriving at once, and it helps to keep them separate. The first is cognitive: AI can now hold more variables in view, and act on them faster, than any control room of people could manage. The second is physical: robots built in shapes no human resembles can now enter spaces that injure or kill an unprotected person. When I wrote about AI and labor in 2018, I [expected both cognitive and physical work to be exposed](https://michaelleifman.substack.com/p/what-i-got-right-and-wrong-about), brain and brawn together. What I did not expect was the sequence. Cognitive automation arrived first and fast, through language models, while physical automation lagged, slowed by the stubborn difficulty of getting machines to manipulate the real world. The power sector is one revealing exception to the general rule. Here, physical capability is arriving early in a specific set of niches, not because the robotics got easier, but because the human was never a safe option to begin with. ## The first engine: cognition at a scale people cannot reach Hannah Kaplan, writing on this platform, describes the shift underway in grid operations as a move [from reactive reliability to dynamic intelligence](https://www.aixenergy.io/from-reactive-reliability-to-dynamic-intelligence-how-ai-is-building-the-grid-we-actually-need/). For most of its history the grid was operated reactively. Something failed, an alarm sounded, and a crew responded. The system Kaplan describes is different in kind. It coordinates millions of devices at the edge of the grid, rooftop solar, home batteries, electric vehicles, and smart thermostats, in something close to real time, shaping supply and demand continuously rather than waiting for the next emergency. That coordination is not a faster version of what a human dispatcher did. It is a task no dispatcher could do at all. Consider what happened in California on the evening of July 29, 2025\. More than [100,000 home batteries discharged together for two hours, delivering an average of about 535 megawatts to the grid](https://electrek.co/2025/08/04/california-grid-gets-a-record-power-assist-from-a-100k-home-battery-fleet/?ref=aixenergy.io), the output of a midsize power plant, assembled out of appliances sitting in tens of thousands of garages. The event was a test of California’s Demand Side Grid Support program, coordinated by the state’s Energy Commission and its grid operator, with the batteries pooled by aggregators including Sunrun and Tesla. The act of conducting that many devices, second by second, so that they behave like a single plant, is software doing something no staff of people could. The same pattern appears in the queue of projects waiting to connect to the grid, which I covered in Part 3\. Reviewing an interconnection request means running power-flow studies to see what a new generator or large load would do to the surrounding network, work that has historically taken human engineers months per case and produced a backlog measured in years. AI systems can run those studies at a speed that compresses the arithmetic from months toward minutes, which matters because the queue, not the supply of equipment, is now one of the binding limits on how fast new power can come online. Kaplan points to the same kind of capability in wildfire country, where utilities use predictive models to score ignition risk across an entire service territory and decide where to cut power before a fault becomes a fire. Scoring a whole territory continuously, weighing weather, vegetation, and equipment condition together, is once again a problem of more variables than a person can hold at once. ## The second engine: places the body cannot go The robots showing up in the power sector look nothing like people, and that is the point. I have made a version of this argument before: we long ago stopped designing our fields for the ox and our roads for the horse, and there is [little reason to design the next era of work around the human form](https://michaelleifman.substack.com/p/what-i-got-right-and-wrong-about) either. A human shape is the right design only when the work was built around human bodies, and the work in question here was not. It happens inside narrow gaps, atop tall structures, underwater, and in radiation fields, so the machines that do it take whatever form the task demands. The substation is another such place. In 2024 the utility Avangrid began a pilot at two of its [United Illuminating substations in Connecticut](https://www.tdworld.com/substations/article/21282583/avangrid-pilots-mobile-robot-dog-to-advance-substation-inspections-with-artificial-intelligence?ref=aixenergy.io), sending a four-legged robot named Spot, made by Boston Dynamics and nicknamed Sparky, on inspection rounds to read gauges and capture thermal images of transformers and breakers, with software that flags anomalies. The value, as the utility described it, is partly that the robot can inspect far more often than a crew making periodic visits, building a record of how equipment behaves across seasons and loads. It is also that every inspection the robot performs is one a person does not have to perform inside an energized yard. The same constraint operates inside the machines themselves. A large generator is normally inspected by pulling the multi-ton rotor out of the stator, a process so costly and slow that the examination was historically [left out of outage schedules](https://www.powermag.com/minimize-risks-with-robotic-generator-inspections-and-comprehensive-testing/?ref=aixenergy.io) when time was short. Today a robot small enough to pass through the [roughly one-inch air gap between rotor and stator](https://www.mdaturbines.com/services/generator/robotic-generator-in-situ-inspection/?ref=aixenergy.io) crawls the length of the machine carrying cameras and test instruments, performing the inspection with the rotor left in place, and GE Vernova offers a related tool that scans a generator’s retaining rings for stress-corrosion cracks without removing them. This is inspection, not repair, and the distinction matters. Getting a robot to see inside a place a person cannot reach turned out to be far easier than getting one to fix what it finds there, which is the same lesson the wider robotics industry has been learning, that locomotion was the solved problem and manipulation the hard one. In-place robotic repair inside turbines and generators exists today mostly at the prototype stage. The same shift is visible on wind turbines, where the leading edges of blades erode under years of rain and grit a hundred meters in the air. That repair has meant rope-access technicians, who climb or hang from the blade and can work only in calm weather. The company [Aerones](https://www.compositesworld.com/products/aerones-unveils-modular-robot-for-level-1-3-wind-turbine-maintenance-repair?ref=aixenergy.io) now runs a ground-controlled robotic system, suspended by cables, that cleans, sands, fills, and recoats eroding blades, and can work in higher winds and a wider temperature range than a person on a rope can. By 2025 the company projected servicing [more than 10,000 turbines](https://www.s2ginvestments.com/insights/welcome-aerones?ref=aixenergy.io), with deployments at US wind plants including two in Oklahoma. Notably, Aerones is explicit that it is [not removing the technician](https://aerones.com/Services/Leading-Edge-Repair/?ref=aixenergy.io): the robot does the leading-edge work while technicians operate it from the ground and handle the repairs the machine cannot, a division of labor worth holding onto for what follows. The hardest limit of all is radiation, because it is a biological ceiling that no protective equipment lifts. Inside a nuclear plant, the surfaces of the reactor refueling cavity must be cleaned and decontaminated during outages, work historically done by people after the water was drained, slowly, and at the cost of [radiation dose to those workers](https://www.world-nuclear-news.org/Articles/First-applications-for-Diakonts-decontamination-r?ref=aixenergy.io). The company Diakont sends a hybrid swimming-and-crawling robot into the flooded cavity instead, operated by a small team from the edge of the refueling floor. At the [Perry plant in Ohio](https://diakont.com/case-studies/nuclear-solutions/reducing-dose-and-saving-time-with-robotic-cavity-decon/?ref=aixenergy.io), the robotic decontamination was thorough enough that no manual cleaning was needed afterward, and it kept people out of the cavity entirely. Robotic inspection of reactor internals and of the thousands of narrow tubes inside a steam generator follows the same logic, sending a machine where a dose limit forbids a person. Described separately, the two engines read like different stories, but they are beginning to act as one. Imagine an inspection robot inside a grid-scale battery enrolled in a program like California’s, picking up the early thermal signature of a cell starting to fail. The same coordination layer that conducts a hundred thousand batteries could take that one unit offline and rebalance the rest of the fleet within seconds, faster than any operator watching a screen could react. The physical engine reaches what no person could reach, and the cognitive engine acts on what it finds at a speed no person could match. That combination, more than either engine alone, is where the capability becomes something new. ## The edge of what does not exist yet Everything to this point either is in service or is close to it. The more interesting boundary is the one just beyond, where the most consequential capabilities are still being invented. Putting power lines underground would end much of the storm and wildfire damage that overhead wires suffer, but at [more than six million dollars a mile](https://www.constructiondive.com/news/DOE-explores-moving-power-lines-underground/705485/?ref=aixenergy.io) it is so expensive that utilities rarely do it. In January 2024 the Department of Energy’s advanced research arm launched a thirty-four-million-dollar program, [GOPHURRS](https://arpa-e.energy.gov/technologies/programs/gophurrs-0?ref=aixenergy.io), to cut that cost by shifting undergrounding from digging to drilling. One of its dozen projects, GE Vernova’s [SPEEDWORM](https://www.utilitydive.com/news/DOE-explores-moving-power-lines-underground/704637/?ref=aixenergy.io), is a robot that mimics the way an earthworm moves to dig and lay cable in a single pass, with the goal of installing a thousand feet of line in two hours from the back of a standard pickup truck. It does not exist as a working machine yet. If it comes to work, it would give the power system a capability it has never had, the ability to put lines underground at a cost that makes undergrounding ordinary rather than exceptional. The harder version of the same problem appears during a storm, when drones can map the damage while it is still too dangerous to send people, but the repair itself cannot begin. After Hurricane Helene, Duke Energy [used drones to help restring a transmission line](https://illumination.duke-energy.com/articles/new-helicopters-smarter-drones-duke-energys-storm-response-upgrades?ref=aixenergy.io), and robots can now inspect energized lines on their own, as EPRI’s “Ti” does [gliding along the wire above an energized transmission line in Ohio](https://eprijournal.com/robots-take-on-tough-tasks-in-transmission-and-distribution/?ref=aixenergy.io), with a few systems able to make limited repairs on live conductors while a crew stands by. What does not yet exist is the capability that would matter most in a storm, a machine that could find and mend a fault on its own while the weather still made the site too dangerous for people. Until then the repair waits for the storm to pass and for crews to reach the poles, which is why a bad storm is still measured in days without power, and why the storm, not the technology, sets the clock. ## What gets created, and what does not Both engines create capability that did not exist, and it is tempting to read that as unambiguous good news for employment, since new capability has historically meant new work. Some of that holds here. The orchestration layer needs people who can build and run it, the robots need operators, the data they produce needs interpreting, and the wind example shows the pattern at its most benign, with the dangerous task removed and the worker repositioned rather than dismissed. New capability is not the same as new employment, though, and the new work rarely appears where the old work was. The cognitive engine is real but bounded. Kaplan’s own assessment is that AI, distributed resources, and virtual power plants [do not replace wires or generation, but buy time and optionality](https://www.aixenergy.io/from-reactive-reliability-to-dynamic-intelligence-how-ai-is-building-the-grid-we-actually-need/), and that the industry needs precision at scale more than it needs more pilots. Its progress depends on data quality, on regulatory alignment, and on institutional trust, none of which arrives on a software schedule. The physical engine is bounded by the gap between seeing and fixing: the inspection is here, the in-place repair mostly is not. And where a robot removes a person from a hazard, it also removes the hours that person was paid for, even when no one would wish those particular hours back on a human being. Part 3 of this series drew two lines worth recalling. One separated the moment a capability becomes possible from the much later moment institutions actually absorb it, since licensing, liability, labor agreements, and regulatory minimums all move slowly. The other separated the tasks a technology takes over from the jobs it eventually ends, since a role can shed its most dangerous tasks long before anyone counts a position as gone. The capabilities in this part bear on the first line in a particular way. Because so many of them do work no one was doing before, reaching where a person never could or coordinating more than a person ever tracked, there is no incumbent role to protect, no contract or license or staffing minimum standing in the way, and so the institutional resistance that slows displacement is weakest, almost by definition, exactly where the capability is newest. The harder questions, what institutions do as these tools mature and how the loss of tasks becomes the loss of roles, belong to the final part, which turns from what AI and robotics can do to what utilities, regulators, workers, and investors should do about it. The capabilities described here are neither science fiction nor uniformly mature. A robot on an inspection route and a fleet of home batteries acting as a power plant are in service now, while in-place repair inside a turbine, a line laid by a burrowing robot, and the full orchestration of a dynamic grid are closer to the frontier than to the norm. That mixture is exactly what makes the trajectory easy to underestimate. Each piece looks like a pilot, a demonstration, a research grant. Put together, they describe a power system in which the limits once set by what a human body could survive, and by how many variables a human mind could track, are quietly being lifted. The decade ahead looks slow until it does not. ## Sign up for AIxEnergy AIxEnergy helps leaders understand the physical economy of artificial intelligence by connecting energy, infrastructure, markets, policy, and governance into clear strategic insight. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### The Millisecond Menace URL: https://www.aixenergy.io/the-millisecond-menace/ Last updated: 2026-07-29T19:42:34.000Z by [Morgan Bazilian](https://www.linkedin.com/in/morganbazilian/?ref=aixenergy.io) and [Denis Kouroussis](https://www.linkedin.com/in/denis-kouroussis/?ref=aixenergy.io) Global data center electricity consumption reached an estimated [415 terawatt-hours in 2024](https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai?ref=aixenergy.io), roughly 1.5 percent of all electricity generated worldwide. The International Energy Agency (IEA) projects that figure will nearly double to [945 terawatt-hours by 2030](https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary?ref=aixenergy.io), growing at around 15 percent per year, more than four times the rate of every other sector combined. In the United States, [Lawrence Berkeley National Laboratory](https://newscenter.lbl.gov/2025/01/15/berkeley-lab-report-evaluates-increase-in-electricity-demand-from-data-centers/?ref=aixenergy.io) estimates that data centers consumed 176 terawatt-hours in 2023, representing 4.4 percent of total national electricity use, with projections ranging from 325 to 580 terawatt-hours by 2028\. The forces behind this expansion are familiar: larger models, faster chips, and seemingly relentless scale-out. What receives far less attention is what happens once all that power arrives at the site. 0:00 /0:35 1× The answer the industry seems to be converging on, in its more forward-looking configurations, is 800 V direct current (DC) distribution fed by solid-state transformers (SST) that convert medium-voltage (MV) alternating current (AC) directly to the DC bus in a single step. [NVIDIA announced 800 V DC as its target rack architecture at Computex 2025](https://developer.nvidia.com/blog/nvidia-800-v-hvdc-architecture-will-power-the-next-generation-of-ai-factories/?ref=aixenergy.io), specifying it for Kyber rack-scale systems scheduled for full production in 2027\. But, beneath the efficiency gains lies a control-systems problem the industry has yet to fully confront: the interaction between millisecond-scale Graphics Processing Unit (GPU) power transients, cascaded power electronics, and the distributed impedances of the DC bus. This is an engineering problem that will determine whether these architectures deliver on their promise. ## The Case Against Conventional Architecture 💡 Conventional data center power architecture moves electricity through multiple conversion stages before it reaches GPUs, adding cost, long equipment lead times, and 6–8% energy losses—enough to waste 6–8 MW in a 100 MW facility. The emerging 800 V DC architecture, using solid-state transformers to feed racks directly from medium-voltage utility service, could collapse that equipment stack and improve efficiency, but it introduces a more complex control-systems challenge that must still be proven at scale. The traditional path from a medium-voltage utility feed to a GPU involves four discrete conversion steps: a medium-voltage transformer, low-voltage switchgear, an uninterruptible power supply, and a power distribution unit. Each adds losses, cost, and a procurement lead time of its own. Together they dissipate 6 - 8% of input energy before it reaches the compute tray. At 100 MW of grid input, a conventional 480 V Alternating Current (AC) architecture wastes 6 to 8 MW in distribution hardware alone, capacity that could otherwise run GPUs. The supply-chain burden of this equipment stack has become severe. [Wood Mackenzie's second-quarter 2025 survey](https://www.woodmac.com/news/opinion/mind-the-gap-tackling-supply-chain-challenges-in-the-electric-td-sector/?ref=aixenergy.io) found standard power transformers averaging 128 weeks for delivery and generator step-up units running to 144 weeks, with switchgear at 44 weeks. The [North American Electric Reliability Corporation](https://www.theinvadingsea.com/2025/12/05/power-grid-electricity-supply-chain-data-centers-transformers-circuit-breakers-utilities/?ref=aixenergy.io) recorded lead times crossing 120 weeks for power transformers in 2024, with larger transformers stretching to 210 weeks. Demand for high-voltage power transformers has grown 116 percent since 2019, according to Wood Mackenzie, and generator step-up units by 274 percent, faster than domestic manufacturing capacity has been able to respond. For a data hall that requires four MV transformers, multiple switchgear assemblies, Uninterruptible Power Supply (UPS) systems, and dozens of Power Distribution Units (PDUs), these timelines dominate the construction schedule in ways that site preparation and civil works cannot offset. The 800 V DC approach collapses this equipment stack. By connecting a solid-state transformer directly to the 34.5 kV utility feed and delivering 800 V DC to the rack, the intermediate conversion chain disappears. Companies including [Heron Power Electronics](https://www.heronpower.com/?ref=aixenergy.io), [Delta Electronics](https://www.deltapowersolutions.com/en/mcis/news-2025-delta-demonstrate-hvdc-power-cooling-networking-to-drive-ai-data-center-evolution-at-ocp-global-summit.php?ref=aixenergy.io), and [Enphase Energy](https://investor.enphase.com/news-releases/news-release-details/enphase-energy-announces-development-iq-solid-state-transformer?ref=aixenergy.io) are developing Solid-State Transformer (SST) platforms claiming conversion efficiencies above 98 percent, against the 93 to 94 percent typical of the conventional chain. A [2025 simulation study by Xu et al.](https://arxiv.org/pdf/2601.16502?ref=aixenergy.io) published on arXiv provides one of the more rigorous technical evaluations of this architecture, demonstrating stable SST-based 800 V DC operation under realistic Artificial Intelligence (AI) workload profiles in real-time digital simulation. It also documents the complexity of coordinating the multiple nested control loops that SST operation requires, a complexity the industry tends to understate. Heron Power Electronics recently published a [technical blueprint](https://www.heronpower.com/?ref=aixenergy.io) proposing a 12 MW data hall building block powered by four 4.2 MW solid-state transformers, each connecting to a 34.5 kV feed and delivering 800 V DC to forty compute racks in a 4-to-make-3 redundant topology. The company claims that eliminating the conventional MV transformer, switchgear, UPS, and PDU cuts equipment and installation costs by 65 to 90 percent, reduces construction timelines by roughly three months, and eliminates the dedicated electrical room. Whether those figures hold across diverse project conditions remains to be demonstrated at scale. ## When GPUs Breathe Together 💡 AI training clusters behave unlike conventional data center loads: thousands of GPUs spike, pause, and surge together, creating synchronized power swings that can propagate from racks to the utility interconnect and reach tens or hundreds of megawatts. Current stabilization methods can reduce those swings, but they often burn energy simply to hold power steady, meaning some of the efficiency gains promised by 800 V DC architectures may be offset by the need to control GPU-driven transients. Modern GPU architectures bear little resemblance to traditional electrical loads. A data center chiller ramps over seconds; conventional server demand shifts gradually with workload. AI training clusters behave differently. During matrix computations, GPU power spikes; during data transfers and synchronization, it drops sharply. Checkpointing operations can briefly reduce load to near zero before snapping back to full demand. A 2025 paper from Microsoft, OpenAI, and NVIDIA, ["Power Stabilization for AI Training Datacenters" by Choukse et al.](https://arxiv.org/abs/2508.14318?ref=aixenergy.io), documents that these fluctuations span 30 to 100 percent of thermal design power (TDP) within windows as short as 0.2 to 2 seconds, with intra-batch variations occurring at millisecond granularity. Distributed AI training synchronizes thousands of GPUs through collective operations such as AllReduce; every GPU in a cluster steps through the same iterations in lockstep. The facility's power draw does not fluctuate randomly in ways that average out. [Oracle's cloud infrastructure team](https://blogs.oracle.com/cloud-infrastructure/behind-the-scenes-gpu-power-smoothing-ai-training?ref=aixenergy.io) has documented seeing this signature propagate from the rack level to the data center level and ultimately to the utility interconnect, where swings can reach tens or even hundreds of megawatts. The Choukse paper notes that the frequency spectrum of these oscillations, concentrated between 0.2 and 3 Hz, falls close to known resonant modes of turbine-generator shafts and long transmission lines, raising the prospect of mechanical stress in upstream generation equipment. [SemiAnalysis has reported](https://newsletter.semianalysis.com/p/ai-training-load-fluctuations-at-gigawatt-scale-risk-of-power-grid-blackout?ref=aixenergy.io) that multiple utility providers have recorded harmonic impacts from synchronized computing loads, with amplitude growing proportionally to cluster size. The Choukse paper documents two mitigation strategies currently deployed in production, and their energy costs are instructive. The first, called Firefly, injects controlled secondary workloads during low-power communication phases to sustain a more uniform draw. It works, but as the authors acknowledge, energy burned on artificial computations that produce no useful output is simply wasted. The second uses NVIDIA's GB200 Minimum Power Floor feature, which prevents GPU power from dropping below a specified fraction of thermal design power even when no useful work is running. At a floor set to 90 percent of TDP, the paper reports a total energy overhead of 10.5 percent. For a 100 MW training cluster, that represents 10.5 MW of continuous power consumption that produces no computational output, burned purely to prevent the consequences of load transients. They are in production because unmitigated swings at critical grid frequencies pose genuine risks to interconnection agreements and, at sufficient scale, to grid stability. A system that promises 50 percent lower distribution losses but requires 10 percent energy overhead for load stabilization has surrendered a significant portion of its claimed advantage, before accounting for the additional complexity of coordinating GPU-level controls with SST-level and battery-level controls across a distributed DC bus. ## The Architecture of Stability 💡 In an 800 V DC architecture, high-C-rate batteries integrated at the solid-state transformer can smooth the data center’s grid-facing load, but they may not respond fast enough to clamp the sharpest voltage excursions at distant racks. That means stability likely requires a coordinated hierarchy: local capacitance near the racks for sub-millisecond events, bulk storage at the SST for larger slower swings, and carefully tuned control loops so the battery, SST, and rack converters do not fight each other. The conventional response to power transients in data centers has been to place a battery buffer behind the UPS, centralizing storage at the conversion stage. In that configuration, the storage sits at a low-impedance node close to the source of regulation, and fast transients are absorbed before they propagate. The 800 V DC architecture offers a different option: integrating a high-C-rate battery directly onto the DC bus at the SST. Heron's blueprint co-locates a battery buffer unit with each Heron Link, coupling it directly to the 800 V DC bus without an additional DC-to-DC converter, so it can respond to load fluctuations faster than any mechanical switching device and present a smooth draw to the grid. But, a transient absorber is only as effective as the impedance path between it and the transient source allows. Storage close to the load, with low series inductance and resistance, can respond fast and clamp hard. Storage at the SST, separated from each rack by the full length of the DC distribution cable runs, sees a higher-impedance path. By the time charge traverses that network, the voltage excursion may have already occurred at the rack. The practical result is that SST-integrated bulk storage smooths the load profile as seen from the grid while doing comparatively less to clamp the fastest voltage excursions at the rack itself. The two storage roles may therefore be complements rather than substitutes: modest local capacitance at the rack handles sub-millisecond events that distribution impedance would otherwise allow to propagate, while bulk energy storage at the SST handles the slower, larger swings that constitute the bulk of grid-visible fluctuation. The [Xu et al. simulation study](https://arxiv.org/pdf/2601.16502?ref=aixenergy.io) addresses SST control architecture in detail. The paper models a cascaded H-bridge front-end rectifier converting medium-voltage AC to intermediate DC buses, with dual-active-bridge (DAB) converters providing galvanic isolation and regulating the 800 V DC output. The control scheme is structured hierarchically: an outer voltage loop generates current references for an inner decoupled DQ current controller in the rectifier stage, while the DAB employs a simpler single-loop phase-shift controller. When the integrated battery and the SST's conversion stage share regulation of the same 800 V DC bus, a further complication emerges. Two control systems acting simultaneously on one node can, if designed without explicit coordination, interpret each other's corrections as disturbances and produce the oscillatory instability the storage was meant to prevent. This is a well-documented pathology in cascaded power electronics, traceable to the foundational work of [Middlebrook and Cuk in 1976](https://premiermag.com/wp-content/uploads/2022/01/H2PToday1911%5Fdesign%5FTexasInstruments%5Fpart10.pdf?ref=aixenergy.io), who showed that two individually stable subsystems can become unstable when cascaded if the output impedance of the upstream stage is not significantly lower than the input impedance of the downstream stage across all relevant frequencies. In the SST context, the minor-loop gain, defined as the ratio of the SST's output impedance to the load's input impedance, must be verified across the full operating envelope. As rack power demand pulses, the effective input impedance changes, and stability margins satisfied at the nominal design point may be violated at another. ## Cables, Bus Bars, and Hidden Resonances 💡 In 800 V DC data centers, cables, bus bars, and rack converters form a lightly damped resonant network where SST control actions meant to stabilize voltage can instead excite hidden oscillations that AC engineering intuition may miss. The move to 800 V DC changes the electrical character of the distribution network in ways that are not always intuitive to engineers trained on AC systems. DC cables and bus bars present primarily resistive and inductive impedance. A 50-meter cable run to a remote rack may present 20 to 30 microhenries of series inductance, modest in isolation but consequential in aggregate. Multiplied across dozens of parallel paths and combined with the input capacitance of rack-level DC-DC converters, the result is a distributed LC network with multiple resonant modes. These resonances matter because the SST's control loops operate at frequencies that can excite them. A voltage control loop with 200 Hz bandwidth, responding to a load step, injects energy from DC through several hundred hertz. If a cable-capacitor resonance falls within this range, the correction can excite the bus rather than stabilize it. The problem is compounded by the relative absence of damping in DC systems. AC distribution benefits from the inherent damping of line-frequency transformers, skin-effect losses, and resistive load components. DC systems optimized for efficiency minimize resistance everywhere, producing high-Q resonances that persist longer and ring to higher amplitudes. The tools and intuitions developed for 60 Hz AC distribution do not transfer directly to a DC network driven by fast-switching power electronics. ## The Regulatory Environment 💡 As AI data centers grow into hundreds-of-megawatt grid loads, regulators are beginning to treat them less like passive customers and more like active power-electronics systems that must prove ride-through capability, dynamic stability, and grid-compatible performance through commissioning tests conventional facilities were never designed to meet. These technical challenges do not exist in isolation from the regulatory environment. As AI campus loads grow to hundreds of megawatts on single feeders, grid operators are raising interconnection requirements. ERCOT, which operates the Texas grid and is processing more large-load interconnection requests than any other US operator, has become a practical reference point. [ERCOT's June 2025 Market Notice M-B062325-01](https://www.ercot.com/services/comm/mkt%5Fnotices/M-B062325-01?ref=aixenergy.io) formally requested voltage ride-through capability data from all data center loads of 75 MW or greater, citing reliability concerns about large electronic loads during grid disturbances. The December 2025 implementation of PGRR115 established mandatory interconnection requirements for such loads, including dynamic modeling standards. [ERCOT's December 2025 constraints report](https://www.ercot.com/files/docs/2025/12/23/2025-Report-on-Existing-and-Potential-Electric-System-Constraints-and-Needs.pdf?ref=aixenergy.io) notes that the operator is tracking more than three times as many large-load interconnection requests as in 2024, with the total queue pointing toward 181 GW by 2030, more than double the grid's current peak demand of 85 GW. IEEE 2800, the interconnection standard for inverter-based resources, is increasingly referenced as the model for what regulators will require of active power electronics at the grid interface. Facilities built on conventional passive-load architecture cannot meet these requirements without significant additional equipment. An SST with integrated battery storage, when properly designed, addresses them from the outset. The SuperBBU architecture in Heron's blueprint is designed to absorb AI training transients directly on the 800 V DC bus and provide sufficient hold-up time for backup generator startup during grid disturbances. But native capability and demonstrated capability are different things. The latter requires commissioning tests that conventional electrical infrastructure has never been asked to perform. ## Monitoring and the Operational Shift 💡 800 V DC AI data centers will require continuous, high-speed waveform monitoring and real-time control tuning because their electrical “normal” is a shifting dynamic envelope where instabilities can emerge faster than conventional power monitoring—or human operators—can respond. If 800 V DC SST architectures cannot be designed once and left alone, they must be continuously monitored and periodically retuned. This is a genuine departure from how electrical infrastructure has historically been operated. Traditional power monitoring focuses on steady-state quantities, voltage, current, power factor, harmonic content, sampled at low rates and compared against static thresholds. AI data center power systems violate each of these assumptions. Load changes occur faster than conventional monitoring can capture. Normal is not a fixed operating point but a dynamic envelope that shifts with workload, ambient temperature, and component aging. Instabilities can develop within the bandwidth of the control loops driving them, far faster than any human can intervene. What is needed is continuous system identification: waveform capture at millisecond or sub-millisecond resolution, synchronized across the distribution network, feeding real-time analytics capable of identifying emerging instabilities before they become hardware failures. A voltage transient at one rack correlated with a current spike at the SST tells a different story than either measurement alone. AI data center power infrastructure must be continuously engineered while operating, with feedback loops from monitoring to control tuning to system modification. The boundary between commissioning and ongoing operation needs to be substantially narrowed. One platform that addresses this monitoring requirement directly is Volta Insite. The platform captures full waveforms at 20 kHz and above at every feed and branch circuit simultaneously, providing the sub-millisecond resolution across the distribution network that SST-based stability analysis requires. Domain-trained AI agents perform continuous predictive analysis using LLM-based reasoning, correlating power quality anomalies, voltage ripples at the SST output, current transients at the rack, across the full system before individual events compound into hardware failures. ## A Practical Path Forward 💡 800 V DC data center architectures can deliver real efficiency gains, but only if the industry treats them as dynamic control systems requiring impedance-based stability analysis, coordinated storage controls, GPU-like commissioning tests, and continuous high-speed monitoring from rack to grid. The 800 V DC solid-state transformer architecture is not inherently unstable. It is inherently dynamic and demanding of a level of control-systems rigor that the data center industry has not historically applied to electrical infrastructure. Closing that gap requires several changes in practice. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/06/ChatGPT-Image-Jun-10--2026--10_25_09-AM.png) **Figure 1**: This diagram compares the conventional 480 V AC data center power chain with an SST-based 800 V DC architecture. The 800 V DC path can reduce conversion stages, losses, and equipment footprint, but it also turns AI power delivery into a dynamic control problem as synchronized GPU load swings interact with cables, converters, batteries, and monitoring systems. Impedance-based stability analysis must become a standard design deliverable. Every interface, SST to battery, SST to distribution bus, bus to rack, needs to be characterized across frequency and verified against Middlebrook criteria or more sophisticated equivalents across the full operating envelope. Control loop interactions need to be co-simulated with realistic models of communication latency and sensor noise. Distributed energy storage requires coordinated control from the start, droop sharing, virtual impedance shaping, or explicit bandwidth hierarchy, rather than as a corrective measure applied after integration problems surface. Commissioning protocols need to include dynamic testing. Static load tests confirm that a system can deliver rated power. They say nothing about whether it remains stable while doing so. Step loads, pulse trains, and synthetic GPU-like load signatures should be standard elements of acceptance testing, alongside millisecond-resolution waveform monitoring synchronized across the full distribution network. The efficiency gains of 800 V DC distribution are real and material. But, the transition asks something the industry has not fully acknowledged: that data center electrical infrastructure is no longer a static distribution problem but a dynamic control problem, operating at power levels and response times where the margin for error is measured in megawatts and the timescale of failure is measured in milliseconds. The GPUs driving AI's computational expansion breathe in synchronized rhythms that propagate through power electronics, cables, and bus bars to the utility grid itself. Managing that breathing requires a synthesis of power electronics, control theory, and operational intelligence that goes substantially beyond what has historically been asked of the people who build and run data centers. The infrastructure that rises to that challenge will underpin the next decade of AI development. ## Sign up for AIxEnergy AIxEnergy helps leaders understand the physical economy of artificial intelligence by connecting energy, infrastructure, markets, policy, and governance into clear strategic insight. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### The Cyber-Physical Frontier of AI URL: https://www.aixenergy.io/cyber/ Last updated: 2026-08-05T03:29:55.000Z On June 2, 2026, the United States crossed a quiet but important threshold in the governance of artificial intelligence. The federal government created a voluntary pathway for leading AI developers to provide pre-release access to advanced models for national security and cybersecurity review. The review window was limited, the framework was not mandatory, and the order did not create a licensing regime for frontier AI. Yet the signal was clear: the most capable AI systems are no longer being treated only as commercial products. They are beginning to be treated as infrastructure-adjacent technologies with national consequences. Article Summary 0:00 /79.680748 1× The immediate issue was not synthetic consciousness, autonomous weapons, or any of the more theatrical fears that often dominate public discussion of AI. The immediate issue was software. More precisely, it was the growing ability of frontier AI systems to discover, reason about, and potentially operationalize software vulnerabilities at a speed and scale that existing institutions were not built to absorb. A model that helps a trusted security team find unknown defects in critical software can be a defensive breakthrough. The same model, released without adequate controls or stolen from its developer, can become a force multiplier for adversaries. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/06/ChatGPT-Image-Jun-4--2026--03_01_04-PM.png) That is the dual-use problem in its most practical form. It is also why the issue belongs at the center of the AIxEnergy conversation. The public debate still treats AI and energy mainly as a load-growth story: data centers need power, chips need cooling, transmission queues are filling, utilities are revising forecasts, and large technology companies are searching for firm electricity supplies. Those pressures are real. They are already reshaping utility planning, generation procurement, gas infrastructure, nuclear strategy, hydropower contracting, behind-the-meter development, and demand flexibility. But the load-growth story is only the visible edge of a deeper transition. AI is not merely consuming electricity. It is beginning to enter the operating layer of the systems that produce, deliver, price, protect, and govern electricity. The grid is becoming more digital at the same moment AI is becoming more capable. That convergence changes the meaning of cybersecurity. Cybersecurity is no longer only about protecting networks from intrusion. It is increasingly about preserving the integrity of the machine layer beneath modern life. ## The Old Bargain: Fast Software, Deferred Security Modern civilization rests on software that was not designed for the weight it now carries. Banks, hospitals, water systems, airports, pipelines, telecom networks, public agencies, data centers, cloud platforms, manufacturing plants, and electric utilities all depend on code assembled over decades. Some of that code is modern, tested, and well maintained. Much of it is not. It includes legacy systems, vendor packages, open-source libraries, embedded firmware, remote-access tools, identity platforms, cloud services, operational databases, field devices, scripts, and interfaces that were never built for a permanently contested digital environment. For years, the software economy lived under a tacit bargain. Build quickly, add features, ship updates, expand functionality, and patch later. Cybersecurity became the compensating industry around that bargain. Firewalls, endpoint protection, vulnerability scanners, managed detection, threat intelligence, penetration testing, incident-response retainers, cyber insurance, identity platforms, and compliance programs grew around the fact that the underlying software base remained structurally fragile. The premise was not that the system was secure. The premise was that insecurity could be managed. AI challenges that premise because it changes the economics of finding flaws. Historically, serious vulnerability discovery required scarce expertise. A capable researcher needed to understand the target, inspect code or binaries, infer system behavior, test inputs, trace failure conditions, and determine whether a weakness could be exploited. Automation helped, but high-end vulnerability research still depended on specialized human judgment, patience, and domain knowledge. Frontier AI weakens those constraints. A sufficiently capable model can analyze large codebases, identify suspicious implementation patterns, reason across dependencies, generate test cases, inspect authentication flows, compare code against known weakness classes, and assist with exploitability analysis. It may not replace elite human researchers, but it can multiply them. It can also give less capable actors access to methods they could not previously use. That is the central cyber transition. Vulnerability discovery is moving from expert craft toward industrial-scale cognitive automation. The first-order effect is more findings. The second-order effect is shorter time between discovery and exploitation. The third-order effect is institutional overload. It is one thing for AI to identify thousands of weaknesses. It is another for vendors, hospitals, utilities, banks, water systems, and public agencies to validate, prioritize, patch, and monitor those weaknesses before adversaries act. This is where the issue stops being only technical. Discovery can move at model speed. Remediation still moves at organizational speed. The gap between those two speeds may become one of the defining infrastructure risks of the AI era. ## The Vulnerability Window is Becoming the Critical Variable Every vulnerability has a life cycle. It begins as a latent defect, becomes a security issue when someone discovers it, becomes an operational problem when someone can exploit it, becomes systemic when the affected software is widely deployed, and becomes a crisis when defenders cannot patch before attackers move. AI does not create that cycle. It compresses it. In the pre-AI cyber model, defenders could still rely on friction. Attackers needed time, skill, infrastructure, target knowledge, and operational discipline. Even when criminal groups or state actors had advanced capabilities, the work still had bottlenecks. AI reduces those bottlenecks by making reconnaissance, code analysis, exploit-hypothesis generation, phishing, malware variation, configuration review, and vulnerability triage faster and cheaper. This does not mean every AI-assisted operation will succeed. Cyber operations still require access, persistence, command infrastructure, operational security, target selection, and execution. But AI shifts the frontier by increasing the number of actors who can attempt sophisticated work, increasing the number of systems that can be scanned intelligently, and increasing the pressure on defenders who already struggle with patch backlogs, asset inventories, vendor dependencies, and operational constraints. The most important metric in AI-era cybersecurity may become the vulnerability window: the period between the moment a serious flaw becomes discoverable and the moment it is effectively mitigated across exposed systems. That window is especially hard to close in critical infrastructure. A consumer application can often be patched rapidly. A substation environment, hospital network, water treatment plant, pipeline control environment, or grid operations platform may require vendor coordination, safety validation, scheduled maintenance windows, backup procedures, operator training, regulatory awareness, and careful testing before changes can be made. The defender’s paradox is therefore acute. AI can help find more vulnerabilities, but finding more vulnerabilities does not automatically make society safer. It can simply enlarge the known-risk backlog unless the institutions responsible for repair become faster, better coordinated, and better resourced. A world that accelerates discovery without accelerating remediation does not become more secure. It becomes more aware of its fragility. This is why the June 2026 policy signal matters. The issue is not whether one executive order is sufficient. It is not. The deeper point is that government and industry are beginning to recognize that frontier AI cyber capability cannot be treated as an ordinary software release. If a model can materially improve the discovery of dangerous vulnerabilities in widely used software, then release discipline becomes an infrastructure question. ## Why the Electric Grid is Uniquely Exposed The electric system is not simply one critical infrastructure sector among many. It is the enabling platform beneath the others. Hospitals need electricity. Water systems need electricity. Telecom networks need electricity. Emergency response needs electricity. Fuel logistics need electricity. Financial systems need electricity. Data centers need electricity. Modern government needs electricity. The digital economy sits on top of the power system. That makes grid cybersecurity different in kind. A cyber incident in a single company can be costly, disruptive, and damaging. A cyber incident that affects electricity operations can propagate into public safety, economic continuity, fuel supply, communications, emergency response, and national security. The grid is also becoming more complex at the same moment AI demand is creating new load growth, new interconnection pressure, new data center clusters, and new operational volatility. The system itself is changing. It is moving from a centralized machine built around large thermal and hydro plants toward a more distributed, inverter-rich, data-intensive, software-mediated system. Solar, wind, batteries, electric vehicles, heat pumps, flexible buildings, smart meters, grid-enhancing technologies, distributed energy resources, data centers, and advanced forecasting systems all increase the number of digital interfaces involved in planning and operations. Each interface creates value. Each also creates exposure. The traditional grid was already cyber-physical. SCADA systems, energy management systems, distribution management systems, protective relays, outage management platforms, market systems, telecommunications networks, and field devices have long connected digital systems to physical consequences. But the AI-native grid adds a new layer: probabilistic, adaptive, data-hungry intelligence that may influence planning, operations, maintenance, markets, and emergency response. This is not inherently bad. AI may be necessary to manage the complexity of the modern grid. The system is becoming too dynamic for purely deterministic tools and slow planning cycles. Operators need better forecasting, anomaly detection, asset intelligence, situational awareness, and decision support. Utilities need tools that can help interpret weather volatility, distributed resource behavior, customer load changes, equipment conditions, cyber alerts, and data center demand at speeds traditional workflows cannot match. But AI also changes the control surface. A load forecast is not just a forecast if it informs dispatch, procurement, demand response, emergency operations, or capital planning. A distributed energy resource orchestration platform is not just software if it can shape megawatt-scale load behavior. A data center flexibility system is not just a commercial optimization tool if it can respond to price or grid signals at scale. An AI assistant is not just an assistant if it can read internal documents, query operational databases, draft configuration changes, open tickets, recommend switching actions, or trigger workflows. The practical distinction is not between AI and non-AI. It is between advisory AI and agentic AI. It is between systems that inform humans and systems that act through permissions. This is where many energy organizations will underestimate the risk. They will evaluate the model when they should evaluate the model’s authority. A modest model with broad access may be more dangerous than a powerful model confined to a sandbox. The question is not only how capable the AI is. The question is what the AI can touch. ## The New Risk Stack: From Prompt Injection to Physical Consequence AI systems introduce a risk stack that traditional cybersecurity programs were not built to manage. The first layer is prompt injection. Large language models blur the boundary between instructions and data. A malicious instruction hidden inside an email, webpage, document, ticket, code comment, vendor advisory, or retrieved file can manipulate the model’s behavior. In a passive chatbot, this may produce a bad answer. In an agentic system connected to tools, it can cause data exposure, unsafe actions, or compromised workflows. The second layer is insecure output handling. AI-generated output may be treated as trusted when it is not. If a model writes code, SQL, shell commands, configuration files, policy language, relay-setting suggestions, or control recommendations, those outputs must be validated before execution. Otherwise, the model becomes an untrusted code generator operating inside a trusted environment. The third layer is training data poisoning and retrieval poisoning. AI systems depend on data. If training data, fine-tuning data, embeddings, retrieval sources, logs, tickets, telemetry, inspection records, market data, or asset records are manipulated, model outputs can be corrupted. In the energy sector, poisoned data can distort load forecasts, asset health models, wildfire risk tools, DER coordination systems, interconnection screening, or cyber defense triage. The fourth layer is excessive agency. This is the risk created when AI systems are granted too much autonomy or too many permissions. An AI agent that can read, write, execute, approve, connect, or modify systems can create consequences beyond the intent of its designers. The problem is not intelligence alone. It is intelligence plus authority. The fifth layer is model theft. A frontier model, fine-tuned model, prompt architecture, tool configuration, embedding database, or energy-specific AI system may encode sensitive knowledge. If stolen, it may reveal operational patterns, system dependencies, code vulnerabilities, asset conditions, defensive methods, or vendor relationships. A compromised model can become an adversary’s map. The sixth layer is overreliance. Humans may trust AI outputs because they are fluent, fast, and confident. In grid operations, this is especially dangerous. An AI recommendation can be useful without being authoritative. Operators need decision support, not decision displacement. These risks are no longer speculative. NIST’s AI Risk Management Framework gives organizations a structure for governing, mapping, measuring, and managing AI risk. NIST’s Generative AI Profile extends that structure to the distinctive risks of generative systems. NIST’s 2026 critical-infrastructure profile concept points toward the next stage: translating AI risk management into the operational context of infrastructure sectors. OWASP’s GenAI security work identifies concrete LLM application vulnerabilities, including prompt injection, insecure output handling, training data poisoning, excessive agency, overreliance, supply-chain vulnerabilities, and model theft. The energy sector should not read these frameworks as compliance paperwork. It should read them as early maps of a new operating terrain. The danger is not that AI will suddenly make the grid unmanageable. The danger is that energy organizations will adopt AI into critical workflows faster than they adapt their cyber, data, governance, and operating disciplines. ## The Cyber-Physical Clearinghouse Problem The hardest problem in AI-enabled vulnerability discovery is not discovery. It is coordinated repair. Consider a plausible case. A frontier AI system identifies a serious vulnerability in a software component used across hospital backup-power systems, municipal utility control environments, building automation systems, and distributed energy resource platforms. The model flags a pattern. Researchers validate the issue. The vendor confirms the flaw. The affected product versions are widespread. What happens next cannot be left to improvisation. The vulnerability must be validated. False positives must be removed. Affected versions must be identified. Vendors must develop patches or mitigations. Asset owners must know whether they are exposed. Critical infrastructure operators must receive enough information to act, but not so much that the vulnerability is handed to attackers. Smaller organizations may need direct technical assistance. Public disclosure must be timed. Exploitation must be monitored. Emergency mitigations may be needed before patches are available. This is not a model capability problem. It is an institutional design problem. AI creates the possibility of discovering vulnerabilities faster than society can fix them. That means any serious defensive AI strategy must include a clearinghouse function: a trusted mechanism for moving from vulnerability discovery to validation, prioritization, remediation, and monitoring across sectors. The clearinghouse cannot be merely a reporting inbox. It needs technical expertise, secure channels, sector-specific triage, vendor coordination, legal clarity, disclosure protocols, and support for resource-constrained infrastructure operators. It must understand that a vulnerability in a consumer application and a vulnerability in a rural hospital network are different kinds of events. It must also understand that a small municipal utility may not have the staff, budget, or test environment to respond like a major investor-owned utility or cloud provider. This is where national cybersecurity and energy policy meet. A society that accelerates AI-enabled vulnerability discovery without building remediation capacity may increase awareness while leaving risk unresolved. That is not resilience. It is illuminated fragility. ## Release Discipline for Frontier Cyber-Capable Models The phrase “AI regulation” is too broad to be useful here. The better term is release discipline. Not every AI system requires the same controls. A small model summarizing public documents is not equivalent to a frontier model that can discover unknown vulnerabilities in widely used software. A model with no tools is not equivalent to an agent that can scan networks, write code, call APIs, execute commands, or interact with production systems. A research model available to a controlled group is not equivalent to a widely released system that can be copied, fine-tuned, wrapped, jailbroken, or stolen. The right governance architecture should be capability-based. The highest scrutiny should apply to systems that can materially improve offensive cyber capability, especially against critical infrastructure, cloud platforms, identity systems, operational technology, widely used open-source packages, firmware, or software supply chains. Evaluation should ask practical questions. Can the model identify previously unknown vulnerabilities in real codebases? Can it reason from a bug to an exploit path? Can it chain multiple weaknesses? Can it generate reliable proof-of-concept exploit code? Can it automate reconnaissance? Can it improve phishing, malware development, privilege escalation, persistence, or lateral movement? Can it increase the capability of a low-skilled actor? Can it operate through tools without close human supervision? If the answer is yes, then the release process should not look like an ordinary product launch. It should include structured red teaming, staged access, abuse monitoring, secure development practices, independent evaluation for high-consequence capabilities, vulnerability disclosure coordination, tool-use constraints, model security controls, and post-release monitoring. The strongest capabilities may need tiered access limited to trusted defenders, researchers, vendors, and infrastructure operators. This does not mean freezing innovation. It means recognizing that some AI capabilities are dual-use in the classic sense. They can support legitimate defense, research, and software improvement. They can also cause harm if distributed carelessly. The governance challenge is not to suppress the capability, but to control the pathway by which it enters the world. ## Energy Organizations Need an AI Cyber-Control Plane Energy companies should not wait for perfect federal rules. The operational exposure is already arriving through vendors, employees, cloud platforms, analytics tools, grid modernization programs, and informal use of public AI systems. The first task is inventory. Every energy organization needs to know where AI is being used, what data it touches, what systems it connects to, what vendors are involved, whether outputs are used in decisions, and whether the AI system can take actions. Shadow AI is now a material enterprise risk. The second task is consequence classification. AI tools should be tiered by operational consequence. Low-risk tools can move quickly. High-consequence tools need deeper review. Any AI system connected to grid operations, cyber defense, code repositories, identity systems, market participation, customer data, operational technology, engineering records, or emergency response should receive elevated scrutiny. The third task is permission design. AI systems should have their own identities. They should not casually inherit broad human credentials. They should operate under least privilege. Their access should be scoped, logged, time-limited, and revocable. High-impact actions should require human approval. Read-only should be the default posture for operational environments unless there is a strong reason to move beyond it. The fourth task is runtime containment. AI agents should operate in sandboxes. Tool access should be constrained. External content should be treated as untrusted. Outputs should be validated before execution. The system should separate instructions from data wherever possible, even though large language models do not naturally enforce that boundary. The architecture should assume prompt injection will occur and limit the blast radius. The fifth task is data integrity. AI models are only as trustworthy as the data systems around them. Energy organizations need provenance, access control, anomaly detection, data quality checks, change logs, and clean recovery paths for training data, retrieval systems, telemetry, asset records, inspection data, and operational feeds. The sixth task is vendor assurance. Vendors should disclose AI functionality, model dependencies, data retention practices, tool permissions, update processes, security testing, incident notification procedures, and subcontractor exposure. A vendor adding AI to a grid-relevant product is not merely adding a feature. It is changing the risk profile of the product. The seventh task is incident response. Energy organizations need playbooks for AI-specific cyber events: prompt injection, model compromise, data leakage, poisoned retrieval systems, unsafe agent behavior, malicious code generation, vendor AI breach, model theft, and AI-enabled social engineering. Recovery plans should include disabling AI functions, revoking credentials, rolling back model updates, restoring clean data, and communicating with regulators and partners. This is the energy-sector version of AI governance. It is not a board memo about ethics. It is an operating architecture. ## Defensive AI May be the Only Way to Keep Pace The answer to AI-enabled cyber risk cannot be less AI. Defenders will need AI to keep pace. AI can help identify vulnerabilities, review code, summarize threat intelligence, correlate logs, prioritize alerts, generate detection rules, assist incident response, inspect infrastructure-as-code templates, translate vendor advisories into asset-specific guidance, and help smaller organizations understand what matters. It can also help bridge the cultural gap between IT security teams and operational technology teams. That bridge is essential in energy. Cyber teams may understand attack chains but not grid operations. Grid operators may understand physical consequences but not cloud identity, software supply chains, or AI-specific vulnerabilities. AI systems, if governed well, can help translate between those domains. They can map a software weakness to an operational consequence. They can identify which substations, control centers, DER platforms, market systems, or vendor dependencies are most exposed. They can support faster triage during the first hours of an incident, when confusion is high and time is scarce. The goal is not to remove humans from the loop. It is to give humans a better loop. In the AI-native grid, the best systems will combine machine-speed analysis with human accountability. AI should shorten the distance from signal to understanding. It should not collapse the distance from recommendation to action without governance. ## The New Doctrine: Govern Agency The central doctrine for critical infrastructure AI should be simple: govern agency, not just intelligence. Model capability matters. But the greater operational risk often comes from the interaction between model capability, data access, tool access, permissions, institutional dependence, and human overtrust. A high-capability model in a sealed research environment may pose less immediate infrastructure risk than a less capable agent with access to production systems, credentials, workflows, and operational data. This doctrine has several implications. AI assurance must move beyond benchmark performance and include authority mapping. What can the system read? What can it write? What can it execute? What can it trigger? What can it expose? What can it change? Those questions matter as much as accuracy scores, latency, or user experience. AI governance must also be continuous. A model update, new plugin, new retrieval source, new workflow integration, or new vendor feature can change the risk profile without changing the product name. AI security must include the supply chain: models, datasets, prompts, embeddings, APIs, tools, agents, cloud infrastructure, hardware, firmware, and vendors all form part of the system. AI risk must also be measured by consequence. A hallucinated travel recommendation is not the same as a hallucinated grid operating recommendation. A data leak from a public marketing tool is not the same as a data leak from an operational technology environment. A prompt injection in a public chatbot is not the same as a prompt injection in an AI assistant connected to engineering records, incident tickets, or control-room workflows. The final requirement is reversibility. AI deployment in infrastructure must preserve the ability to degrade gracefully. Operators should be able to disable AI functions, fall back to conventional tools, revoke credentials, restore clean data, and maintain operations under degraded conditions. The test of AI governance is not whether the system works when everything is normal. It is whether the organization can control it when something goes wrong. ## From Electricity Demand to Intelligent Infrastructure The AI-energy debate began with megawatts. It will not end there. Load growth is real. Data centers are reshaping utility planning. Power availability is becoming a strategic constraint on AI deployment. But the deeper transition is that AI is entering the operating fabric of infrastructure itself. It is becoming part of how systems forecast, optimize, defend, repair, trade, and coordinate. That means the electric grid is no longer only supplying AI. It is being changed by AI. In the old grid, hidden defects were physical: a failing transformer, a cracked insulator, an overloaded line, a mis-set relay, a corroded connector, or a tree too close to a conductor. In the AI-native grid, the hidden defect may be a poisoned dataset, a vulnerable software library, an over-permissioned agent, a compromised model, a manipulated forecast, a prompt injection buried in a maintenance document, or a vendor feature that quietly creates a new pathway into operational systems. The physics of the grid still matter. Frequency, voltage, inertia, reactive power, protection, thermal limits, and stability remain fundamental. But the digital layer around those physics is thickening. The security of that layer now shapes the reliability of the physical system. This is the new infrastructure condition. AI makes software more powerful, and it makes software failure more consequential. It can help defenders repair the digital foundation of modern life. It can also help adversaries find the cracks faster. The outcome will depend less on slogans about innovation or regulation than on whether institutions can build disciplined pathways for release, access, testing, remediation, and operational control. The frontier is not artificial intelligence alone. It is intelligent infrastructure. The first rule of intelligent infrastructure is that cognition must be governed before it is connected to consequence. ## Sign up for AIxEnergy AIxEnergy helps leaders understand the physical economy of artificial intelligence by connecting energy, infrastructure, markets, policy, and governance into clear strategic insight. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## Notes 1. Courtney Rozen, Ismail Shakil, and Bhargav Acharya, “Trump Administration to Ask US AI Firms to Voluntarily Submit Models for Cybersecurity Tests,” Reuters, June 2, 2026. 2. Matt O’Brien, “Trump Signs an Executive Order That Invites Vetting of Top AI Models for National Security Risks,” Associated Press, June 2, 2026. 3. Sanya Mansoor, “Trump Signs Executive Order Seeking Early Access to New AI Releases,” The Guardian, June 2, 2026. 4. National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1 (Gaithersburg, MD: National Institute of Standards and Technology, January 2023), [https://doi.org/10.6028/NIST.AI.100-1](https://doi.org/10.6028/NIST.AI.100-1?ref=aixenergy.io). 5. National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, NIST AI 600-1 (Gaithersburg, MD: National Institute of Standards and Technology, July 2024), [https://doi.org/10.6028/NIST.AI.600-1](https://doi.org/10.6028/NIST.AI.600-1?ref=aixenergy.io). 6. National Institute of Standards and Technology, “Concept Note for an AI RMF Profile on Trustworthy AI in Critical Infrastructure,” April 7, 2026. 7. OWASP Foundation, OWASP Top 10 for Large Language Model Applications, Version 2025, OWASP GenAI Security Project, 2025. 8. MITRE, MITRE ATLAS: Adversarial Threat Landscape for Artificial-Intelligence Systems, accessed June 4, 2026. 9. National Institute of Standards and Technology, The NIST Cybersecurity Framework (CSF) 2.0, NIST CSWP 29 (Gaithersburg, MD: National Institute of Standards and Technology, February 26, 2024), [https://doi.org/10.6028/NIST.CSWP.29](https://doi.org/10.6028/NIST.CSWP.29?ref=aixenergy.io). 10. Miles Brundage et al., “Frontier AI Auditing: Toward Rigorous Third-Party Assessment of Safety and Security Practices at Leading AI Companies,” arXiv, January 16, 2026, revised February 7, 2026, [https://doi.org/10.48550/arXiv.2601.11699](https://doi.org/10.48550/arXiv.2601.11699?ref=aixenergy.io). ## Bibliography Associated Press. “Trump Signs an Executive Order That Invites Vetting of Top AI Models for National Security Risks.” By Matt O’Brien. June 2, 2026. Brundage, Miles, Noemi Dreksler, Aidan Homewood, Sean McGregor, Patricia Paskov, Conrad Stosz, Girish Sastry, et al. “Frontier AI Auditing: Toward Rigorous Third-Party Assessment of Safety and Security Practices at Leading AI Companies.” arXiv. January 16, 2026\. Revised February 7, 2026\. [https://doi.org/10.48550/arXiv.2601.11699](https://doi.org/10.48550/arXiv.2601.11699?ref=aixenergy.io). Mansoor, Sanya. “Trump Signs Executive Order Seeking Early Access to New AI Releases.” The Guardian. June 2, 2026. MITRE. MITRE ATLAS: Adversarial Threat Landscape for Artificial-Intelligence Systems. Accessed June 4, 2026. National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework (AI RMF 1.0). NIST AI 100-1\. Gaithersburg, MD: National Institute of Standards and Technology, January 2023\. [https://doi.org/10.6028/NIST.AI.100-1](https://doi.org/10.6028/NIST.AI.100-1?ref=aixenergy.io). National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. NIST AI 600-1\. Gaithersburg, MD: National Institute of Standards and Technology, July 2024\. [https://doi.org/10.6028/NIST.AI.600-1](https://doi.org/10.6028/NIST.AI.600-1?ref=aixenergy.io). National Institute of Standards and Technology. “Concept Note for an AI RMF Profile on Trustworthy AI in Critical Infrastructure.” April 7, 2026. National Institute of Standards and Technology. The NIST Cybersecurity Framework (CSF) 2.0\. NIST CSWP 29\. Gaithersburg, MD: National Institute of Standards and Technology, February 26, 2024\. [https://doi.org/10.6028/NIST.CSWP.29](https://doi.org/10.6028/NIST.CSWP.29?ref=aixenergy.io). OWASP Foundation. OWASP Top 10 for Large Language Model Applications. Version 2025\. OWASP GenAI Security Project, 2025. Rozen, Courtney, Ismail Shakil, and Bhargav Acharya. “Trump Administration to Ask US AI Firms to Voluntarily Submit Models for Cybersecurity Tests.” Reuters. June 2, 2026. ### Recursive AI and the Cognitive Grid URL: https://www.aixenergy.io/anthropic/ Last updated: 2026-07-13T01:38:03.000Z The governance problem begins before machines take control. It begins when AI starts shaping the decisions humans believe they are still making. _This post is for paying subscribers only._ ### The New Deal: How Hyperscalers’ Partnership Models Are Rewiring Clean-Power Procurement URL: https://www.aixenergy.io/the-new-deal-how-hyperscalers-partnership-models-are-rewiring-clean-power-procurement/ Last updated: 2026-07-29T19:40:13.000Z Hyperscalers are shifting from power buyers to grid architects. Through five partnership models they shape clean energy procurement and grid build-out—deciding whether AI demand locks in fossil fuels or accelerates zero-carbon supply. _This post is for paying subscribers only._ ### The Prompt Box Is Part of the Energy System: Why Sustainable AI Is Not Just a Data Center Problem URL: https://www.aixenergy.io/the-prompt-box-is-part-of-the-energy-system/ Last updated: 2026-07-09T18:02:41.000Z Artificial intelligence has made computation feel almost weightless. A paragraph appears in seconds. A memo rewrites itself. A market scan becomes a briefing note. A half-formed thought becomes a slide, an image, a policy comment, a product description, or a strategic plan before the user has fully decided whether the work needed to exist. Article Summary 0:00 /50.94161 1× That is the magic of generative AI. It is also the trap. The easier it becomes to generate language, images, summaries, code, and analysis, the easier it becomes to confuse activity with value. AI can help people think more clearly, move faster, and reduce wasteful work. It can also help organizations produce more material than anyone needs, at a scale that quietly turns individual convenience into system-level demand. ## The Visible Infrastructure The energy debate around AI has focused, understandably, on visible infrastructure: data centers, substations, transmission corridors, cooling systems, backup generation, power purchase agreements, land use, water consumption, and carbon emissions. That focus is justified. Data centers are no longer a quiet corner of the commercial building stock. They are becoming a new class of industrial load: fast-growing, geographically concentrated, power-hungry, and politically consequential. The numbers explain why. The International Energy Agency estimates that global data center electricity consumption was about 415 terawatt-hours in 2024, or roughly 1.5 percent of global electricity use. By 2030, the IEA projects that data center electricity demand could more than double to around 945 terawatt-hours, slightly more than Japan’s total electricity consumption today.¹ In the United States, the increase is already visible. Lawrence Berkeley National Laboratory estimates that U.S. data center electricity use rose from 58 terawatt-hours in 2014 to 176 terawatt-hours in 2023, and could reach 325 to 580 terawatt-hours by 2028\. Depending on overall electricity demand growth, that would put data centers at roughly 6.7 to 12 percent of total U.S. electricity consumption by 2028.² This is no longer an abstract sustainability concern. It is showing up in real places, on real grids, and in real capital commitments. On May 26, 2026, Reuters reported that I Squared Capital acquired ten U.S. data center facilities from Cogent Fiber for $225 million, with plans to commit another $1 billion for upgrades, expansion, and acquisitions. The facilities provide 53 megawatts of power capacity across nine markets and are aimed in part at the growing demand for AI inference: the everyday use of trained AI models closer to customers, rather than only the massive centralized training runs that dominate the public imagination.³ That distinction matters. The AI infrastructure buildout is not only about a few giant training campuses. It is also about the diffusion of inference across the economy: more queries, more applications, more embedded AI features, more automated workflows, and more computation hidden inside ordinary digital life. The large training run may be the dramatic event, but inference is the habit. And routine is where demand accumulates. ## The Invisible Infrastructure of Use The infrastructure story is only half the story. The other half begins at the prompt box, where millions of small human decisions determine how much useful work AI performs and how much low-value computation it creates. AI is usually described as a technology. That description is incomplete. AI is also a behavioral system. It changes how people ask questions, produce documents, make decisions, avoid decisions, communicate authority, and scale work. Its environmental footprint is shaped not only by chips, cooling towers, transformers, and clean-energy contracts, but also by habits. The physical infrastructure matters enormously. So does the culture of use that infrastructure enables. The analogy is the automobile. The environmental impact of cars does not end at the factory gate. Manufacturing a car matters. Fuel efficiency matters. Battery chemistry matters. Road infrastructure matters. But so does how people drive. A household that combines trips, avoids unnecessary travel, maintains the vehicle, and chooses walking or transit when appropriate produces a different environmental outcome than a household that treats every impulse as a separate car trip. AI is moving through the economy in a similar way. Model efficiency matters. Data center design matters. Clean electricity procurement matters. Water stewardship matters. But use also matters. A prompt can substitute for work that would otherwise require more time, travel, meetings, analysis, or administrative overhead. It can also multiply work that never needed to exist. ## Substitution or Expansion? That distinction may become one of the most important sustainability questions in the AI economy: is AI substituting for activity, or expanding it? If AI substitutes intelligently, it can be an efficiency tool. It can compress research time, improve building operations, reduce unnecessary meetings, optimize logistics, support grid planning, and improve the quality of consequential decisions. If it expands indiscriminately, it becomes a rebound machine. Energy economists have long understood the rebound effect. When a technology becomes more efficient, the cost of using it often falls. When the cost falls, people often use more of it. Sometimes efficiency gains dominate, and society still saves energy overall. Sometimes additional use erodes the savings. AI has its own version of this problem. As the cost of generating language, images, code, summaries, and analysis approaches zero, the amount of generated output can explode. That is the paradox of AI sustainability. The environmental impact of a single prompt may be small, and in some systems it is becoming smaller. Google recently estimated that a median Gemini Apps text prompt consumes about 0.24 watt-hours of electricity, emits 0.03 grams of carbon dioxide equivalent, and consumes 0.26 milliliters of water.⁴ Those numbers are far smaller than many popular estimates of AI use, but they do not settle the system question. A single prompt is not the system. The system is the number of prompts, the complexity of the model invoked, the length of the context window, the length of the output, the hardware used, the data center’s power usage effectiveness, the cooling method, the local water context, the carbon intensity of electricity at the time of computation, and the downstream work the AI output causes people to create. A drop of water is negligible. A culture of infinite prompting is not. ## Efficiency Is Not Enough For the energy sector, this requires a conceptual shift. We tend to think in terms of supply-side interventions: build cleaner generation, improve data center efficiency, site loads near low-carbon resources, recover waste heat, add storage, strengthen transmission, and use advanced cooling. All of that matters. A data center with a power usage effectiveness, or PUE, of 1.1 uses far less overhead energy for cooling and power distribution than a facility with a PUE of 1.6\. PUE measures total facility energy divided by IT equipment energy. A PUE of 1.1 means that for every 1.0 megawatt used by servers, about 0.1 megawatts are used for supporting infrastructure. A PUE of 1.6 means 0.6 megawatts of overhead for every 1.0 megawatt of IT load. That difference matters at hyperscale. But technical efficiency alone cannot settle the question. If each unit of computation becomes cleaner and cheaper while total computation grows faster, absolute electricity demand can still rise. The result is familiar across energy history: efficiency improves the machine, while behavior determines the system. AI will be no different. ## Demand Architecture Sustainable AI needs a better vocabulary. The next frontier is not only green data centers. It is demand architecture: the deliberate design of interfaces, workflows, defaults, procurement standards, and organizational norms that shape how much AI computation is actually used and whether that use creates value. Today’s AI interfaces are optimized for ease, speed, and engagement. They invite the user to ask anything, regenerate endlessly, and produce more. That is understandable from a product-growth perspective, but from an energy-systems perspective the interface is not neutral. A tool that encourages vague prompting, repeated regeneration, maximal output, and low-purpose content is shaping load. The prompt box is an energy device. That may sound strange, but so did the idea that thermostats were grid assets. Then utilities learned that millions of small behavioral signals — a degree of cooling here, a delayed water heater there, an electric vehicle charging session shifted by two hours — could add up to system-level consequences. AI usage will follow a similar pattern. No single user is the grid. But aggregated behavior becomes demand. ## Low-Value AI Use at Scale The sustainability field risks making this too moralistic. The point is not to shame people for using AI. Used well, AI can reduce waste. It can help planners understand interconnection queues, help building operators detect drift, help utilities summarize regulatory filings, help engineers compare scenarios, and help organizations move faster with fewer meetings and less duplication. The problem is not AI use. The problem is low-value AI use at scale. A utility planner who uses AI to identify inconsistencies across load forecasts may improve institutional learning. A sustainability team that uses AI to synthesize hundreds of pages of disclosure requirements may save real labor. A project developer who uses AI to compare interconnection alternatives may reach better decisions faster. A writer who uses AI to test an argument, sharpen structure, and identify missing evidence may produce clearer work with fewer wasted cycles. The same tool can also create institutional drag. It can generate ten versions of a memo nobody needed. It can flood inboxes with synthetic prose. It can replace judgment with volume. It can encourage more reports, more dashboards, more updates, more drafts, and more meetings simply because the marginal cost of producing them has collapsed. That is not intelligence. It is throughput without discipline. ## The Substitution Test The sustainability question is not simply, “Should I use AI?” It is, “What is this AI use replacing?” If AI replaces unnecessary travel, duplicative analysis, low-value administrative labor, or a week of fragmented research, the environmental ledger may be favorable. If AI merely adds more content to a world already drowning in content, the ledger changes. The output may be digital, but the inputs are physical: electricity, cooling, water, land, chips, steel, concrete, substations, transformers, and generation capacity. AI has made language cheap. It has not made infrastructure free. ## Four Principles for Purposeful AI Use My friend and colleague [Zach Ayvazian](https://www.linkedin.com/in/zachary-ayvazian-4831b322b/?ref=aixenergy.io), Founder and CEO of SureWay AI, has developed a simple but powerful set of principles for using AI with more discipline. They are not rules of abstinence. They are rules of better use: a way to make AI more useful, less wasteful, and more aligned with the work that actually needs to be done. ### 1\. Iterate Instead of Regenerating “Try again” is the AI equivalent of throwing away a nearly finished draft and ordering a new one. A better approach is to say what should change: make it shorter, preserve the argument, add a counterexample, remove jargon, strengthen the opening, or test the claim against evidence. Iteration keeps the work anchored. Regeneration often produces novelty without improvement. ### 2\. Right-Size the Task Not every task warrants AI. If the answer is obvious, write it yourself. If the problem requires judgment rather than generation, use AI as a sparring partner, not a substitute. If the output will not be read, used, or acted upon, do not generate it. The greenest AI output is the one that never needed to exist. ### 3\. Use AI to Reduce Work Before prompting, ask whether the output will replace something you would otherwise have done or simply create a new obligation. The best use cases compress effort, improve decisions, or reduce downstream waste. The worst ones create synthetic work products that require more human review, more coordination, and more storage while adding little value. ### 4\. Move From Personal Habit to Platform Design These principles should not remain personal habits. They should become design principles for AI platforms, procurement standards, and institutional norms. Enterprise AI systems could make model selection more transparent, routing simple tasks to smaller models and reserving frontier models for higher-value work. Interfaces could encourage clarification before generation. Tools could display approximate energy or carbon ranges, not as scolding devices, but as feedback. Organizations could define when to summarize, when to draft, when to search, when to use retrieval, when to use a smaller model, and when not to use AI at all. This matters because inference is likely to become more distributed, embedded, and ordinary. The next wave of AI demand may not look like one dramatic supercomputer. It may look like millions of small decisions made by people and software agents every hour: summarize this, rewrite that, generate another image, compare these options, draft a response, search again, make it more polished, make it more executive, make it more human. ## From Personal Habit to Platform Design These principles should not remain personal habits. They should become design principles for AI platforms and procurement standards for institutions. Enterprise AI systems could make model selection more transparent, routing simple tasks to smaller models and reserving frontier models for higher-value work. Interfaces could encourage clarification before generation. Tools could display approximate energy or carbon ranges, not as scolding devices, but as feedback. Organizations could define when to summarize, when to draft, when to search, when to use retrieval, when to use a smaller model, and when not to use AI at all. This matters because inference is likely to become more distributed, embedded, and ordinary. The next wave of AI demand may not look like one dramatic supercomputer. It may look like millions of small decisions made by people and software agents every hour: summarize this, rewrite that, generate another image, compare these options, draft a response, search again, make it more polished, make it more executive, make it more human. Each action is small. The pattern is not. ## Purposeful Computation There is a deeper lesson here for the energy transition. Sustainability is rarely achieved by improving one component in isolation. A more efficient car can still increase vehicle miles traveled. A more efficient light bulb can still contribute to more illuminated space. A more efficient AI model can still enable a flood of low-value computation. This does not mean efficiency is futile. It means efficiency must be paired with governance, design, and culture. For AI, the physical system and the behavioral system are now coupled. Data center developers, utilities, regulators, hyperscalers, enterprise customers, and individual users are all part of the same demand chain. The server hall and the prompt box belong to one system. One is made of concrete, copper, silicon, water loops, switchgear, and fiber. The other is made of habits, incentives, defaults, deadlines, curiosity, anxiety, and ambition. The environmental cost of AI will not be solved only by cleaner electrons or better chips. It will also be shaped by whether we learn to use intelligence intelligently. That is the standard AI now has to meet: not less use, but better use; not digital abundance for its own sake, but purposeful computation; not endless generation, but substitution, judgment, and restraint. AI is here to stay. The question is whether we build a culture of use worthy of the infrastructure now being built to support it. ## Sign up for AIxEnergy AIxEnergy analyzes how artificial intelligence is reshaping electricity demand, grid operations, and energy infrastructure, with a focus on system risk, policy, and institutional decision-making. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## Endnotes 1. International Energy Agency, *Energy and AI* (Paris: IEA, 2025), “Energy Demand from AI,” accessed May 26, 2026\. The IEA estimates global data center electricity consumption at about 415 TWh in 2024 and projects demand could reach about 945 TWh by 2030, slightly more than Japan’s current electricity consumption. 2. Arman Shehabi et al., *2024 United States Data Center Energy Usage Report* (Berkeley, CA: Lawrence Berkeley National Laboratory, December 2024). The report estimates U.S. data center electricity use at 176 TWh in 2023 and projects 325–580 TWh by 2028, equal to roughly 6.7–12 percent of total U.S. electricity consumption depending on demand growth. 3. Milana Vinn, “I Squared Bets on AI Inference with $225 Million Data Center Buy from Cogent,” *Reuters*, May 26, 2026\. The deal includes ten U.S. data center facilities across nine markets, 53 MW of power capacity, and a planned $1 billion commitment for upgrades, expansion, and acquisitions. 4. Jeff Dean et al., “Measuring the Environmental Impact of AI Inference,” Google Cloud Blog, August 21, 2025\. Google estimates that a median Gemini Apps text prompt consumes 0.24 Wh of electricity, emits 0.03 gCO₂e, and consumes 0.26 mL of water. ### East Asia’s AI-Energy Moment: The Cognitive Grid Meets the Shadow Grid URL: https://www.aixenergy.io/asia/ Last updated: 2026-07-10T17:29:43.000Z The first map of the digital economy was drawn in fiber. It followed submarine cables, cloud regions, software talent, tax regimes, enterprise customers, and latency-sensitive corridors. Electricity mattered, but it mostly sat in the background, treated as an input the digital economy could quietly absorb. Artificial intelligence is breaking that settlement. A modern AI campus is not merely a data center. It is a multi-hundred-megawatt industrial load, a cooling system, a water question, a substation project, a transmission-planning problem, a semiconductor demand signal, a permitting case, and a political test. At gigawatt scale, it begins to look less like digital real estate and more like a new industrial district: one that must be energized, cooled, connected, financed, permitted, and publicly justified before a single model can be trained or a single inference served. That is the larger meaning of [*Energy and AI in East Asia*](https://iea.blob.core.windows.net/assets/431d7d22-039e-46d3-9832-090a8b4075fc/EnergyandAIinEastAsia.pdf?ref=aixenergy.io), the new report prepared jointly by the International Energy Agency and the Korea Energy Economics Institute. The report is not simply a regional supplement to the global debate over data-center electricity demand. It is a field guide to one of the places where the AI-energy transition becomes fully visible. East Asia is where the semiconductor supply chain, data-center growth, power-system constraints, renewable integration challenges, energy security concerns, and industrial-policy ambitions converge. 0:00 /1:38 1× In my earlier AIxEnergy article, “[The New Geography of AI Infrastructure](https://www.aixenergy.io/the-new-geography-of-ai-infrastructure/),” I argued that global AI capacity is no longer following only the logic of chips, fiber, cloud regions, and customers. It is being reorganized by power availability, water, cooling, grid reliability, permitting, transformers, substations, political legitimacy, and what I called the Shadow Grid: the emerging layer of parallel infrastructure forming around AI campuses when the public grid alone cannot deliver the speed, certainty, or durability that AI developers require. The IEA/KEEI report gives that thesis an East Asian geography. It shows that the AI-energy story is not one story but two. On one side is AI for energy: artificial intelligence as a tool for renewable siting, power-system forecasting, grid operation, predictive maintenance, demand response, industrial optimization, and energy innovation. On the other side is energy for AI: the need to power the data centers, chip supply chains, cooling systems, backup systems, and grid connections that make AI physically possible. This duality matters because AI is becoming both the optimizer and the load. It is the cognitive layer that can help operate more complex power systems, and it is also one of the forces making those power systems harder to plan and operate. It can help forecast renewable generation, detect equipment anomalies, optimize virtual power plants, and improve demand-side flexibility. But it also brings new loads that can concentrate quickly, move strategically, and overwhelm traditional planning assumptions. That is the defining tension of the AI-energy era. Artificial intelligence may help build the Cognitive Grid. But the growth of AI infrastructure is also accelerating the emergence of the Shadow Grid. ## East Asia as the Physical Stack of AI East Asia matters because it is one of the few regions where the entire physical stack of AI is visible at once. The report notes that the region plays a central role in global semiconductor manufacturing. This is not a secondary detail. The AI economy begins long before a model is trained. It begins with high-purity quartz, silicon wafers, lithography, foundries, packaging, accelerators, servers, storage systems, networking equipment, cooling equipment, and power electronics. The “cloud” is not weightless. It is a chain of physical conversions. East Asia sits near the center of that chain. The region manufactures much of the chip capacity on which the AI economy depends. It also hosts fast-growing AI adoption, national AI strategies, dense industrial and metropolitan loads, and constrained electricity systems. Japan and Chinese Taipei are island systems. Korea is physically a peninsula but electrically closer to an island because it lacks cross-border interconnections. China is the exception: a continental-scale system with far greater internal grid breadth, though still facing its own transmission, siting, and regional-balancing challenges. That makes East Asia not merely a participant in the AI race but one of its proving grounds. The region is both producing the hardware foundations of AI and confronting the electricity consequences of AI deployment. In this sense, the IEA/KEEI report extends the “new geography” argument. The old map of AI advantage was built around chips, models, software talent, and cloud regions. The new map is built around the ability to turn electricity into intelligence. That requires not just data centers, but credible delivery of power, cooling, water, fiber, land, permits, equipment, and public trust. ## From Cloud Geography to Power-System Geography The report’s most important contribution is that it treats data-center growth as a power-system issue, not merely a digital-infrastructure issue. This distinction is essential. Data centers may represent a modest share of global electricity consumption in aggregate, but electricity systems do not operate on global averages. They operate through local substations, transmission corridors, feeders, reserve margins, voltage constraints, interconnection queues, and utility planning processes. A data center does not connect to “global demand.” It connects to a particular node on a particular grid. That is why AI load is becoming geographically disruptive. It is not evenly distributed. It clusters near fiber, customers, cloud ecosystems, labor pools, and metropolitan demand. But those are not always the places with the most available power, the strongest transmission, the easiest permitting, or the best cooling conditions. Korea illustrates the problem sharply. The report notes that the Seoul Metropolitan Area consumes a large share of national electricity and hosts most of the country’s data centers, even though renewable generation potential and surplus generation are more prominent in other regions. Jeolla has substantial solar and wind resources, yet a much smaller share of data centers. The result is a spatial mismatch between where clean energy potential exists and where digital load wants to locate. This is the East Asian version of the global pattern described in “The New Geography of AI Infrastructure.” AI compute wants to follow latency, customers, capital, and ecosystem density. Power systems want new loads to follow available generation, unconstrained networks, and deliverable capacity. The two maps rarely align. That mismatch is where policy begins. Korea’s consideration of locationally differentiated grid connection tariffs for data centers is therefore more than a technical pricing adjustment. It is an early signal that governments are beginning to treat data-center siting as a system-planning problem. If new AI loads locate in congested metropolitan regions, they can intensify grid stress. If they locate near available generation, lower-congestion regions, or renewable-rich areas, they may help rebalance the system. The same question now faces every major AI economy: should compute be allowed to locate wherever the digital market wants, or should grid conditions, cost causation, water availability, emissions impact, and reliability value shape where it goes? The answer will define the next generation of AI infrastructure policy. ## The Cognitive Grid: AI as the Operating Layer The first half of the IEA/KEEI report examines AI for energy, and this is where it connects directly to the Cognitive Grid. The Cognitive Grid is not an autonomous grid in which machines replace operators. It is a more intelligent grid in which forecasting, sensing, simulation, anomaly detection, optimization, and decision support improve the speed and quality of human and institutional judgment. It is the layer that helps grid operators, utilities, regulators, developers, and customers see a more complex system with greater clarity. The report’s examples point in this direction. AI can support renewable-energy siting by integrating geospatial data, weather patterns, ecological constraints, grid access, permitting history, and community factors. It can improve wind and solar forecasting by using weather data, satellite imagery, sensor measurements, and historical generation patterns. It can support predictive maintenance by identifying equipment abnormalities before failures occur. It can optimize battery dispatch, virtual power plants, electric-vehicle charging, and demand-side flexibility. These are not speculative uses. They are the early architecture of the Cognitive Grid. Forecasting is the most immediate example. In a renewable-heavy system, forecast accuracy is not a convenience. It is a reliability resource. Better forecasts reduce balancing costs, improve reserve planning, lower curtailment, and help operators manage uncertainty. As solar, wind, batteries, electric vehicles, heat pumps, and data centers all expand, the value of high-quality forecasting rises. Predictive maintenance is another core element. Traditional maintenance asks what failed. Preventive maintenance asks what might fail on a schedule. Predictive maintenance asks what the system is telling us now. AI makes that third question more powerful by detecting weak signals in large volumes of operational data. But the Cognitive Grid is not just about algorithms. It is about governance. A power system is a high-consequence environment. AI tools must be explainable, auditable, secure, bounded, and integrated into operator workflows. In the energy sector, the right model is not “move fast and break things.” It is “move carefully and learn faster.” That is why advisory AI will come before autonomous AI. The near-term future is not a grid run by a black-box model. It is a grid in which operators are surrounded by better forecasts, better anomaly detection, better scenario analysis, better visualization, and better decision support. The IEA/KEEI report implicitly supports this path. It recognizes AI’s value while warning about cybersecurity, data quality, privacy, hallucination, and the need for governance. That is the right balance. The Cognitive Grid is not the abandonment of human authority. It is the augmentation of institutional intelligence. ## The Shadow Grid: When AI Load Outruns the Public Grid If the Cognitive Grid is the intelligence layer, the Shadow Grid is the parallel infrastructure layer. The Shadow Grid emerges when AI developers cannot rely on the public grid alone to deliver the speed, certainty, or political durability they need. It can include co-located generation, private substations, dedicated renewable procurement, backup generation fleets, storage, special tariffs, curtailment-ready service, green-energy pathways, and cost-containment arrangements designed to prevent local communities from absorbing the full burden of AI growth. In the United States, the Shadow Grid is becoming visible through co-location cases, backup-generation arrangements, large-load tariffs, and emergency reliability discussions. But the IEA/KEEI report suggests that East Asia may develop its own version. The ingredients are present. East Asia has dense urban load centers, constrained land, high energy import dependence, limited cross-border interconnection in several countries, growing data-center demand, and strong national interest in AI and semiconductor competitiveness. Where the public grid cannot expand quickly enough, large AI loads will look for alternative structures. Where clean-energy procurement is difficult, they will seek dedicated pathways. Where grid connection is constrained, they will negotiate around location, timing, flexibility, and cost. Korea’s locational tariff discussion is an early public-grid response to a Shadow Grid pressure. Singapore’s green-energy pathways, though outside the report’s core East Asia focus, point to another model: data-center capacity is increasingly allocated with energy-performance obligations attached. Japan’s constraints may push yet another pathway, where data-center siting, nuclear restarts, offshore wind, storage, and grid reinforcement become part of a single strategic conversation. The Shadow Grid is not necessarily bad. It can bring investment, flexibility, backup capability, and accelerated clean-energy procurement. But it can also create risks. If poorly governed, it can shift costs to other customers, increase emissions through backup generation, fragment planning, or create private reliability systems that do not align with public needs. The policy challenge is not to prevent the Shadow Grid from emerging. It is already emerging. The challenge is to make it transparent, accountable, reliable, low-emissions, and useful to the broader system. ## East Asia’s Five-Regime Fit The IEA/KEEI report can also be read through the five infrastructure regimes from “The New Geography of AI Infrastructure.” The first regime is constrained digital hubs. East Asia has several places where digital density remains valuable but power, land, cooling, permitting, or public acceptance increasingly ration growth. Tokyo, Seoul, Taipei, and Singapore each face versions of this problem. Incumbency still matters, but marginal megawatts now require justification. The second regime is sovereign AI builders. China, Korea, Japan, and Chinese Taipei all treat AI and semiconductors as strategic industrial priorities. Their energy policies can no longer be separated from their compute strategies. Low-emissions electricity, grid reliability, chip manufacturing, data-center siting, and national AI capability are converging. The third regime is spillover zones. As established hubs become constrained, data-center growth will look to lower-cost, lower-congestion, or power-rich areas. Within Korea, non-metropolitan regions may play this role if policy signals are strong enough. Across Asia, secondary markets may gain importance as constrained digital centers curate or ration capacity. The fourth regime is reliability-risk systems. Islanded and quasi-islanded systems face acute operational challenges because they have fewer cross-border balancing options. Large AI loads, variable renewables, and electrification all increase the need for forecasting, flexibility, and operational coordination. Japan’s frequency split and Korea’s lack of interconnection make this regime especially relevant. The fifth regime is the Shadow Grid. East Asia’s combination of large AI ambitions, constrained urban grids, energy-security concerns, and corporate low-carbon procurement needs makes parallel infrastructure likely. The question is whether it develops in a coordinated way or as a fragmented workaround. Seen this way, East Asia is not an exception to the new geography of AI infrastructure. It is one of its clearest demonstrations. ## The Policy Lesson: Compute Must Become Grid-Aware The report’s policy recommendations point toward a larger principle: compute must become grid-aware. Data centers can no longer be treated as ordinary commercial buildings. At small scale, they are real estate. At large scale, they are infrastructure events. They can alter transmission plans, capacity needs, substation investments, local emissions, water use, and cost-allocation debates. They can also provide flexibility, backup capability, demand response, and clean-energy procurement if designed correctly. Grid-aware compute means several things. It means connection processes should distinguish real projects from speculative requests. It means deposits, milestones, and queue rules should reward readiness. It means tariffs should reflect cost causation and location. It means data-center operators should disclose enough operational information for utilities and regulators to plan responsibly. It means large loads should be evaluated not only by annual energy use, but by peak demand, ramp behavior, voltage sensitivity, backup arrangements, flexibility potential, and emergency operating characteristics. It also means AI developers should understand that electricity is not a commodity available everywhere on demand. It is a delivered service embedded in local infrastructure. The megawatt that matters is not the theoretical megawatt. It is the deliverable megawatt, at the right node, at the right time, under the right reliability conditions, with the right political and environmental legitimacy. This is where the Cognitive Grid and Shadow Grid intersect. The Cognitive Grid helps the public system operate with greater intelligence. The Shadow Grid reflects the private effort to secure infrastructure certainty when public systems move too slowly. The future will not be one or the other. It will be a hybrid architecture. Public grids will remain essential, but they will be surrounded by more private generation, storage, backup systems, flexible load arrangements, and dedicated clean-energy pathways. The central question is whether that hybrid architecture strengthens the system or fragments it. ## Conclusion: The New Geography Becomes Operational *Energy and AI in East Asia* is valuable because it moves the AI-energy debate from abstraction to system reality. It shows that AI is not just increasing electricity demand. It is forcing governments, utilities, regulators, and technology companies to rethink the relationship between digital ambition and physical infrastructure. East Asia is a preview of this future because the region compresses the full AI-energy problem into one geography. It manufactures the chips. It hosts fast-growing AI markets. It depends heavily on imported fuels. It has dense cities and constrained grids. It has islanded and quasi-islanded power systems. It has major renewable ambitions and complex siting challenges. It has national industrial strategies that increasingly depend on compute capacity. The lesson is not that AI will overwhelm the power system everywhere. The lesson is that AI will expose the weak seams in power systems wherever planning, permitting, transmission, generation, cooling, water, and regulation cannot move on the same clock as digital demand. That is why the AI race will not be won by models alone. It will be won by places that can convert electricity into intelligence reliably, affordably, cleanly, and quickly enough to matter. The first cloud era rewarded places that could assemble fiber, customers, capital, software talent, and trust. The AI era will reward places that can assemble power, cooling, land, permits, equipment, public legitimacy, and operational intelligence. In that world, the Cognitive Grid becomes the nervous system of the energy transition, and the Shadow Grid becomes the pressure signal that tells us where the public system is not moving fast enough. East Asia now sits at the center of that test. Its challenge is not merely to power AI. Its challenge is to decide what kind of energy system the AI era will require — and whether that system will be planned, visible, and broadly beneficial, or improvised in the shadows. ### The Cognitive Grid Begins as an Interpretive Layer URL: https://www.aixenergy.io/the-cognitive-grid-begins-as-an-interpretive-layer/ Last updated: 2026-07-13T06:06:30.000Z AI, Large Loads, and the First Phase of Grid Intelligence _This post is for paying subscribers only._ ### The Megawatt Is the New Microchip URL: https://www.aixenergy.io/the-megawatt-is-the-new-microchip/ Last updated: 2026-07-10T17:29:50.000Z For most of the digital age, computation seemed weightless. Software lived in the cloud. Intelligence arrived through a screen. The physical systems underneath—power plants, substations, chillers, transformers, water systems, fiber routes, backup generators, and permitting offices—stayed mostly out of view. 0:00 /0:49 1× Artificial intelligence is ending that illusion. The U.S. Energy Information Administration’s May 19, 2026 analysis of the Annual Energy Outlook 2026 gives the shift statistical force. EIA projects that electricity consumed by data-center servers alone will reach 446 billion to 818 billion kilowatthours by 2050\. Servers accounted for an estimated 7 percent of commercial-sector electricity consumption in 2025\. By 2050, EIA projects that data-center server electricity use will rise to 22 percent to 33 percent of all commercial-building electricity use. In the High Electricity Demand case, standalone data centers—represented in EIA’s “other buildings” category—consume 581 billion kilowatthours of server electricity in 2050. The important word is servers. EIA’s headline range does not describe total data-center facility consumption. It describes the electricity used by servers themselves. Cooling, ventilation, power conversion, backup systems, and the rest of the facility follow behind. The server is the visible load. The supporting infrastructure is the shadow it casts. ## The Commercial Building Changes Function That distinction matters because EIA is not merely forecasting that data centers will use more power. It is showing that computation is changing the character of the commercial building stock. Offices, schools, hospitals, stores, warehouses, hotels, and laboratories are now being joined—and in some cases overtaken—by buildings whose central purpose is to convert electricity into computation. The commercial building is no longer just a container for human activity. Increasingly, it is a machine for producing intelligence. That machine has a physical appetite. Its input is electricity. Its byproduct is heat. Its enabling systems include cooling, land, fiber, substations, transformers, switchgear, water, backup power, capital, permits, cybersecurity, and public trust. Its output may be digital, but its constraints are not. This is the deeper lesson of the AI-electricity transition. The AI race will not be won only by the firms with the best models, the largest chip inventories, or the most sophisticated software teams. It will be won by those that can turn a megawatt into compute reliably, affordably, visibly, and on time. ## From End Use to Planning Regime EIA’s analysis is powerful because it marks a shift in how the energy system now sees computation. For AEO2026, EIA updated its Commercial Demand Model to report data-center server electricity use separately from broader commercial computing. That may sound like a technical adjustment, but it is more than that. It is a signal that a load once buried inside ordinary commercial activity has become large enough to stand on its own. Energy transitions often become visible this way. At first, the new thing is too small to matter. Then it becomes an adjustment. Then it becomes a category. Eventually, it becomes a planning regime. Data-center servers are now crossing that line. The clearest evidence is commercial electricity intensity. EIA projects that commercial-sector electricity use per square foot will exceed the 2003 historical high of 14.9 kWh per square foot for the first time in 2031–2032\. For decades, newer buildings often meant better lighting, controls, motors, chillers, and ventilation. Efficiency helped hold down electricity intensity. Data centers change that story. The most advanced and economically strategic commercial buildings may now be among the most electricity intensive. This is not a failure of efficiency. It is a change in function. These buildings are not merely lighting rooms and conditioning air for people. They are running machines continuously. ## Flat Does Not Mean Simple EIA also assumes that data-center server load is essentially flat across the day. For a national long-term model, that is reasonable. Many data centers are designed to operate at high utilization around the clock. But flat does not mean simple. A 300 MW load that runs steadily can still require major transmission upgrades, firm generation capacity, voltage support, contingency planning, backup coordination, and a careful decision about who pays. A flat load can still be a system-shaping load. That is where the national forecast meets the project-level reality. EIA shows the size of the wave. A separate analysis of hyperscale data-center development shows why that wave will not arrive smoothly. Large projects increasingly face a full “constraint stack”: power availability, interconnection timing, water supply, cooling design, permitting, local governance, and public acceptance. These constraints no longer resolve one at a time. They interact. A project can secure land and capital, design a high-efficiency facility, and still stall because interconnection studies stretch into years or because local officials demand proof of long-term water supply. That is the missing bridge between national energy modeling and real infrastructure execution. A billion kilowatthours in an outlook becomes, on the ground, a substation, a transformer order, a transmission study, a cooling design, a water permit, a zoning hearing, a tariff filing, a fuel contract, and a public meeting. ## The New Metric Is Time-to-Compute The old development sequence was linear: find land, secure power, design the building, obtain permits, build, energize, operate. AI infrastructure is breaking that sequence. Power, water, cooling, permits, and community acceptance now move together. If one slips, the whole project slips. This is why the decisive metric is not announced megawatts. It is time-to-compute. Time-to-compute is the period required to turn a proposed data-center project into reliable operating computational capacity. It includes land control, utility service, interconnection, transmission deliverability, transformer availability, cooling design, water certainty, backup configuration, permitting, cybersecurity readiness, public acceptance, and energization. A site with a two-year construction schedule but an eight-year power solution is not a two-year project. A site with electricity but no credible cooling path is not ready. A site with land and capital but no public legitimacy is not bankable infrastructure. This changes the geography of AI competitiveness. The winning regions will not necessarily be those with the cheapest nominal electricity or the largest parcels of land. They will be the places where the full infrastructure stack can align. Power without water may fail. Water without transmission may fail. Land without permitting may fail. Fiber without reliability may fail. Capital without trust may fail. ## The Bottleneck Is the Stack In the AI era, the bottleneck is rarely one thing for long. The bottleneck is the stack. Cooling makes this unavoidable. EIA assumes that space-cooling requirements in data-center floorspace can be as much as 2.9 times as energy intensive as non-data-center floorspace. In the High Electricity Demand case, EIA projects that space-cooling electricity consumption is 84 billion kilowatthours higher than in the Counterfactual Baseline case in 2050 to support more intensive data-center operations. Cooling is often treated as a support system. It is not. In AI infrastructure, cooling is part of the compute engine. Every watt consumed by a server becomes heat. That heat must be removed, absorbed, rejected, reused, or tolerated. As rack densities rise, cooling becomes a determinant of siting, electricity demand, water exposure, permitting risk, and resilience. The tradeoff is brutal because it is physical. Evaporative cooling can reduce electricity demand but consume significant water. Dry cooling can reduce water use but increase electricity demand and perform less efficiently during extreme heat. Hybrid systems provide flexibility, but they do not eliminate the tradeoff. The constraint-stack analysis makes the deeper point: the water footprint of computation depends not only on the facility, but also on the electricity system that supplies it. A developer can reduce on-site water use by shifting to dry cooling, but if that increases electricity demand in a region supplied by water-consuming thermal generation, some of the water burden may simply move upstream. ## The Constraint Does Not Disappear. It Moves. The system does not always eliminate the constraint. It relocates it. That same boundary problem appears in power procurement. A data center may claim clean energy on an annual basis, but the grid must still deliver reliable electricity every hour. A campus may add on-site generation, but that generation may depend on public gas networks, emissions rules, interconnection agreements, and emergency coordination. A facility may sit behind the meter, but it is not outside the system. This is why the shadow grid is emerging. When public-grid interconnection takes too long, when transmission is constrained, when generation development lags data-center construction, and when local approvals become uncertain, developers look for parallel pathways. They assemble dedicated renewables, batteries, private substations, backup fleets, on-site generation, special tariffs, fuel contracts, and hybrid energy campuses. This is not necessarily a rejection of the public grid. It is a rational response to compound constraint. But it creates a governance problem. Private power is not entirely private if the public system must stand ready when it fails. Backup generation is not merely backup if it becomes part of emergency response. Behind-the-meter assets are not irrelevant to public reliability simply because they sit behind a customer meter. ## The Shadow Grid Must Become Visible The shadow grid must be made visible before it becomes too large to govern. EIA’s efficiency assumptions add another layer of uncertainty. In the Counterfactual Baseline case, EIA assumes that after 2040 servers become increasingly efficient, reducing average annual operational power draw by 10 percent every three years beyond historical efficiency trends. Even then, continued growth in server installations drives overall consumption higher. In the High Electricity Demand case, EIA makes no such additional efficiency assumption and assumes AI servers become a larger share of installed server stock. That difference is the hinge of the AI electricity debate. Efficiency may improve, but demand may grow faster. Energy history has seen this before. More efficient steam engines did not end coal use; they expanded the uses of steam. More efficient lighting did not end electricity demand for illumination; it made light ubiquitous. More efficient computation may not reduce electricity demand if cheaper intelligence creates more reasons to compute. ## Demand Visibility Becomes Infrastructure Intelligence This is why utilities and regulators cannot treat AI load growth as a conventional forecast problem. Historical commercial-sector trends will not reveal where the next cluster of GPU-intensive campuses will land. Existing interconnection queues do not distinguish serious projects from strategic placeholders. Annual energy projections do not capture the difference between a site that can be energized in thirty months and one trapped behind years of grid dependency. The next planning model must combine utility data with external signals: land transactions, zoning activity, water filings, tax incentives, transformer orders, generator procurement, fiber expansion, chip supply, construction permits, developer behavior, and regional policy shifts. This is not speculation. It is the new frontier of demand visibility. ## Risk Must Follow Causation The cost-allocation issue will become unavoidable. If servers rise from 7 percent of commercial-sector electricity consumption to as much as one-third by 2050, the commercial class itself becomes less coherent. A small business, a hospital, a grocery store, a university server room, a colocation facility, and a hyperscale AI campus do not impose the same grid problem. Some are ordinary customers with digital equipment. Others are regional infrastructure events. Data centers can be valuable customers. They can support new generation, anchor transmission investment, strengthen tax bases, accelerate advanced cooling, and help pay for infrastructure that improves broader reliability. They can also create stranded-cost risk, speculative queue congestion, water conflict, emissions backlash, and public resistance if their costs are socialized while their benefits are privatized or exported. The principle should be simple: growth is welcome, but risk must follow causation. Developers that require extraordinary infrastructure should provide extraordinary transparency, milestone discipline, credit support, and contractual commitment. Utilities should not be forced to choose between economic development and customer protection using tools designed for ordinary load additions. Regulators should distinguish real projects from speculative claims, firm demand from flexible demand, and network benefits from project-specific costs. ## Governance Is Part of the Infrastructure The broader implication is that AI infrastructure is forcing electricity planning to become more institutional, not merely more technical. Public trust now determines whether physical systems can be built. A project may have engineering merit and still fail if a community believes it will raise rates, drain water, rely on polluting backup power, or provide too little local benefit. A regulator may support innovation and still reject a structure that shifts too much risk to existing customers. Governance is not paperwork after the infrastructure is designed. Governance is part of the infrastructure. That is also why the grid will become cognitive before it becomes autonomous. The near-term value of AI in the power sector is not handing critical infrastructure to opaque machine control. It is better visibility, faster interpretation, probabilistic forecasting, anomaly detection, scenario analysis, operator decision support, and executive intelligence. The cognitive grid is not a robot dispatcher. It is an intelligence layer that helps people and institutions see what is changing before the system runs out of time. ## The Race to Turn Megawatts Into Compute EIA has shown that data-center servers are becoming large enough to reshape the commercial building stock. The constraint-stack analysis shows why that server demand will not become operating compute automatically. The deeper lesson is that the megawatt has become the new microchip. It is not sufficient by itself, but without it—and without the physical systems that make it usable—the rest of the AI stack cannot function. The next decade of AI leadership will not belong simply to those who announce the most models, buy the most chips, or reserve the most land. It will belong to those who can assemble the full infrastructure stack into working systems: electricity, cooling, water, land, fiber, capital, equipment, permitting, reliability, cybersecurity, and trust. That is where the digital future becomes physical. That is where the AI race will be won or lost. ### The New Geography of AI Infrastructure URL: https://www.aixenergy.io/the-new-geography-of-ai-infrastructure/ Last updated: 2026-07-13T06:06:44.000Z AI is making computation physical again: global AI capacity will follow not just chips and fiber, but power, water, cooling, permits, grid reliability, and the emerging Shadow Grid that turns energy infrastructure into the new map of AI advantage. _This post is for paying subscribers only._ ### The Power Sector’s AI Decade Part III URL: https://www.aixenergy.io/the-power-sectors-ai-decade_part_iii/ Last updated: 2026-07-29T19:41:40.000Z *This is the third of a five-part series, hosted by* [*Brandon Owens*](https://www.aixenergy.io/author/brandon/) *at* [*AIxEnergy*](https://www.aixenergy.io/)*.* [*Part 1*](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fi/) *introduced three distinctions (vendor capability versus utility adoption, reliability versus safety, task versus role) and a twelve-row sample of the US electricity workforce.* [*Part 2*](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fii/) *developed the four-stage mechanism by which AI capability translates into workforce change in regulated industries and traced it through aviation, freight rail, and maritime. This part builds the full role-by-role inventory across the major positions in the US electricity industry and the institutional infrastructure surrounding it.* For most of the twentieth century, a meter reader at a US electric utility walked or drove a route (typically a few hundred residential and commercial accounts in a day), read the glass dial on each meter, recorded the figure, and turned the readings in to the billing office at the end of the shift. In May 2006, the Bureau of Labor Statistics counted 10,510 of them in the US electric power industry.\[1\] 0:00 /0:43 1× On December 19, 2007, President Bush signed the Energy Independence and Security Act, whose Title XIII established the federal Smart Grid policy framework and authorized federal cost-share for advanced metering investments. Pacific Gas and Electric had begun deploying SmartMeter devices the year before, in a program the California Public Utilities Commission expanded substantially in 2009; the rollout would reach roughly nine million meters by 2012\. Other utilities followed on their own timelines, encouraged by Smart Grid Investment Grants from the American Recovery and Reinvestment Act. By 2011, 23 percent of US electricity customers had advanced meters; by 2016, 47 percent; by 2022, 72 percent.\[2\] By May 2025, the BLS count of meter readers in the US electric power industry was 3,320\. The role had contracted by nearly 70 percent in nineteen years. It had not been outsourced. It had not been deskilled. It had not been negotiated away. It had been hollowed out from within by a piece of technology that did the human’s task without a human in the loop. The meter reader sits at the closed end of an arc that most of the rest of the US electric power workforce is still partway along. Advanced metering infrastructure is not artificial intelligence. The cellular smart meter and its database are a generation older than the deep learning systems entering utility operations today, and the technology that displaced the meter reader is closer in spirit to the dial telephone than to ChatGPT. The pattern around the displacement, though, is what matters for the analysis below. A task-specific technology, deployed against a role whose institutional defenses were thin, produced a steady, asymmetric, near-total contraction in role headcount over fifteen-to-twenty years without a single regulatory crisis. That pattern is the predictor. [Part 1 ](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fi/)of this series introduced three distinctions that organize how AI capability reaches the utility workforce: vendor AI capability versus utility adoption, reliability functions versus safety functions, and task content versus role count. Part 2 developed a four-stage mechanism (capability maturation, task-content erosion, institutional-defense attenuation, and role-level headcount change) and traced it through aviation, freight rail, and maritime, three other safety-critical regulated industries that have moved through some version of the same arc on different clocks. The meter reader is the case where all four stages have run to completion: the technology matured, the task content eroded, the institutional defenses attenuated (CPUC opt-out programs notwithstanding, no contract or regulation prevented the substitution), and the headcount fell. Most of the rest of the US electric power workforce sits earlier on that arc, at stage one, two, or three. Where each role sits, how much of its task content has shifted, how much of its institutional armor still holds, and how much of its headcount has moved or will move varies systematically across the industry. The inventory below maps that variation, role by role. The employment figures use Bureau of Labor Statistics Occupational Employment and Wage Statistics data for industry code 221100, May 2025 vintage, unless a row notes otherwise.[\[3\]](#%5Fftn3) The dominant-AI-system column uses Brandon Owens’s six-cluster taxonomy from [*Artificial Intelligence Has Entered the Grid Quietly*](https://www.aixenergy.io/artificial-intelligence-has-entered-the-grid-quietly-what-u-s-utilities-are-actually-doing-with-ai-why-most-deployments-remain-constrained-and-what-that-reveals-about-the-future-of-crit/) (January 2026).[\[4\]](#%5Fftn4) Where role-specific employment cannot be separated within an occupational code, the row says so. Several rows describe a task-content shift that follows the progression [Shailesh Jain](https://www.aixenergy.io/author/shailesh/) develops in [*From Control to Cognition*](https://www.aixenergy.io/from-control-to-cognition-training-the-grid-for-the-age-of-ai/) (October 2025): from direct judgment by the operator, to augmented judgment with AI in support, to supervisory judgment over learning systems with humans in the governance loop. Jain’s characterization of where this leads is precise: *operators do not vanish; they ascend.* Protection engineers validate adaptive settings rather than recalculating them. System planners become curators of models that evolve continuously rather than authors of studies frozen in time. Where the progression maps cleanly onto a row in the matrix below, the row cites the stage; elsewhere the row describes the shift in plain terms. Each row in the matrices below has eight columns. The first names the role, with its BLS Standard Occupational Classification (SOC) code where one applies. The second sketches representative tasks. The third gives approximate employment, drawn from the BLS Occupational Employment and Wage Statistics survey for NAICS 221100 in May 2025 unless a row specifies otherwise. The fourth tags the role for the institutional obligation it carries: R for reliability functions (keeping the grid stable and the lights on), S for safety functions (preventing harm to people or equipment), or O for operations and back-office functions. Roles carrying more than one obligation have combined tags. The fifth names the dominant AI system reaching the role today, identified by which of Brandon Owens’s six clusters it belongs to and, where deployment evidence allows, the specific technology underlying it. The sixth describes the substantive shift AI is producing in the work the role performs day to day. The seventh marks the direction of role-level headcount with an arrow: a single up arrow for growing, a horizontal arrow for stable, a single down arrow for contracting, two down arrows for contracting sharply. The eighth gives the time horizon over which the change in the seventh column is expected to play out: near covers the next two to five years, medium five to fifteen, and long fifteen years and beyond. ## Generation Operations A plant operator at a modern US combined-cycle gas turbine in 2026 spends much of a shift watching a distributed control system display that has changed substantively in the past decade. At plants where utilities have deployed condition-monitoring stacks, vibration sensors on the compressor and turbine bearings, thermal imaging on the generator stator, and pressure transducers throughout the steam cycle feed a continuous stream of data into long short-term memory models trained on the unit’s own historical sensor record. The models flag anomalies before they reach alarm thresholds. The operator’s job, at such a plant, has shifted from scanning the instruments for an anomaly toward validating the AI’s flag, deciding whether to dispatch a technician, and judging whether to derate or trip the unit. Most US fossil generation has not yet reached this stack. The direction of travel, at the plants that have, is clear. That is the direction of the task-level shift in conventional generation. The role-level story runs differently. Power plant operator headcount inside the US electric utility industry has been contracting for two decades, and the dominant driver is fleet consolidation rather than AI: gas capacity additions have slowed, coal retirements have accelerated, and per-megawatt staffing in modern combined-cycle plants is lower than in the units they replace. AI lets each operator handle more equipment, which makes the smaller staffing pattern workable. It does not independently drive layoffs in the near term. **Nuclear** is a structurally different case. Reactor operators carry licenses from the Nuclear Regulatory Commission (NRC) that codify their task competencies against specific operating procedures, and those licenses are not candidates for near-term relaxation. AI provides early-warning signals on core and coolant parameters, but it cannot reduce the human-in-the-loop requirement under current regulations. Task content shifts incrementally, toward validation of digital-display outputs rather than direct reading of analog instruments. The reactor operator workforce holds steady. The rate-limiting step for change here is not technology availability but the revision of NRC operator licensing, and licensing revision at the NRC has historically taken decades to work through. In the Jain framework introduced above, the role is in early transition from direct judgment by the licensed operator toward augmented judgment with AI in support, and the change should play out over fifteen years and more rather than five. **Renewables** look different again. Wind turbine technicians and solar photovoltaic technicians work in industries where total employment is growing because buildout is growing, even as the task content within each role compresses through AI-driven inspection planning, computer-vision blade defect detection from drone imagery, and AI-flagged inverter fault dispatch. A wind tech arrives at a turbine knowing which unit needs attention, what is likely wrong, which parts to bring, and which sites to hit in what sequence, work that previously required scheduled rotation and field-level diagnosis. The physical climb-and-repair work remains dexterity-bound, and the dexterity requirement is the moat. Most of the workforce adjacent to generation is not in the rows below. The construction crews who build new combined-cycle plants and renewable projects, the contracted maintenance teams who perform major overhauls, and the technical specialists who support commissioning are coded in NAICS 23 (Construction) and NAICS 541 (Professional and Technical Services), outside the utility industry code. The 418,900 workers BLS counts in NAICS 221100 in May 2025 is roughly twenty percent of the workforce the US Energy and Employment Report counts in electric power generation across the full employer universe.[\[5\]](#%5Fftn5) The matrix below counts the twenty percent. ### How AI Changes Work Across the Power-Sector Labor Stack AI does not affect every energy occupation in the same way. The near-term shift is concentrated in monitoring, diagnostics, forecasting, and documentation, while licensed, physical, and safety-critical work remains more resistant to full automation. | Role | Current Tasks | Employment | R/S/O | Dominant AI System | Task-Content Shift | Role Direction | Horizon | | ------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------ | ----- | ---------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------- | ----------------------- | | Power plant operator 51-8013 | Monitor DCS; adjust fuel and cooling parameters; respond to equipment alarms; coordinate with maintenance. | 19,690 | R/S | Asset monitoring; vibration sensors; thermal imaging; LSTM early-warning. | Operator no longer scans the instrument wall for anomalies; responds to AI-surfaced alerts and decides whether to dispatch a technician, derate, or trip the unit. | ↓ | Medium | | Nuclear power reactor operator 51-8011 | Monitor reactor core and coolant; execute NRC-licensed procedures; conduct surveillance testing. | 4,030 | S | Asset monitoring; physics-informed ML for reactor diagnostics; NRC-supervised digital instrumentation and control. | Licensed procedure execution persists; AI provides early-warning signals but cannot reduce the human-in-the-loop requirement. | ↔ | Long | | Nuclear technician 19-4051 | Radiation surveys in containment and auxiliary buildings; calibrate radiation monitors; fuel handling support. | 3,600 | S | Asset monitoring; computer vision for remote hot-cell inspection; ML anomaly detection in radiation telemetry. | Routine area-survey work partially automates via remote monitoring; technician shifts toward exception-driven investigation. | ↔ | Long | | Wind turbine service technician 49-9081 | Gearbox, generator, and yaw maintenance; SCADA fault troubleshooting; blade and tower inspections. | 4,960 | R | Asset monitoring; computer-vision blade defect detection from drone imagery; vibration-analysis ML; SCADA-integrated predictive maintenance. | Scheduled climbs give way to condition-based intervention; AI pre-identifies which units need work, the likely fault, and required parts. | ↑ | Near task / Medium role | | Solar PV installer / technician 47-2231 | Module and inverter mounting; commissioning tests; O&M diagnostics on string, module, and inverter performance. | 4,000 NAICS 221100; broader solar workforce of \~280,000 sits in NAICS 23 per USEER. | R | Asset monitoring; forecasting and planning; computer vision for module defect detection; aerial thermography; AI-driven inverter optimization. | O&M moves from periodic site visits to AI-flagged condition-based dispatch; installation labor remains dexterity-bound. | ↑ | Near task / Medium role | | Plant shift supervisor / operations lead 51-1011 subset | Direct shift crew operations; turnover and tailboard briefings; approve operating procedure deviations; coordinate outage planning. | \~2,000–3,000 Generation subset of 5,740 first-line supervisors in NAICS 221100\. | R/S | Asset monitoring; grid optimization; AI-aggregated plant-status dashboards; LLM-assisted shift-turnover documentation. | Less time compiling shift logs and outage windows; more time on crew assignments, deviations from procedure, and outage coordination—work that carries personal liability and union contract implications. | ↓ | Medium | R/S/O indicates the role’s dominant exposure: routine, safety-critical, or operationally constrained work. Arrows indicate directional pressure on the role, not a deterministic employment forecast. ## Transmission and Distribution Field Workforce In June 2023, [Southern Company received one of the first conditions-based FAA waivers](https://dronelife.com/2023/06/26/southern-company-gets-bvlos-waiver-for-fully-remote-drone-operations-with-skydio-dock/?ref=aixenergy.io) authorizing remote-operated, beyond-visual-line-of-sight (BVLOS) autonomous drone operations at critical infrastructure sites across its system. The opening of Part 1 of this series described that deployment. Two and a half years later, the [FAA published a notice of proposed rulemaking under Part 108 in August 2025](https://www.federalregister.gov/documents/2025/08/07/2025-14992/normalizing-unmanned-aircraft-systems-beyond-visual-line-of-sight-operations?ref=aixenergy.io) to normalize BVLOS operations across critical infrastructure inspection without per-mission waivers. Final rulemaking is pending. When the rule lands, the inspection task collapses to its essentials: the work that previously required a pilot-observer team in the field can be done by a technician in a control room reviewing AI-classified findings from autonomous flights. That is the largest near-term task-content shift in the cluster, and it does not produce a corresponding role-level contraction in line installer headcount. Physical conductor work is dexterity-constrained (energized line work, hardware replacement, splicing, switching), and the relevant question for the next decade is whether dexterous robotics can do energized physical tasks. Research on legged and crawling robots for energized work is active but not yet near commercial deployment. The expected outcome over the next decade is task compression around the physical work rather than substitution of the physical work itself. **Vegetation management** sits at the intersection of this cluster and the wildfire-mitigation Owens cluster. Most utility arborist work is contracted, coded in NAICS 561730 (Landscaping Services) rather than NAICS 221100, and the dominant deployment pattern at large investor-owned utilities is risk-prioritized intervention informed by satellite imagery, weather modeling, and ignition-risk scoring rather than cycle-based trimming on a fixed schedule. The role headcount grows because wildfire-driven program scope is expanding at large Western utilities. The task content within each role shifts toward following AI-prioritized work orders rather than walking a route. **Meter reading** is the role whose contraction the opening of this part traced from 10,510 in 2006 to 3,320 in 2025: the easy case where all four stages of the mechanism have run, the institutional defenses were thin, and the substituting technology was task-specific enough to displace the work almost entirely. **AMI installation** is the workforce that built out the smart meter infrastructure across the past two decades. It sat almost entirely outside the utility industry code. The installation crews were specialty contractors (Wellington Energy, Grid One Solutions, MasTec utility services among the largest), coded in Construction or Administrative Support, not in NAICS 221100\. The row below names the contractor pattern rather than estimating an in-utility count that does not exist. ### How AI Reshapes Transmission, Distribution, and Field Operations Work In grid field operations, AI is less likely to replace skilled physical work than to compress inspection, routing, diagnosis, documentation, and prioritization. The deepest pressure falls on work built around route-based observation or one-time infrastructure deployment cycles. | Role | Current Tasks | Employment | R/S/O | Dominant AI System | Task-Content Shift | Role Direction | Horizon | | --------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------ | ----- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------- | ----------------------------- | | Electrical power-line installer and repairer 49-9051 | Patrol T&D circuits for damage; string and terminate conductors; replace insulators, deadends, and cross-arms; perform energized line work under qualified-worker authorization. | 62,470 | S | Outage prediction and reliability support; wildfire mitigation; BVLOS drones for damage assessment; LiDAR conductor-clearance analysis; AI-generated work-order prioritization. | Physical conductor work remains dexterity-bound. Patrol, damage assessment, and work-order triage compress under drone imagery and AI classification; crews arrive with damage maps and priority rankings already drafted. | ↔ | Near task / Medium role | | Substation / relay repairer 49-2095 | Test and calibrate protective relays; perform scheduled maintenance on transformers, breakers, and switchgear; troubleshoot control-circuit faults; maintain NERC CIP compliance documentation. | 12,920 | R/S | Asset monitoring; dissolved-gas analysis sensors; partial-discharge sensors; AI-driven condition-based maintenance scheduling. | Fixed-interval preventive maintenance gives way to condition-based maintenance. The technician validates model-recommended intervention timing. | ↔ | Medium | | Transmission line patrol inspector 49-9051 subset | Conduct aerial and ground patrols of transmission circuits; identify conductor damage, hardware failures, and encroachments; document findings for work-order generation. | Subset of 62,470 line workers Many patrol functions are performed by contracted aviation services outside NAICS 221100. | R/S | Outage prediction; wildfire mitigation; BVLOS autonomous drones; LiDAR conductor-sag analysis; AI image classification for damaged insulators and right-of-way intrusions. | Pre-flight planning and post-flight inspection drafting compress under AI image classification. Pilot-observer functions give way to remote supervision of AI-flagged findings. | ↓ | Near task / Medium role | | Underground cable splicer 49-9051 subset / 49-2093 | Locate underground cable faults; splice and terminate medium- and high-voltage cable; perform partial-discharge testing; maintain installation and repair records. | Not separable in OEWS Subset of 62,470 workers in 49-9051 and 49-2093; 40 in NAICS 221100. | S | Outage prediction; asset monitoring; partial-discharge sensors with ML anomaly classification; GIS-fused fault-location algorithms. | Locating and diagnosing cable faults compresses. Physical jointing under live or de-energized conditions remains dexterity-bound and credential-bound. | ↔ | Medium-long | | Utility arborist / vegetation management worker 37-3013; typically contract | Trim and remove vegetation in T&D rights-of-way; assess tree-related fire and reliability risk; document clearance compliance. | Not separable in NAICS 221100 Most workers are coded in NAICS 561730, Landscaping Services, under contract to utilities. | R/S | Wildfire mitigation and vegetation management; LiDAR encroachment modeling; satellite and aerial imagery for risk scoring; AI-driven prioritized trim scheduling. | Cycle-based trimming yields to risk-prioritized intervention informed by satellite imagery, weather modeling, and ignition-risk scoring. | ↑ | Near planning / Medium role | | Meter reader 43-5041 | Read residential and commercial meters for accounts outside AMI coverage; investigate suspected tampering and non-AMI billing exceptions; field-verify customer-reported readings. | 3,320 Down from 10,510 in 2006. | O | Asset monitoring. AMI infrastructure itself is the displacement technology, rather than an AI overlay. | Route-based reading has been substantially eliminated by remote reading. Residual functions include non-AMI accounts, off-cycle reads, and tamper investigations. | ↓↓ | Near / substantially complete | | AMI installation technician Contracted; outside NAICS 221100 | Install AMI meters at residential and commercial accounts during rollout; verify meter-form compatibility and commissioning; coordinate with utility back office on activation. | Not separable in OEWS AMI installation is typically performed by specialty contractors coded in NAICS 23 or NAICS 56. | R | Asset monitoring; large-load integration through AMI as enabling infrastructure; AI-routed deployment scheduling; computer-vision meter-form verification; automated commissioning workflows. | This is a rollout-buildout workforce rather than a steady-state utility role. Volume tracks utility AMI deployment cycles, then declines as the installation base saturates. | ↔ → ↓↓ | Near-medium | R/S/O indicates the role’s dominant exposure: routine, safety-critical, or operationally constrained work. Arrows indicate directional pressure on the role, not a deterministic employment forecast. In this table, the strongest labor-market pressure appears where work is primarily observational, route-based, or tied to one-time deployment cycles. ## Control Room and Dispatch In July 2025, [the California Independent System Operator and OATI launched a generative-AI pilot inside CAISO’s outage management system](https://www.publicpower.org/periodical/article/oati-launches-generative-ai-platform-pilot-power-grid-operations-with-caiso?ref=aixenergy.io). CAISO’s chief information and technology officer framed the deployment in terms of “improving situational awareness and freeing up time for other important tasks” for operators. In the pilot, the operator remains the decision authority over every action the system recommends. What changes, where the system is in use, is what the operator does between recommendations: less time compiling situational awareness from raw data, more time interpreting AI-generated summaries and validating recommended actions. Similar operator-assist pilots are active at PJM and SPP. The pilots are pilots, not pervasive deployments, and the practical experience of most US dispatchers in 2026 still looks much closer to direct judgment than to supervisory judgment. The control room cluster is where Jain’s *operators do not vanish; they ascend* characterization is most directly in tension with the role-level arrows below. Both halves of the tension are real. At the leading edge, the task content of the dispatcher’s day has begun to shift from direct judgment toward supervision of automated outputs, in the way Jain describes for the near term, with the operator retaining final authority over every action the system recommends. Most dispatchers have not yet seen this shift. The matrix’s downward role-level arrow does not contradict that. It locates the same trajectory on a longer clock. The supervisory operator is a smaller workforce than the directly-judging operator was, even when each individual operator is doing more cognitively demanding work than before. Stage two of the four-stage mechanism produces both effects at once. The certification framework (the [NERC System Operator Certification Program](https://www.nerc.com/pa/Train/SysOpCert?ref=aixenergy.io)) that credentialed the directly-judging operator does not yet describe the supervisory operator’s actual work, and that gap is the institutional pressure that eventually carries the cluster into stage three. A second pattern in this cluster runs in the opposite direction. Distribution system operators working with advanced distribution management systems (ADMS) and distributed energy resource management systems (DERMS) are a role being created rather than displaced. Distribution-level operations have not historically required real-time dispatch judgment at the scale that transmission has; the proliferation of distributed solar, behind-the-meter storage, and electric vehicle charging is creating the workload that the role exists to handle. AI handles the combinatorial complexity that humans cannot manage at scale. The role grows because the system it operates is growing. ### How AI Reshapes Dispatch, Market, ADMS, and SCADA Control-Room Work In utility control rooms, AI changes the tempo of work before it changes the accountability structure. Real-time operators, reliability coordinators, traders, ADMS operators, and SCADA technicians increasingly supervise, validate, and challenge machine-generated recommendations, while certified human judgment remains the institutional backstop. | Role | Current Tasks | Employment | R/S/O | Dominant AI System | Task-Content Shift | Role Direction | Horizon | | ---------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------- | ----- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------- | ----------- | | Power dispatcher / balancing authority operator 51-8012 | Monitor real-time frequency and tie-line flows; issue switching orders; adjust generation dispatch to maintain balance; respond to contingency events with the reliability coordinator. | 5,000 | R/S | Forecasting and planning; grid optimization; generative-AI operator assist; AI-enhanced EMS with real-time contingency analysis; reinforcement-learning dispatch optimization. | Direct real-time dispatch judgment moves toward supervision of AI-recommended actions. Operators build situational awareness from AI-generated summaries as leading deployments move into assisted-operation modes. | ↓ | Medium | | Reliability coordinator NERC-certified institutional role | Maintain wide-area situational awareness across multiple balancing authorities; issue reliability directives; coordinate emergency response; oversee transmission-loading relief procedures. | \~500–1,000 nationally Not in NAICS 221100; concentrated at ISOs/RTOs and a small number of large multi-balancing-authority utilities. | R/S | Outage prediction; grid optimization; AI-enhanced contingency screening; real-time security-constrained dispatch with ML overlays; generative situational summarization for cross-region coordination. | Event-by-event judgment moves toward supervisory oversight of AI-recommended actions across larger footprints. NERC RC certification remains the institutional defense. | ↔ | Medium-long | | Energy market trader / scheduling coordinator 13-2099 / 13-2051 subset | Submit day-ahead and real-time energy bids; manage congestion through financial transmission rights; monitor ISO market dashboards; reconcile settlements. | Not separable in OEWS A large share sits at power-marketing affiliates and trading houses outside NAICS 221100\. | R | Forecasting and planning; ML-driven price and congestion forecasting; LLM-assisted regulatory filing review; algorithmic bid optimization. | First-pass forecasting and bid construction increasingly become model-generated. The trader’s value shifts toward challenging model assumptions on edge cases and managing counterparty relationships. | ↓↓ | Near-medium | | Distribution system operator / ADMS operator Emerging role | Operate ADMS for feeder switching, fault location, isolation, and restoration; supervise DERMS-aggregated dispatch; respond to grid-edge events with automated topology reconfiguration. | \~500–1,500 Subset of 5,000 power distributors and dispatchers in NAICS 221100; ADMS-specific count is still emerging. | R | Grid optimization and DER coordination; outage prediction; AI-augmented ADMS with predictive feeder modeling; DERMS for aggregated DER dispatch; FLISR automation; real-time topology estimation. | Dispatch responsibility expands to encompass DER coordination and dynamic feeder reconfiguration. AI handles combinatorial complexity that humans cannot manage at scale. | ↑ | Near-medium | | SCADA analyst / EMS technician 17-3023 subset | Maintain SCADA and EMS point databases; troubleshoot communication outages between substations and control centers; support alarm-response engineering and OT cybersecurity. | \~2,000–3,000 Subset of 6,020 electrical and electronic engineering technologists in NAICS 221100\. | R/O | Asset monitoring; grid optimization; AI-enhanced anomaly detection in SCADA telemetry; automated alarm filtering and prioritization; ML-based point-quality screening. | AI triages and prioritizes alarms before the technician sees them. The technician investigates flagged anomalies and manages the rapidly growing universe of sensors feeding SCADA. | ↔ | Medium | R/S/O indicates the role’s dominant exposure: routine, safety-critical, or operationally constrained work. Arrows indicate directional pressure on the role, not a deterministic employment forecast. In this cluster, AI compresses forecasting, contingency screening, alarm triage, and bid construction, while certified operators and technical specialists remain accountable for reliability-critical decisions. ## Engineering and Planning OEWS does not distinguish protection engineers from planning engineers from interconnection engineers from general electrical engineers. All sit under Standard Occupational Classification codes 17-2071 (Electrical Engineers) and 17-2072 (Electronics Engineers), and the 20,330 electrical engineers BLS counts in NAICS 221100 in May 2025 is a single bucket. The cluster below subdivides the bucket using utility staffing pattern conventions and trade association estimates, with each row identified as an approximation. **Protection engineering** is a safety function rather than a reliability function. Protective relays coordinate fault interruption across the transmission and distribution system, and the settings that govern that coordination exist to prevent equipment damage and personnel hazard during electrical disturbances. The institutional architecture around the role reflects that: NERC PRC reliability standards specify minimum performance requirements, professional engineering liability follows the engineer’s signature on the coordination study, and the standards themselves take years to revise. AI applied to relay setting optimization is a research frontier (Pacific Northwest National Laboratory and other national laboratories and academic groups have published on machine-learning approaches to protective relays), but utility deployment remains limited. Where these tools reach production use, the shift in the work will follow the Jain progression directly: *protection engineers validate adaptive settings rather than recalculating them*. Headcount in the role holds, on a horizon governed by the same institutional logic that holds nuclear operators in place. The engineer signs the settings, and the signature carries professional engineering liability that no AI vendor will assume. **Planning engineering** is a reliability function. Planners produce the studies that demonstrate the grid will remain stable under contingencies, but they do not sign off on operating decisions in real time, and their institutional protection is correspondingly thinner than the protection engineer’s. At utilities that have adopted AI-assisted scenario generation, ML-driven load and resource forecast automation, and LLM-assisted regulatory filing drafting layered on top of conventional tools like PSS/E and PowerWorld, transmission planning studies that previously took weeks of manual scenario setup are compressing to days or less. Most utility planning departments have not adopted these tools at scale. Where adoption proceeds, the planner becomes, in Jain’s framing, *a curator of models that evolve continuously rather than an author of studies frozen in time*. Regulatory sign-off on the resulting plans, and the liability-bearing engineering review that produces it, remain human work. The role itself contracts over the next five to fifteen years as per-study labor intensity falls and study production capacity becomes less of a binding constraint on the planning function. **Power electronics and energy storage engineering** are growing roles. Inverter-based resource deployment, grid-forming inverter development, and utility-scale battery integration require engineering capacity that did not exist at scale a decade ago. AI runs control scheme simulations in minutes instead of days, letting engineers explore design alternatives that were previously impractical. The engineer’s value-add shifts from running the simulations to verifying that AI-generated control schemes meet grid-code requirements and translating the technical findings into market and regulatory documents. Role-level headcount grows; per-MW labor intensity falls. ### How AI Reshapes Grid Engineering, Planning, Interconnection, and Permitting Work In the engineering and planning layer of the power sector, AI compresses modeling, scenario generation, document drafting, and first-pass analysis. The human role does not disappear so much as migrate toward validation, liability-bearing judgment, stakeholder negotiation, and interpretation of increasingly complex machine-generated results. | Role | Current Tasks | Employment | R/S/O | Dominant AI System | Task-Content Shift | Role Direction | Horizon | | ------------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | -------------- | ----------------------- | | Protection engineer 17-2071 subset | Set protective relay coordination studies; review fault data and post-event analysis; specify relay hardware; validate settings against NERC PRC standards. | Subset of 20,330 electrical engineers Protection-specific count is not separable in NAICS 221100. | R/S | Outage prediction; physics-informed ML for relay-setting optimization; AI-assisted post-event fault analysis. | Settings recalculation compresses. Engineers validate adaptive settings generated by physics-informed models; physical inspection and liability-bearing review remain human. | ↔ | Long | | Power systems planning engineer 17-2071 / 17-2072 | Conduct annual transmission planning studies, including N-1 and N-2 contingencies; perform generation interconnection impact studies; prepare load forecasts for integrated resource plans; support FERC Form 715 filings. | Subset of \~20,550 engineers Planning-specific count is not separable in NAICS 221100. | R | Forecasting and planning; ML-driven load and resource-forecast automation; AI-assisted production-cost modeling in PSS/E and PowerWorld; LLM-assisted regulatory filing drafting. | Study workflows compress. Planners become curators of model outputs and validators of scenarios, while regulatory sign-off and liability-bearing review remain human. | ↓ | Medium | | Transmission interconnection engineer 17-2071 / 17-2072 | Perform interconnection impact studies for generation and load additions; conduct power-flow, contingency, and stability analyses; submit data into ISO/RTO study processes. | Subset of \~20,550 engineers Additional headcount sits at ISOs/RTOs and engineering consultancies outside NAICS 221100. | R | Forecasting and planning; grid optimization; AI-assisted load-flow, transient-stability, and EMT analyses; cluster-study automation; LLM-assisted study-report drafting. | Engineers shift from study execution to validation of automated results, scoping of cluster studies, and interpretation of system impacts. | ↓ | Near task / Medium role | | Power electronics / grid modernization engineer 17-2072 | Specify inverter and converter control architectures; conduct grid-forming inverter studies; support HVDC and FACTS commissioning; develop technical standards for IBR interconnection. | \~1,000–2,000 utility-side estimate Small in NAICS 221100; most power-electronics engineering sits at OEM manufacturers in NAICS 335\. | R | Grid optimization and DER coordination; ML-assisted inverter-control design; AI-driven HVDC/FACTS operation studies; physics-informed ML for grid-forming inverter analysis. | Simulation-heavy work compresses. Engineers become curators of model libraries and validators of AI-generated control schemes. | ↑ | Medium | | Energy storage engineer 17-2072 / emerging | Design battery storage systems for utility and market participation; develop dispatch and value-stacking strategies; analyze battery degradation and warranties; perform siting and sizing studies. | \~500–1,500 Not separately coded in OEWS; emerging role across NAICS 221100. | R | Forecasting and planning; grid optimization; ML-driven battery dispatch optimization; degradation models; value-stacking optimization; AI-assisted siting and sizing. | Optimization computations become heavily AI-driven. Engineers focus on model validation, financial value-stack analysis, and regulatory compliance for storage market participation. | ↑ | Medium | | Reliability and resource adequacy analyst Institutional; ISOs/RTOs | Conduct stochastic loss-of-load expectation studies; determine reserve-margin requirements; perform capacity-expansion modeling; participate in FERC and state proceedings on resource adequacy. | \~300–600 across seven ISOs/RTOs Not in NAICS 221100; subset of roughly 6,400 ISO/RTO headcount. | R | Forecasting and planning; stochastic production-cost modeling; capacity-expansion modeling; ML-assisted scenario generation; probabilistic resource-adequacy assessments. | Model-run volume expands. Analysts become designers of scenarios and validators of results rather than the primary source of computational labor. | ↔ | Medium | | Environmental and permitting engineer 17-2081 / 19-2041 | Prepare environmental impact assessments and permit applications; coordinate with agencies on NEPA and state-equivalent reviews; conduct site characterization and stakeholder engagement; monitor compliance. | Several thousand cross-NAICS Not separately coded in NAICS 221100; many environmental engineers and scientists sit in NAICS 541620 under contract to utilities. | R | Forecasting and planning; LLMs for NEPA and state-permitting document drafting; AI-assisted regulatory compliance analysis; satellite imagery for environmental impact assessment. | Permit-application drafting compresses under LLM assistance. Environmental review judgment, agency negotiation, and public engagement remain human. | ↔ | Medium | R/S/O indicates the role’s dominant exposure: routine, safety-critical, or operationally constrained work. Arrows indicate directional pressure on the role, not a deterministic employment forecast. In this cluster, AI reduces the labor intensity of modeling, study execution, drafting, and scenario generation, while increasing the premium on engineering judgment, certification, interpretation, and accountability. ## Customer-Facing and Back Office A [Gartner survey of 321 customer service and support leaders conducted in October 2025](https://www.gartner.com/en/newsroom/press-releases/2025-12-02-gartner-survey-finds-only-20-percent-of-customer-service-leaders-report-ai-driven-headcount-reduction?ref=aixenergy.io) found that 20 percent had already reduced agent staffing due to AI. The survey was cross-industry; utility-specific compression has not been independently documented at the same precision, and utility customer service operations have historically lagged commercial sectors in adopting automation at the customer interface. The mechanism is the same. First-line inbound call handling for outage reporting, billing disputes, and rate structure questions is the substitutable task content, and LLM-based chatbots and outage self-service portals are the substituting technology. The CSR who remains handles exceptions: the calls the AI cannot resolve, the customers it loses, the regulatory complaints that escalate from the first two. That is the clearest near-term candidate for the same complete substitution that has already played out for the meter reader. Two other roles in the cluster run in the opposite direction. **Cybersecurity** is the cluster’s growth story. AI-driven security operations center monitoring, ML-based anomaly detection in operational technology and information technology traffic, and LLM-assisted threat intelligence synthesis compress the alert triage workload. At the same time, the AI-enabled attack surface is expanding. AI-driven phishing, deepfakes, and automated reconnaissance create threat volume that the analyst role exists to address. CIP compliance work persists regardless of AI deployment levels. Headcount grows. **Data analytics** is the back-office counter-trend. Routine reporting and dashboard creation increasingly self-service through AI-assisted tools. The analyst shifts toward data-pipeline architecture, model validation, and business-question framing, and utility headcount in the role grows as the AI deployment surface expands across the rest of the organization. The cluster’s nuanced case is the **rate case economist**. First-draft testimony, data-request responses, and commission-order analysis are high-LLM-exposure tasks. At firms that have adopted LLM-assisted drafting and analysis, the economist’s role is shifting toward reviewing AI-generated drafts and standing behind the analysis on cross-examination, rather than writing the drafts from scratch. But the credentialing value of cross-examination testimony and the liability exposure in rate case filings sustain the human role. Not a role the large language model eliminates. A role whose work the large language model substantially prefabricates. ### How AI Reshapes Customer, Regulatory, Analytics, and Cyber Utility Work The strongest near-term AI pressure falls on high-volume, text-heavy, and rules-based utility functions: customer service, billing, regulatory drafting, reporting, and first-pass analysis. At the same time, cyber and advanced analytics roles become more valuable as AI expands both operational capability and attack surface. | Role | Current Tasks | Employment | R/S/O | Dominant AI System | Task-Content Shift | Role Direction | Horizon | | ----------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------ | ----- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------- | ----------- | | Customer service representative 43-4051 | Handle inbound calls for outage reporting and billing disputes; arrange payments; answer rate-structure questions; escalate complex cases to regulatory specialists. | 18,580 | O | Large-load integration; forecasting and planning; LLM-based chatbots for first-line handling; AI-assisted billing dispute resolution; outage self-service portals. | First-line inbound call volume is the primary LLM substitution target. Customer service representatives increasingly handle exception escalations the AI cannot resolve. | ↓↓ | Near | | Demand response / DSM program manager 11-9199 subset | Design and manage demand-side management and energy-efficiency programs; coordinate cost-recovery filings; oversee measurement and verification; manage implementer contracts. | Subset of 4,280 managers DSM, EE, and DR counts are not separable in NAICS 221100; a large parallel workforce sits at program implementer firms. | R | Forecasting and planning; large-load integration; AI-driven customer segmentation; ML-based M&V; generative AI for outreach drafting; predictive analytics for opt-in and persistence. | Marketing drafting, M&V analysis, and routine reporting carry high LLM exposure. Program design judgment and regulatory commission engagement remain human-centered. | ↑ | Medium | | Rate case economist / regulatory analyst 13-2051 / 19-3011 subset | Prepare cost-of-service studies for rate cases; draft testimony on rate design and cost allocation; analyze commission orders for compliance; respond to data requests. | Very small Subset of financial and investment analysts, 1,840 in NAICS 221100, and economists, roughly 150 in NAICS 221100\. | R | Forecasting and planning; LLMs for filing drafting and data-request response; AI-assisted commission-order analysis; automated cost-of-service model execution. | First-draft testimony, data-request responses, and commission-order analysis carry high LLM exposure. The economist shifts toward expert validation of AI-generated analysis, while cross-examination, credentialing, and liability sustain the role. | ↓ | Near-medium | | Billing and metering analyst 43-3031 subset | Reconcile AMI meter data with billing records; investigate billing exceptions and disputes; audit rate-class assignments; calculate net-metering and time-of-use bills. | Subset of 2,710 clerks Billing-specific subset is not separable from bookkeeping, accounting, and auditing clerks in NAICS 221100\. | O | Asset monitoring; AMI-fed billing automation; ML-based billing exception detection; LLM-assisted dispute resolution. | Routine billing-exception triage substantially automates. Analysts handle complex multi-jurisdictional cases and large-customer issues requiring human judgment. | ↓↓ | Near | | Data analyst / business intelligence analyst 15-2051 | Build reporting dashboards for operations, finance, and regulatory functions; perform ad hoc AMI, SCADA, and customer-data analysis; support business-unit decisions with quantitative analysis. | 1,060 | O | Forecasting and planning; AutoML for utility-specific dashboards; LLM-based SQL generation; AI-assisted data-quality monitoring. | Routine reporting and dashboard creation increasingly become self-service through AI tools. Analysts shift toward data-pipeline architecture, model validation, and business-question framing. | ↑ | Medium | | Cybersecurity analyst 15-1212 | Monitor OT and IT environments for intrusion and anomaly; collect CIP compliance evidence and prepare for audits; investigate alerts; lead incident response; assess vulnerabilities and coordinate OT patching. | 1,150 | R/O | Outage prediction; AI-driven SOC monitoring; ML-based anomaly detection in OT/IT traffic; LLM-assisted threat-intelligence synthesis; automated incident-response playbooks. | Alert triage and threat-intelligence synthesis compress. Analysts shift toward threat hunting and CIP compliance work, while the AI-enabled attack surface expands. | ↑↑ | Near-medium | R/S/O indicates the role’s dominant exposure: routine, safety-critical, or operationally constrained work. Arrows indicate directional pressure on the role, not a deterministic employment forecast. In this cluster, AI substitutes most directly for high-volume administrative and first-pass analytical work, while increasing the value of cyber, data architecture, and expert judgment. ## Project Developers and Interconnection In March 2025, [FERC Commissioner David Rosner wrote to the seven independent system operators and regional transmission organizations](https://www.ferc.gov/news-events/news/commissioner-rosners-letters-isosrtos-regarding-interconnection-automation?ref=aixenergy.io) urging adoption of automation in interconnection study processes. The letter cited a specific result: a Midcontinent ISO automation tool called SUGAR had reproduced a two-year manual cluster study in ten days. Similar tools are emerging at Southwest Power Pool (the GridUnity platform) and at developer technical teams; an Amazon Web Services multi-agent architecture for connection impact assessment was published in November 2025, and a Grid-Mind system using LLM orchestration for sequential power flow, contingency, transient stability, and electromagnetic transient analyses is under peer review at *IEEE Transactions on Smart Grid*. This is the most dramatically compressed task content in the matrix. It is not a future scenario. It is current production deployment at one of the seven major US wholesale market operators. And it is not yet producing a corresponding role-level contraction in the interconnection queue analyst role, because the FERC Order 2023 cluster study reform expanded the volume of studies the analysts perform faster than the per-study automation reduces the labor each one requires. National interconnection queue backlogs sit above 2,000 gigawatts as of the most recent [Lawrence Berkeley National Laboratory *Queued Up* report](https://emp.lbl.gov/queues?ref=aixenergy.io). The analyst headcount holds for the next two to five years while the queue is large, then contracts over the five-to-fifteen-year window as automation reaches the per-study tasks faster than new studies enter the queue. That gap between the task content and the role count is the task-versus-role distinction from Part 1 visible at its starkest. The task content of the interconnection queue analyst’s day is shifting now, in months, through deployment of tools that demonstrably work. The role count holds, in years, because the institutional pressure (queue volume, regulatory deadline) sustains it. When the queue clears, the role contracts. Most of the development workforce around interconnection (originators, financiers, engineering consultants, project managers, EPC supervisors) sits in NAICS 237 (Specialty Trade Contractors) and NAICS 541 (Professional and Technical Services) depending on firm structure. None of it is counted in NAICS 221100, and there is no single dataset that cleanly represents the full developer universe across fuel types. The role headcount column in the table below uses qualitative direction arrows rather than estimated counts where the count is genuinely not knowable. ### How AI Reshapes Project Development, Interconnection, and Commercial Contracting Work In the project-development layer of the power sector, AI compresses siting analysis, queue studies, financial modeling, contract review, and compliance tracking. But the decisive work remains institutional: navigating interconnection queues, negotiating counterparties, managing stakeholder risk, and converting technical findings into financeable projects. | Role | Current Tasks | Employment | R/S/O | Dominant AI System | Task-Content Shift | Role Direction | Horizon | | ---------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------- | ----- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------- | ----------------------- | | Project developer / originator | Identify generation siting opportunities; negotiate land options and PPA structures; manage permitting and stakeholder engagement; develop financial models for investment committee review. | Not separable in OEWS Developer workforce typically appears in NAICS 237 or NAICS 541, depending on firm structure. | R | Large-load integration; forecasting and planning; LLMs for permit drafting; AI siting tools using GIS constraint mapping; ML-driven financial-model scenario analysis. | Siting analysis and financial modeling carry high LLM exposure. The originator’s value shifts toward relationship judgment, stakeholder strategy, and institutional queue navigation that AI does not replicate. | ↑ | Near-medium | | Interconnection queue analyst / study engineer | Prepare and review interconnection impact studies, including power-flow, contingency, and stability analysis; track queue position and cluster-study schedules; manage FERC-compliant data submissions; coordinate with ISO/RTO study teams. | Very small Concentrated in ISOs/RTOs and developer technical teams; not separable in OEWS. | R | Forecasting and planning; grid optimization; AI-assisted cluster-study tools; automated connection-impact analysis; LLM-orchestrated study workflows; multi-agent architectures for interconnection analysis. | Compression is already operational. Study workflows that once took years can be reproduced far faster in AI-assisted environments. Analysts shift from conducting studies to scoping, validating, and interpreting automated outputs, while managing institutional queue processes AI cannot navigate. Queue volume sustains near-term headcount. | ↔ near term↓ as tools mature | Near task / Medium role | | Power purchase agreement / contract manager 13-1041 subset | Negotiate and administer PPAs, EPC contracts, and interconnection agreements; manage milestone tracking and counterparty performance; coordinate with legal on disputes; analyze precedent contracts and market comparables. | Subset of 2,870 compliance officers Large additional workforce sits at developer firms and IPP affiliates outside NAICS 221100\. | R | Forecasting and planning; large-load integration; LLMs for PPA and EPC contract drafting and review; AI-assisted negotiation comparables; automated contract-compliance tracking; NLP-based counterparty risk extraction. | First-draft contract language, standard-provision review, and routine compliance tracking compress under LLM assistance. Counterparty negotiation, legal sign-off, and dispute resolution remain human. | ↑ | Medium | R/S/O indicates the role’s dominant exposure: routine, safety-critical, or operationally constrained work. Arrows indicate directional pressure on the role, not a deterministic employment forecast. In this cluster, AI accelerates the analytical and document-heavy parts of development, while the bottleneck shifts toward institutional judgment, queue navigation, stakeholder trust, negotiation, and accountability. ## Federal Regulators, ISOs, RTOs, and State Commissions In November 2024, PJM Interconnection terminated an internal XGBoost-based net-interchange forecasting project after struggling to maintain the ML staffing capacity required to keep the model in production. The project was technically functional. The constraint was institutional: PJM’s data science and ML engineering bench was not deep enough to sustain ongoing model maintenance, retraining, and validation alongside the queue of other work. PJM published the experience as a case study at the CIGRE 2024 conference. The episode is among the clearest documented examples of a particular dynamic that runs throughout this cluster: institutional defenses against AI’s role-level effects often hold not because the institution has explicitly defended them, but because the institution does not have the internal AI and machine learning capacity to deploy and sustain the systems that would otherwise erode the protected work content. The defense emerges from operational reality, not from intent. The institutions that regulate the US electric power industry and operate its wholesale markets do as much to shape what AI deployment looks like inside utilities as the utilities themselves do. FERC, NERC, the seven ISOs and RTOs, and the fifty state public utility commissions employ technical staff who analyze utility filings, certify operators, audit reliability standards, monitor markets, and produce the orders and rules that govern utility AI investment. The binding constraint on AI deployment inside these institutions is appropriation and statutory mission, not technological readiness. The role-level direction across the cluster is largely flat. Headcount at FERC has fluctuated with federal workforce policy and was reported at approximately 1,500 full-time employees in pre-2025 budget justifications, with subsequent estimates varying.[\[6\]](#%5Fftn6) NERC operates with around 340 staff, supplemented by several hundred at the six regional entities under delegated authority. ISO and RTO headcount across the seven entities sums to roughly 6,400 from 990 filings and human resources reports.[\[7\]](#%5Fftn7) State PUC technical staffing varies enormously across the fifty commissions, from large staffs at the California Public Utilities Commission, the New York Public Service Commission, and the Texas Public Utility Commission down to small commissions with handful-of-staff technical benches; rough national order of magnitude is 5,000–15,000 staff across all functions. The emerging role in the cluster, the AI governance analyst, runs in the opposite direction. The PJM experience and analogous ones at other ISOs/RTOs and large utilities are producing internal teams charged with the model development, validation, MLOps, and governance work that the prior generation of operations and IT staff was not built to handle. The role count is small. It is growing fast. ### How AI Reshapes Regulatory, Reliability, ISO/RTO, and AI-Governance Work In the institutional layer of the power sector, AI compresses filing review, docket analysis, compliance evidence review, market monitoring, contingency screening, and report drafting. But the center of gravity remains human: statutory authority, commissioner discretion, enforcement judgment, NERC certification, market-rule interpretation, and AI governance. | Role | Current Tasks | Employment | R/S/O | Dominant AI System | Task-Content Shift | Role Direction | Horizon | | -------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------ | ----- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------- | ----------- | | FERC technical analyst / attorney-adviser 23-1011 / 15-2051 / 13-2099 subsets; institutional | Analyze utility and ISO/RTO filings for FERC rule compliance; draft commissioner memoranda and proposed orders; conduct economic and engineering analyses for rulemaking; manage formal dockets and intervenor correspondence. | \~1,500 FTE Based on pre-2025 budget justifications; current count is uncertain following federal workforce changes. | R | Forecasting and planning; LLMs for filing analysis; NLP-based intervention drafting; AI-assisted Form 1 and EQR analytics; retrieval-augmented generation over docket archives. | Filing review and order drafting compress. Legal judgment, commissioner-level decision-making, and rulemaking discretion remain human; the binding constraint is appropriation, not capability. | ↔ | Medium-long | | NERC compliance analyst / CIP auditor Institutional | Conduct Critical Infrastructure Protection reliability-standard audits at registered entities; review compliance evidence; investigate potential violations; prepare compliance monitoring and enforcement reports. | \~300–500 CIP-specific audit positions NERC has roughly 340 staff; six regional entities add several hundred additional staff. | R/S | Asset monitoring; LLMs for evidence review and audit-report drafting; AI-assisted anomaly detection in registered-entity submissions; NLP-based reliability-standard interpretation. | Evidence review and compliance documentation carry high LLM exposure. Audit judgment and enforcement findings still require human sign-off under NERC statutory authority. | ↔ | Medium-long | | ISO/RTO market monitor / market analyst | Analyze real-time and day-ahead market results for competitive concerns; produce quarterly State of the Market reports; investigate participant behavior; develop rule-change proposals for market design. | \~200–400 across seven ISOs/RTOs Subset of roughly 6,400 ISO/RTO total headcount; function-specific counts are not consistently separable. | R | Forecasting and planning; ML-based market-behavior anomaly detection; NLP-based regulatory document analysis; AI-assisted data compilation for quarterly reports. | First-pass screening for market manipulation compresses. Market monitors shift toward judgment on ambiguous cases and formal regulatory proceedings requiring documented human analysis. | ↔ | Medium | | ISO/RTO operations analyst / real-time operator Institutional | Operate real-time energy and ancillary-services markets; supervise security-constrained economic dispatch; conduct contingency analysis and reliability operations across the footprint; coordinate outage scheduling with member transmission owners. | \~2,000–3,000 Operations and control-center subset across the seven ISOs/RTOs; part of roughly 6,400 total ISO/RTO headcount. | R/S | Outage prediction; grid optimization; AI-enhanced security-constrained economic dispatch; real-time contingency analysis with ML acceleration; generative-AI operator-assist tools. | The trajectory mirrors utility-side dispatch, but on the wholesale-market timescale. AI handles first-pass screening and recommendation; operators retain final dispatch authority under NERC reliability standards and ISO/RTO tariffs. | ↓ | Medium | | State PUC technical staff / rate analyst Institutional | Review utility rate-case filings; conduct discovery; draft interventions on cost-of-service, rate design, and resource planning; advise commissioners on technical issues; manage formal hearing records. | \~5,000–15,000 total staff Highly heterogeneous across fifty commissions; estimate spans technical, legal, and administrative functions. | R | Forecasting and planning; LLMs for utility-filing analysis; AI-assisted intervention drafting and discovery review; retrieval-augmented generation over commission orders and tariffs; automated comparables analysis for rate cases. | Filing review and discovery work compress. Commission deliberation, public-hearing engagement, and order drafting remain human; smaller PUCs are particularly exposed. | ↓ | Medium | | ISO/RTO data scientist / AI governance analyst Emerging | Develop and validate ML models for ISO/RTO operational and market functions; manage MLOps pipelines and model monitoring; oversee third-party AI deployments in market and operational systems; coordinate cross-functional AI governance. | \~50–200 Small but growing across the seven ISOs/RTOs and a handful of large utilities. | R/O | Forecasting and planning; grid optimization; internal AI model development; validation frameworks; MLOps; third-party AI governance. | This role is being created rather than disrupted. The supervisory function becomes formalized inside institutions where AI capability, reliability accountability, and governance must meet. | ↑↑ | Near | R/S/O indicates the role’s dominant exposure: routine, safety-critical, or operationally constrained work. Arrows indicate directional pressure on the role, not a deterministic employment forecast. In this cluster, AI accelerates institutional analysis but also creates a new governance layer: model validation, auditability, reliability accountability, and human authority over machine-generated recommendations. ## What Patterns Hold Across the Matrix Three patterns run across the forty-row inventory. The first is that the reliability-safety axis from [Part 1 ](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fi/)explains more of the variation in role-level outcomes than any other single variable. Reliability roles (dispatchers, market analysts, billing analysts, customer service representatives, rate case economists, regulatory filing analysts) face headcount contraction over the next two to fifteen years. Safety roles (nuclear reactor operators, line installers, substation and relay repairers, protection engineers) face flat or slow-contracting headcount over fifteen-year-and-longer timeframes. The same machine learning techniques drive task-content compression in both categories, but the institutional architecture surrounding the safety roles is harder to attenuate, and the four-stage mechanism therefore plays out on two timetables at once. The second pattern is the task-versus-role distinction from [Part 1](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fi/) visible at full scale. Task content is shifting in every cluster over the next two to five years, including in roles whose headcount is growing. Wind technicians and solar technicians are growth roles whose daily work content is changing substantially. Interconnection queue analysts are roles whose headcount holds in the near term while their study workflows have already been compressed by two orders of magnitude in production deployment. Cybersecurity analysts are growth roles whose alert triage workload has compressed under AI tooling at the same time their threat environment has expanded. None of these patterns shows up in occupational employment data, because occupational employment data counts roles, not what the people in them do. The third is the shape of the role-level growth in the matrix. Most rows show flat or contracting headcount. A small but consistent set shows growth: cybersecurity, data analytics, power electronics, energy storage engineering, distribution system operations, and the emerging AI governance analyst role at ISOs/RTOs and large utilities. These are not roles where AI replaces other roles. They are roles created or expanded by AI deployment itself: positions that exist to handle the AI-enabled attack surface, to manage the deployment surface across the rest of the utility, to design the inverter-based generation that AI helps integrate, or to keep the ML models the utility increasingly depends on in production. The PJM XGBoost retirement is the cautionary version of the same finding: a utility cannot deploy AI without the workforce to sustain it. The matrix shows the shape of that workforce coming into existence: small today, growing fast, and concentrated in the institutions with the deepest internal engineering benches. **For utilities.** Workforce planning that extrapolates from the current org chart will miss in both directions. The roles utilities will need more of (cybersecurity analysts, data scientists, power electronics and energy storage engineers, distribution system operators, AI governance staff) require talent the utility industry has not historically competed for, on compensation bands that mostly don’t exist inside utility salary structures. The roles utilities will need fewer of (customer service representatives, billing analysts, market traders, dispatchers in the long run, rate case economists) are filled by workers whose retraining is not automatic and whose union representation often is. A workforce plan that does not engage both halves of that rebalance, and does not engage the unions on the contracting side from the start, ends up short on the talent it needs and unable to release the talent it does not. **For regulators.** The certification and licensing frameworks that protect safety-critical roles (NERC System Operator Certification, NRC operator licensing, state union contracts around line work and dispatch) hold against role-level headcount change in the near and medium terms. They hold less well against the gradual erosion of the work content the frameworks were originally written to certify. The competencies the frameworks codify describe a directly-judging operator. The work being done day-to-day at the leading control rooms is already moving toward something else. The certifying authorities can revise the frameworks to describe the work that is now being done, or they can wait until the gap between credential and practice becomes too large to ignore. Both paths end at the same place. The second one ends there with more disruption. **For investors and capital allocators.** AI’s productivity effect on utilities is concentrated in coordination-layer functions (dispatch, planning, market operations, customer service, regulatory filing) and in the institutional infrastructure surrounding the industry (interconnection studies, market monitoring, compliance, technical review). It is less concentrated in the physical-infrastructure operations and maintenance functions where utility capex tends to focus. Utility-AI investment narratives that emphasize physical infrastructure deployment and vendor relationships will miss most of the productivity story. The utilities and ISOs/RTOs with the internal AI and ML engineering capacity to deploy and sustain these systems will capture more of the gain than those pursuing AI primarily through vendor procurement. Part 4 takes up the capability-creation half of the AI and robotics story: the new activities utilities will be able to perform for the first time, drawing on [Hannah Kaplan’s](https://www.aixenergy.io/author/hannah/) [*From Reactive Reliability to Dynamic Intelligence*](https://www.aixenergy.io/from-reactive-reliability-to-dynamic-intelligence-how-ai-is-building-the-grid-we-actually-need/). Part 5 returns to strategy, with specific recommendations for utilities, ISOs and RTOs, regulators, capital allocators, and the workforce itself, and engages directly with the productivity claims in current consultancy framings of AI in utilities. The translation from AI capability to utility workforce change runs through institutional defenses that hollow out from within rather than break from outside. Task content inside roles shifts first and most visibly. Role-level headcount follows on longer timelines that differ systematically between reliability and safety functions. The decade ahead looks slow until it does not. ## Notes \[1\] BLS Occupational Employment and Wage Statistics, SOC 43-5041 (Meter Readers, Utilities) in NAICS 221100 (Electric Power Generation, Transmission and Distribution): May 2006, 10,510; May 2016, [6,230](https://www.bls.gov/oes/2016/may/oes435041.htm?ref=aixenergy.io); May 2023, [3,910](https://www.bls.gov/oes/2023/may/oes435041.htm?ref=aixenergy.io); May 2025, [3,320](https://www.bls.gov/oes/current/oes435041.htm?ref=aixenergy.io). \[2\] Title XIII of the [Energy Independence and Security Act of 2007](https://www.congress.gov/bill/110th-congress/house-bill/6?ref=aixenergy.io) established the federal Smart Grid policy framework. The American Recovery and Reinvestment Act of 2009 authorized Smart Grid Investment Grants. PG&E’s SmartMeter program began in 2006 with California Public Utilities Commission approval expanded in 2009; the program installed approximately nine million meters by 2012\. AMI penetration figures from [EIA Form EIA-861](https://www.eia.gov/tools/faqs/faq.php?id=108&ref=aixenergy.io): 23 percent of US electricity customers in 2011, 47 percent in 2016, 68 percent in 2021, 72 percent in 2022. \[3\] Cluster 6 (project developers and interconnection) draws on NAICS 237 (specialty trade contractors) and NAICS 541 (professional and technical services) where developer firms sit. Cluster 7 (federal regulators, ISOs/RTOs, state commissions) uses IRS Form 990 filings, official ISO and RTO human resources reports, and FERC and NERC budget justifications, because federal and state government employment is coded outside NAICS 221100 and ISOs and RTOs are not consistently coded anywhere in OEWS. OEWS groups protection engineers, planning engineers, interconnection engineers, and general electrical engineers under SOC codes 17-2071 and 17-2072 without further subdivision; estimates within those codes are approximations, identified as such in the relevant rows. \[4\] The six clusters: forecasting and planning; outage prediction and reliability support; wildfire mitigation and vegetation management; asset monitoring and predictive maintenance; grid optimization and DER coordination; large-load integration and data center flexibility. \[5\] [US Energy and Employment Report 2025](https://www.energy.gov/policy/articles/united-states-energy-and-employment-report-2025?ref=aixenergy.io) (DOE, August 2025). Of the 933,800 workers counted in electric power generation across all NAICS codes: 32.5 percent in Construction, 22.0 percent in Professional and Business Services, 20.3 percent in Utilities itself. \[6\] FERC budget justifications pre-2025 cited approximately 1,500 full-time employees. Third-party aggregators (including Prospeo and others) have reported lower estimates following 2025 federal workforce policy changes. The figure in the matrix is approximate and identified as such. \[7\] ISO/RTO headcount from IRS Form 990 filings (most recent available): MISO 1,271; NYISO 714; ISO-NE 764; CAISO 788; SPP 889\. ERCOT 960 from public HR reporting. PJM approximately 1,015 from third-party aggregators (PJM is structured as an LLC and does not file Form 990). NERC headcount approximately 340 from RocketReach (NERC does not file Form 990). Total approximately 6,400 across the seven entities. ### Is AI's Next Frontier Underwater and in Orbit? URL: https://www.aixenergy.io/space_and_sea/ Last updated: 2026-07-29T19:42:39.000Z *by* [*Morgan D. Bazilian*](https://econbus.mines.edu/project/bazilian-morgan/?ref=aixenergy.io)*,* [*Alex Gilbert*](https://payneinstitute.mines.edu/project/alex-gilbert/?ref=aixenergy.io) *and* [*Lt. Col. Jahara “Franky” Matisek (PhD)* ](https://usnwc.edu/Faculty-and-Departments/Directory/Jahara-Franky-Matisek?ref=aixenergy.io) For decades, computing infrastructure was fixed to land, but that assumption is breaking down. Artificial intelligence (AI) is [driving](https://www.brookings.edu/articles/global-energy-demands-within-the-ai-regulatory-landscape/?ref=aixenergy.io) such intense demand for power, cooling, critical [minerals](https://nationalinterest.org/blog/techland/why-the-us-military-could-lose-the-contest-for-materials-crucial-to-ai?ref=aixenergy.io), and physical infrastructure that firms have begun to look offshore – and even off-planet – for the next generation of compute capacity. At the same time, power grids are congested, connecting to the grid takes an exorbitant amount of time, and political resistance to hyperscale data center [infrastructure](https://www.fpri.org/article/2025/11/data-centers-at-risk-the-fragile-core-of-american-power/?ref=aixenergy.io) is rising. An estimated $64 billion in U.S. data center [projects](https://www.datacenterwatch.org/report?ref=aixenergy.io) has already been blocked or delayed due to local opposition over land use, electricity demand, water consumption, and environmental concerns. Article Summary 0:00 /68.732109 1× As land-based data center constraints intensify, firms are searching for new physical environments capable of supporting the next generation of compute infrastructure. In early May, Peter Thiel led a [$140 million financing round](https://finance.yahoo.com/sectors/energy/articles/panthalassa-raises-140-million-power-153900222.html?ref=aixenergy.io) for Panthalassa, a company seeking to build floating data centers on the high seas. Around the same time, Anthropic [announced](https://www.cnbc.com/2026/05/06/anthropic-spacex-data-center-capacity.html?ref=aixenergy.io) a partnership with SpaceX tied to orbital compute ambitions. While this partnership is nascent, it provides some evidence of commercial appetite for orbital compute is getting closer. These developments suggest that some computational infrastructure may soon be heading off-land. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/05/bowenscolo_Ultra-realistic_editorial_magazine_cover_image_for_a_e04b3bf9-350f-4975-b115-d8cca9ff7b79.png) Off-land compute is creating a new geography of strategic competition where commercial cloud architecture, military support systems, intelligence infrastructure, energy networks, and geopolitical rivalry increasingly overlap. Subsea and orbital data centers promise resilience, geographic flexibility, and access to new energy and cooling environments. But they also move computational capacity into domains that are harder to govern, more difficult to defend, and potentially exposed to coercion, espionage, and disruption. China appears to have recognized this shift faster than many Western governments. Beijing is already [positioning](https://nationalinterest.org/blog/techland/chinas-subsea-ambition-powering-the-ai-race-with-global-thermal-debt?ref=aixenergy.io) subsea compute infrastructure tied to offshore energy projects while seeking to [invest](https://spacenews.com/china-backs-orbital-data-center-startup-with-8-4-billion-in-credit-lines/?ref=aixenergy.io) $8.4 billion into orbital computing constellations capable of distributed processing in space. This massive funding is still far from fully inked, but the direction of travel, and the pace China can move in infrastructure makes this a significant signal. The United States still leads in private-sector innovation in this realm, particularly in advanced chips, launch capacity, and cloud architecture. Yet American efforts remain fragmented across regulatory authorities, agencies, and commercial actors that were not designed to treat off-land compute as a national security issue. Existing institutions governing satellites, export controls, maritime infrastructure, and telecommunications were built for separate domains. Off-land compute blurs them. Risks also differ substantially between subsea and orbital systems. Subsea compute expands digital infrastructure into [contested](https://digital-commons.usnwc.edu/cgi/viewcontent.cgi?article=1000&context=ciwag-policy-papers&ref=aixenergy.io) maritime environments already vulnerable to cable sabotage, gray zone activities, and intelligence collection. Orbital compute, by contrast, raises new questions surrounding military AI support, anti-satellite targeting, dual-use escalation, and wartime communications resilience. Treating these as parallel examples of the same phenomenon obscures the fact that they create distinct operational and geopolitical challenges. Washington still has time to shape how these systems develop, but working with allies in these two realms will be necessary. The problem is not legal authority, but strategic coordination and prioritization. As is typical, the technology developments tend to be ahead of the ability of governments to regulate them. Off-land compute sits awkwardly between maritime governance, space policy, telecommunications regulation, export controls, critical infrastructure protection, and national security planning. The countries that dominate the geography of computation may shape the future balance of economic and military power as profoundly as those that once controlled oil, shipping lanes, and undersea cables. ## **Subsea Compute: Efficiency and Physical Denial** 💡 Subsea data centers could ease AI’s power, cooling, land, and permitting constraints by using the ocean as a natural heat sink and placing compute closer to coastal demand centers. But they also introduce new strategic vulnerabilities because they depend on undersea cables, offshore energy systems, and maritime infrastructure that can be targeted, monitored, or disrupted. Subsea data centers attempt to solve several of AI’s emerging infrastructure constraints simultaneously. The ocean provides a stable and effectively unlimited heat sink, reduces cooling costs, and allows compute infrastructure to sit closer to coastal population centers and network nodes. Offshore deployment may be able to partially bypass some of the social license and land permitting increasingly slowing terrestrial hyperscale expansion. The most prominent early experiment came from Microsoft, whose [Project Natick](https://natick.research.microsoft.com/?ref=aixenergy.io) began in 2013 when a former U.S. Navy submariner proposed the concept. Phase 2 of the Project [demonstrated](https://www.blackridgeresearch.com/project-profiles/microsoft-data-center-in-ocean-sea-project-natick-latest-update?ref=aixenergy.io) significant ambitions: a 40-foot sealed cylindrical vessel containing 12 racks, 864 servers, and up to 27.6 petabytes of storage capacity. In June 2018, it was deployed 117 feet below the surface near the Orkney Islands, Scotland. Yet in June 2024, Microsoft [confirmed](https://www.datacenterdynamics.com/en/news/microsoft-confirms-project-natick-underwater-data-center-is-no-more/?ref=aixenergy.io) that Project Natick was no longer active. The primary reason was inflexibility: underwater pods remain sealed for years, [making](https://www.blackridgeresearch.com/project-profiles/microsoft-data-center-in-ocean-sea-project-natick-latest-update?ref=aixenergy.io) it impossible to upgrade GPUs or expand to meet the exploding demands of AI workloads. China, meanwhile, has moved from experimentation to deployment. The Hainan underwater data center [operates](https://merics.org/en/comment/china-commercializing-energy-efficient-underwater-data-centers?ref=aixenergy.io) with a Power Usage Effectiveness (essentially measuring energy efficiency of a data center) of 1.1, with efficiency gains of 40 to 60 percent relative to traditional facilities. More importantly, Beijing appears to view subsea compute as part of a broader effort to integrate offshore energy, digital infrastructure, and maritime industrial capacity. China’s subsea ambitions also extend beyond Hainan. In October 2025, [Highlander](https://hongkongfp.com/2025/10/11/china-trials-energy-saving-underwater-data-centres/?ref=aixenergy.io) was completing a second facility off the coast of Shanghai’s Lin-gang Special Area at a cost of approximately ¥1.6 billion ($226 million), with the facility [designed](https://newatlas.com/energy/wind-powered-underwater-data-center/?ref=aixenergy.io) to draw nearly all of its power from nearby offshore wind farms. Elsewhere, underwater compute concepts are attracting growing interest. A proposed facility off Ulsan, South Korea, with [plans](https://www.datacenterdynamics.com/en/news/underwater-data-center-proposed-off-the-coast-of-ulsan-south-korea-secures-govt-funding-report/?ref=aixenergy.io) for 100,000 seabed servers, secured government funding in early 2026\. In the United States, startups such as Subsea Cloud and DeepGreen Western Passage are [pursuing modular and tidal-energy-powered concepts](https://www.datacenterdynamics.com/en/news/underwater-data-center-proposed-off-the-coast-of-ulsan-south-korea-secures-govt-funding-report/?ref=aixenergy.io) aimed at commercial deployment. Yet even offshore, permitting remains difficult. NetworkOcean, a U.S. startup that sought to deploy underwater units in San Francisco Bay, ran into regulatory [obstacles](https://www.datacenterdynamics.com/en/news/underwater-data-center-proposed-off-the-coast-of-ulsan-south-korea-secures-govt-funding-report/?ref=aixenergy.io), and pivoted toward floating facilities instead. For military planners and strategists, subsea compute infrastructure is a resilience opportunity paired with new vulnerabilities. These systems rely on undersea cables, offshore energy infrastructure, and maritime logistics networks that adversaries may target. Subsea compute nodes will likely simultaneously support civilian cloud processing, AI model training, intelligence functions, and military software, blurring distinctions between commercial and strategic infrastructure. In a crisis or conflict, these systems could be attractive targets for sabotage, espionage, destruction, or gray zone activities. Iran’s willingness to [target](https://nationalinterest.org/blog/techland/when-the-cloud-becomes-a-target-the-future-of-war-is-your-internet?ref=aixenergy.io) Gulf data centers during the 39-day war demonstrated that digital infrastructure is open to coercion. ## **Orbital Compute: Persistent Power, Persistent Exposure** 💡 Orbital compute could provide distributed AI processing powered by near-continuous solar energy, less dependent on strained grids, land-based data centers, or undersea cables. But it also turns compute into dual-use space infrastructure, creating new military, cybersecurity, anti-satellite, and governance risks as the U.S., China, Europe, and private firms race toward deployment. Orbital compute systems pursue different objectives than subsea infrastructure. In low Earth orbit, solar power can be nearly continuous, communications do not depend on seabed cables, and cooling occurs through radiation into space. In theory, orbital systems could provide geographically distributed compute capacity less dependent on vulnerable terrestrial infrastructure and increasingly strained electrical grids. A 2025 [study](https://www.nature.com/articles/s41928-025-01476-1?ref=aixenergy.io) in Nature Electronics found that space-based computing could potentially operate with minimal carbon impact if energy generation and thermal management systems are effectively integrated. [Significant engineering hurdles remain](https://coe.northeastern.edu/news/the-future-of-space-computing-and-the-hurdles-for-orbital-ai/?ref=aixenergy.io), including mass-efficient thermal radiation, space debris mitigation, on-orbit assembly, and radiation-hardened electronics. Yet the pace of investment suggests major firms increasingly view orbital compute as plausible. Starcloud, founded in 2024 by a team including former SpaceX, Airbus, and McKinsey personnel, has emerged as one of the most visible American entrants. In September 2024, the company published a white paper arguing that orbital data centers are, [“feasible, economically viable, and necessary to realize the potential of AI.”](https://www.nature.com/articles/d41586-026-01370-6?ref=aixenergy.io) In November 2025, Starcloud launched Starcloud-1 aboard a SpaceX rocket into low Earth orbit carrying an NVIDIA H100 GPU, reportedly [the most powerful AI-focused processor yet deployed in space](https://www.ycombinator.com/companies/starcloud?ref=aixenergy.io). By December 2025, the company claimed to have [trained a large language model in orbit](https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html?ref=aixenergy.io) using Google DeepMind’s Gemma model. Google has also entered the field through [Project Suncatcher](https://www.semafor.com/article/11/04/2025/google-wants-to-build-solar-powered-data-centers-in-space?ref=aixenergy.io), announced in November 2025\. The initiative proposes fleets of satellites, functioning as a distributed orbital compute architecture capable of processing and transmitting large volumes of data without relying entirely on terrestrial cloud infrastructure. China has again moved faster in this realm than many Western counterparts toward operational deployment. In May 2025, a Long March-2D rocket [launched](https://english.www.gov.cn/news/202505/15/content%5FWS6825452ec6d0868f4e8f28e6.html?ref=aixenergy.io) the first 12 satellites of the “Three-Body Computing Constellation,” a joint effort between ADA Space and Zhejiang Lab. Each satellite [carries](https://www.china-in-space.com/p/chinas-space-enterprises-quietly?ref=aixenergy.io) a large AI model with 8 billion parameters capable of 744 TOPS (tera operations per second), with 100-gigabit inter-satellite laser links, enabling the 12 satellites to jointly attack computing tasks as a distributed system. More importantly, the project suggests Beijing increasingly views orbital compute not merely as commercial infrastructure, but as part of a broader strategy integrating AI, satellite communications, and space-industrial capabilities. Europe has taken a more research-oriented approach. The European Space Agency has [funded](https://www.esa.int/Enabling%5FSupport/Preparing%5Ffor%5Fthe%5FFuture/Discovery%5Fand%5FPreparation/Knowledge%5Fbeyond%5Four%5Fplanet%5Fspace-based%5Fdata%5Fcentres?ref=aixenergy.io) studies with IBM and KP Labs examining orbital data centers capable of processing satellite-generated data directly in space, without requiring constant downlinking to Earth. The ESA’s Φsat-2 satellite, [launched](https://www.cutter.com/article/orbit-data-centers-mapping-leaders-space-ai-computing?ref=aixenergy.io) in 2024, already carries an onboard AI accelerator capable of autonomous cloud detection, vessel identification, and marine anomaly monitoring, providing an early model for orbital edge computing at smaller scale. The speed of development is already becoming apparent to regulators. Since January 2026, the Federal Communications Commission has [received](https://www.gao.gov/products/gao-26-109012?ref=aixenergy.io) multiple applications from U.S. companies seeking approval for satellite constellations intended to function as orbital data centers. For the Pentagon, orbital compute represents both opportunities and vulnerabilities. Unlike subsea systems, which remain physically tethered to maritime infrastructure, orbital compute could provide resilient and geographically distributed processing capacity even when land-based data center infrastructure is attacked. In a major conflict, distributed orbital compute architectures could reduce dependence on vulnerable land-based cloud systems and undersea cable networks while accelerating data processing closer to the point of collection. At the same time, orbital compute could further expand the amount of dual-use infrastructure operating in space. These systems would exist within a domain that is observable and congested. Worse, it can now be [contested](https://mwi.westpoint.edu/space-as-a-gray-zone-the-future-of-orbital-warfare/?ref=aixenergy.io) by various anti-satellite weapons, cyber operations, electronic warfare, and counterspace capabilities. Orbital data centers may become attractive targets during a conflict or crisis because they might support both civilian and military functions simultaneously. Even purely civilian orbital compute infrastructure could become a target during crises as a form of economic coercion. While orbital compute is, today, eye wateringly expensive, traditional economic metrics do not always apply to domains where military dominance is prioritized. Additionally, the launch costs, which will make up a decent part of the overall budgets, are rapidly falling and tied to innovations with low-earth orbit satellites. ## **The Governance Gap and Operational Risk** 💡 Off-land compute is becoming a strategic competition over who controls the next geography of AI infrastructure, where energy, communications, cloud systems, seabed assets, and space networks converge. The United States still has major technological advantages, but it needs coordinated governance, domain-specific rules, and defense planning before subsea and orbital compute become critical—and vulnerable—infrastructure. Subsea and orbital compute systems reflect a new dimension of strategic competition in which computational capacity, energy infrastructure, and communications networks increasingly overlap. China is already deploying subsea systems at scale while simultaneously investing in orbital computing architectures. Meanwhile, the United States but faces greater regulatory friction and institutional fragmentation in translating its substantial technological advantages into operational infrastructure. American institutions still govern compute as if it were geographically fixed to terrestrial infrastructure, but moving compute off-land changes how systems are protected, and how they fail. Washington needs a new governance strategy with three priorities. The first priority is improving strategic coordination across existing regulatory and national security institutions. Those authorities are scattered across institutions built for separate worlds: satellites, submarine cables, export controls, cloud services, maritime infrastructure, energy systems, and foreign investment screening. Off-land compute fuses those worlds together, and that could be better brought together in a more coordinated process. The United States does not need to create a new bureaucracy for every new platform, but it does need a lead interagency process that treats off-land compute as strategic infrastructure rather than bouncing it between licensing offices. While more coordination would be welcome, specific domain differences will need to be accommodated between sea and space. The second priority is to build domain-specific rules: Subsea compute should be viewed through the lens of seabed infrastructure competition. It intersects with submarine cables, offshore energy, coastal permitting, environmental review, fisheries, port access, repair vessels, and maritime security. The 2025 effort by the Federal Communications Commission to [modernize](https://docs.fcc.gov/public/attachments/DOC-413057A1.pdf?ref=aixenergy.io) submarine cable rules shows that Washington already recognizes how communications infrastructure carries national security, law enforcement, foreign policy, and trade risks. Orbital compute is a different problem set. It raises questions about satellite licensing, spectrum, optical links, export controls, cyber resilience, space traffic management, counterspace vulnerability, and dual-use escalation. Finally, the Pentagon must treat off-land compute as future operational infrastructure, not as another tech trend. The U.S. military (and its [allies](https://resiliencemedia.co/europes-armed-forces-are-too-reliant-on-us-cloud-providers-report-finds/?ref=aixenergy.io)) already [depends](https://www.datacenterdynamics.com/en/analysis/data-centers-are-war-infrastructure-now/?ref=aixenergy.io) heavily on commercial cloud providers, and its own cloud security guidance recognizes that cloud providers are responsible for both physical and cyber protection of the systems they operate. That assumption becomes more complicated once compute moves into contested seas and orbit. Subsea systems will become part of the same [vulnerable](https://foreignpolicy.com/2025/06/04/seabed-china-russia-unclos/?ref=aixenergy.io) seabed ecosystem as cables, offshore energy, and maritime logistics. Orbital compute may strengthen wartime resilience, but it also creates new dual-use targets in a domain exposed to anti-satellite weapons, cyber operations, electronic warfare, and counterspace operations. The Pentagon must wargame how these systems are defended, degraded, repaired, and reconstituted before they become indispensable. It also means figuring out how joint warfighting would look if AI and compute were no longer available. In both space and the seas, working with allies will be essential. Both domains have varying and decades-long levels of international collaboration, from the Artemis Accords for space to the UN Convention on Law of the Sea. Still, neither of those frameworks is universally accepted. Compute adds another layer to this complex legal and geopolitical landscape. Compute is moving off the land domain because everyone is looking for power, cooling, scale, and geographic flexibility. Subsea, floating, and orbital systems will likely become important components of digital infrastructure. But they will also move strategically important computational capacity into domains that are harder to govern, more difficult to defend, and increasingly exposed to geopolitical competition. The ability to prioritize these areas and invest in them will be challenging in the short term. Limited government investment is shaped by politics. As an example, there is competition from topics as far ranging as quantum systems or nuclear fusion. The emerging geography of AI infrastructure is potentially vast, stretching from space to the sea floor. If massive spending continues in this sector, these developments can lead the way in developing these formerly inhospitable realms. ## Sign up for AIxEnergy AIxEnergy analyzes how artificial intelligence is reshaping electricity demand, grid operations, and energy infrastructure, with a focus on system risk, policy, and institutional decision-making. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. [*Morgan D. Bazilian*](https://econbus.mines.edu/project/bazilian-morgan/?ref=aixenergy.io) *is the director of the Payne Institute and professor at the Colorado School of Mines, with over 20 years of experience in global energy policy and investment. A former World Bank lead energy specialist and senior diplomat at the UN, he has held roles at NREL and in the Irish government, and advisory positions with the World Economic Forum and Oxford. A Fulbright fellow, he has published widely on energy security and international affairs.* [*Alex Gilbert*](https://payneinstitute.mines.edu/project/alex-gilbert/?ref=aixenergy.io) *is a Fellow at the Payne Institute for Public Policy, where his work spans energy security, seabed and space resource governance, and long-range technology policy. Previously, he was project manager at the Nuclear Innovation Alliance, advancing regulatory modernization and federal policy frameworks on next-generation nuclear reactors. Alex is also adjunct faculty at Johns Hopkins University, and a PhD candidate in Space Resources at the Colorado School of Mines, researching astronaut nuclear safety, lunar commercialization, and off-world resource strategy.* [*Lt. Col. Jahara “Franky” Matisek*](https://usnwc.edu/Faculty-and-Departments/Directory/Jahara-Franky-Matisek?ref=aixenergy.io) *(PhD) is a senior fellow at the Payne Institute for Public Policy and a visiting scholar at Northwestern University. He has published over 200 articles on industrial base issues, strategy, and warfare.* The views expressed are those of the author and do not reflect the official position of the U.S. Air Force, Department of the Air Force, Department of War, or the U.S. Government. ### AI Load Corridors and the Rise of Orchestration Utilities URL: https://www.aixenergy.io/ai-load-corridors-and-the-rise-of-orchestration-utilities-how-the-nextera-dominion-merger-signals-a-new-era-of-ai-infrastructure-utilities/ Last updated: 2026-07-13T06:06:58.000Z AI infrastructure is transforming utilities from electricity providers into infrastructure orchestration platforms. The future advantage may belong to operators that can synchronize power, transmission, cooling, fiber, and capital deployment under conditions of extreme uncertainty. _This post is for paying subscribers only._ ### FERC Meets the Shadow Grid: Deciding when Private Power becomes a Public Grid Problem URL: https://www.aixenergy.io/ferc-and-ai/ Last updated: 2026-07-29T19:40:22.000Z AI data centers are pushing FERC to confront the Shadow Grid, where private power, co-located generation, and hyperscale load blur into public grid obligations, forcing a shift from queue chaos to coordinated infrastructure governance. _This post is for paying subscribers only._ ### The Emerging Grid Dynamics Crisis Beneath AI URL: https://www.aixenergy.io/when-ai-stopped-behaving-like-load-the-emerging-grid-dynamics-crisis-beneath-ai/ Last updated: 2026-07-13T06:07:25.000Z AI infrastructure is no longer just a power-demand story. NERC’s rare Level 3 alert suggests hyperscale AI may alter the dynamic behavior of the grid itself, forcing utilities to rethink reliability, operability, and infrastructure planning in the age of machine-speed load synchronization. _This post is for paying subscribers only._ ### The Power Sector's AI Decade Part II URL: https://www.aixenergy.io/the-power-sectors-ai-decade_part_ii/ Last updated: 2026-07-29T19:41:46.000Z How AI reaches the US electricity workforce depends less on what the technology can do than on how regulated industries absorb new capability in practice. The process has structure. It plays out in stages, each with its own pace, and the stages differ systematically between the two kinds of work discussed in Part 1: the reliability functions whose failures are bounded and recoverable, and the safety functions whose failures kill people. 0:00 /0:42 1× The twelve-role sample in [Part 1](https://www.aixenergy.io/the-power-sectors-ai-decade%5Fpart%5Fi/) showed a pattern that follows from that structural difference: reliability roles change on near and medium horizons, safety roles change on medium and long horizons, and the task content inside every role is already shifting even where role-level headcount is steady or growing. What produces that timing? The translation from AI capability to workforce change in a regulated industry is not a single event. It is a multi-stage process, and each stage has its own pace. The stages are analytically cleaner than real life ever is, but the cleanliness is part of their utility: each stage has a distinct set of signals, and knowing which stage a given role is in tells you what to watch for next. ## **Four stages** In the **first stage**, **AI capability matures inside a specific role** through pilot deployments, vendor partnerships, and early operational integrations that move the technology from controlled demonstration to working production tool. Capability in the lab and capability in a production workflow are not the same thing. This is the stage at which the production-ready version is established. The Southern Company substation drone deployment described in Part 1 sits here, and so do most of the ongoing utility trials: vegetation-management computer vision, outage-prediction analytics for distribution utilities, early DERMS deployments supporting virtual-power-plant aggregation. Much of the present-day evidence of AI in US utility operations is at stage one. In the **second stage**, **the matured capability erodes the substantive content of the human task** inside the role. The worker is still titled the same way, paid the same way, and counted the same way in the BLS occupational data. What they actually do shifts: from direct judgment to supervision of automated systems, from first-pass analysis to exception handling, from problem-solving to result-validation. The role-level headcount does not change. The composition of the workday does. This stage can persist for a decade or longer without producing a single layoff, because the institutional architecture supporting the role is still intact and has not been directly contested. [The July 2025 CAISO pilot with OATI](https://www.publicpower.org/periodical/article/oati-launches-generative-ai-platform-pilot-power-grid-operations-with-caiso?ref=aixenergy.io) is an early stage-two deployment: a generative-AI tool built into the outage management system, with CAISO's chief information and technology officer framing it in terms of "improving situational awareness and freeing up time for other important tasks" for operators. The dispatcher is still the dispatcher; the work is shifting around them. Similar operator-assist and planning-AI pilots are active at PJM and SPP. Stage-two dynamics are visible in utility control rooms now. In the **third stage**, **eroded role content begins to weaken the institutional defenses** that sustained the role's headcount. The licensing requirements were written against tasks described in language that no longer maps cleanly to what the licensed person actually does. The union contracts codified staffing ratios against work that has substantively changed. The regulatory staffing minimums were built on assumptions about operator workload that the capability deployment itself is stressing. The [NERC System Operator Certification Program](https://www.nerc.com/pa/Train/SysOpCert?ref=aixenergy.io), which credentials reliability coordinators, transmission operators, and balancing authority operators against specific task competencies, is a candidate stage-three battleground in waiting. The competencies are defined against work that AI-augmented control rooms are visibly reshaping, and the task content that the certification was built to verify is no longer a clean description of what operators actually do day to day. This is the workforce face of the governance-insufficiency problem [Brandon Owens diagnoses as the latency gap](https://www.utilitydive.com/news/ai-power-grid-governance/750725/?ref=aixenergy.io): as operational judgment migrates into software, the institutional frameworks built around the human exercising that judgment lose their purchase on their original grounds. None of these defenses collapses at once. They attenuate. The arguments for them become harder to make on their original grounds, because the original grounds no longer describe the work that is actually being done. In the **fourth stage**, **the role-level headcount change follows**. When an institutional defense becomes untenable on its original grounds, the change that looked impossible at the outset can arrive quickly. Two corners of the utility workforce appear to have already completed, or nearly completed, a stage-four transition. Meter reading has contracted as advanced metering infrastructure has deployed widely, and the meter-reader role today is significantly smaller than it was a decade ago. Utility customer service representatives are further behind but are under the kind of pressure that has produced stage-four contractions in other industries: [a Gartner survey of 321 customer service and support leaders conducted in October 2025](https://www.gartner.com/en/newsroom/press-releases/2025-12-02-gartner-survey-finds-only-20-percent-of-customer-service-leaders-report-ai-driven-headcount-reduction?ref=aixenergy.io) found that 20 percent had already reduced agent staffing due to AI, and the same general mechanism is operating inside utility customer operations, though the pace of compression in utilities specifically has not been independently documented. Neither transition was announced as a stage-four AI event. Both are playing out anyway, on timelines that are fast relative to the long second-and-third-stage periods that preceded them. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/05/four_stage_mechanism.png) ## **Three precedents** Three regulated industries with safety-critical workforces have moved through some version of this arc, and watching where each of them currently sits is the closest available proxy for what the US power sector is moving toward. **Aviation** moved from three-person to two-person commercial cockpits in the 1980s as autopilot systems matured and the flight engineer role was absorbed by automation. That was a stage-four event. Since then, the copilot role has been moving through stages two and three for several decades. Airbus and Dassault, the two European large-commercial-aircraft manufacturers, formally proposed extended minimum-crew and single-pilot concepts to European regulators in 2017\. A formal rulemaking task on reduced-crew operations was opened. [The European Union Aviation Safety Agency study led by the Royal Netherlands Aerospace Center reported in June 2025](https://www.eurocockpit.eu/news/single-pilot-concept-stalls-safety-grounds-now?ref=aixenergy.io) that equivalent safety could not be demonstrated at current technology levels, and the rulemaking was paused pending further technological development. The institutional defense is currently holding. It is holding under explicit pressure from manufacturers and operators with a clear commercial interest in the change. *Institutions in safety-critical industries do not break; they erode, and then they give way.* **Freight rail** sits in a different position on the same arc. The institutional defense in rail has been actively contested in both directions over the past decade. The Federal Railroad Administration issued a final rule in [April 2024 codifying a two-person minimum crew requirement for freight operations](https://www.federalregister.gov/documents/2024/04/09/2024-06625/train-crew-size-safety-requirements?ref=aixenergy.io), responding in part to the East Palestine derailment and to a long campaign by SMART-TD and BLET. The rule includes carve-outs for certain operations and a special-approval process for one-person crew petitions. The Association of American Railroads and several Class I carriers challenged the rule, and the consolidated case is currently before the 11th Circuit Court of Appeals. A 2025 nullification bill was introduced in the House. Rail is a case where the institutional defense was strengthened by regulatory action even as the underlying technological capability has continued to mature. The resulting tension is producing visible legal and political contestation rather than smooth attenuation. Both directions of stage three are active simultaneously. **Maritime** is moving along the same arc on a different international timeline. The International Maritime Organization's [non-mandatory MASS Code for Maritime Autonomous Surface Ships is on track for adoption in May 2026](https://www.rivieramm.com/news-content-hub/news-content-hub/autonomy-must-be-fit-for-purpose-as-mass-code-is-finalised-87840?ref=aixenergy.io), with a mandatory code targeted for 2032\. Norway, China, and Japan have advanced trials of varying autonomy levels for short-route operations. The current state is closer to formal regulatory framework-building than to commercial deployment, but the international institutional architecture is moving toward accommodating reduced and remote crewing on a defined timeline. Maritime is in earlier stage three than aviation, with the institutional path toward stage four more clearly mapped than in either aviation or rail. The lesson from all three is the same. The institutional defenses around safety-critical workforces do not break under direct technological pressure. They erode through long second and third stages of role-content shift, and the eventual stage-four change happens fast relative to the duration of the erosion that preceded it. The timing of the stage-four transition depends less on technology than on the institutional architecture surrounding the role, and the architecture varies country by country, agency by agency, contract by contract. ## **Mapping onto the US power sector** The four-stage pattern maps onto the workforce matrix from Part 1 with a specific structure. Reliability functions are mostly in stages one and two now and will reach stage four on a five-to-fifteen-year horizon, primarily through coordination-layer consolidation: smaller dispatch teams operating higher-capability systems, smaller market-analysis teams covering broader portfolios, customer-service operations restructured around AI handling first-line interactions, regulatory and compliance functions compressed by LLM-driven drafting and evidence compilation. Safety functions are mostly in stage one now and will pass through stages two and three over the same fifteen-year horizon while maintaining role-level headcount. The work inside those roles will visibly change. The role count will not. The stage-four transition for safety functions arrives on a fifteen-to-thirty-year horizon, and its timing depends on generational turnover in the regulators, operator communities, and legal architecture that sustain the current institutional defenses. Task-level effects are visible throughout this progression, on a much shorter timeline than role-level effects. A wind technician whose role-level headcount is growing will still experience significant task-level change in the next three to five years. A nuclear operator whose role-level headcount is steady through the 2040s will spend much of the intervening period supervising systems that were not part of the job description when they were licensed. The worker's experience of AI is a task-level experience long before the occupational data registers anything. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/05/workforce_horizon_matrix.png) That structure is the through-line of the rest of the series. ## **What this means now** The implications of that timeline arrive at different desks at different times. **For utilities.** The most consequential workforce-planning question facing an integrated utility in 2026 is not how to upskill the existing workforce for AI. It is how to plan succession into reliability-function roles whose headcount targets are likely to compress materially over the next decade, even as replacement-cohort hiring continues against the original targets. The utilities that navigate the transition cleanly will be the ones whose workforce plans treat the matrix in Part 1 as the assumption set rather than the current org chart. The parallel implication for training and development investment is that the largest returns will come from task-level retooling inside persistent roles, not from wholesale retraining of workers whose roles are contracting. **For ISOs, RTOs, and state regulators.** Reliability-function staffing minimums and tariff structures that were set against the workload of pre-AI operations will increasingly look out of step with what the workload actually is. The institutional pressure to relax those minimums will arrive before the institutional comfort to do so. The strongest regulatory posture is to anticipate the pressure rather than to respond to it after the gap becomes embarrassing. Performance-based regulation focused on system outcomes will hold up better through the transition than rigid input-based staffing rules will. The harder regulatory question is on the safety side, where the temptation to relax human-in-the-loop requirements will eventually arrive but should be resisted until the institutional architecture for verifying equivalent safety is built, not assumed. The NERC operator certification framework in particular will face pressure to reflect changed task content long before it faces pressure to reduce licensed headcount, and the former is a reasonable accommodation while the latter should not be granted on the early edge of institutional comfort. **For investors.** The capital-allocation implication is that the binding constraint on AI-driven utility productivity is institutional rather than technological. The companies that will capture the productivity gain are the ones whose internal regulatory, union, and operational governance architectures are built to absorb the transition without litigating each piece of it separately. That is a different screen from what most utility-AI investment frameworks currently use, and it is closer to the screen that worked in industries that have already moved through earlier stages of the same arc. The investment thesis on a utility's AI exposure is more a question about its institutional flexibility than about its vendor relationships. **For workers.** The cognitive-AI exposure pattern visible in [the Stanford Digital Economy Lab's November 2025 ADP payroll analysis](https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/?ref=aixenergy.io), which finds a 16 percent relative decline in employment for workers aged 22 to 25 in the most AI-exposed occupations, has not yet shown up in utility occupational data. The roles in the utility workforce most exposed to that pattern are dispatchers, market analysts, customer service representatives, and rate case economists. The roles least exposed at the role level are the ones tagged safety in the matrix, though the task content in even those roles is changing. The recommendation that follows is not a general-purpose one; it depends on which row of the matrix the worker sits in and on whether the change most likely to affect that worker is a role-level contraction or a task-level shift. For the most exposed cognitive roles, the entry-level concentration of the AI displacement signal makes early-career adaptability the key variable. For the safety-function roles, the change is real but slower at the role level and will arrive first inside the work itself. The human-in-the-loop requirement will be defended for at least another decade; the content of the work inside that requirement will look very different by the end of it. ## **What comes next** **Part 3** builds out the full role-by-role matrix across NAICS 221100 and the surrounding institutional infrastructure. Roughly forty roles, with the four-stage mechanism applied systematically across the industry. **Part 4** takes up the capability-creation half of the AI-and-robotics story: the new activities that utilities will be able to perform for the first time, rather than the existing activities compressed by automation. **Part 5** returns to strategy with specific recommendations for each audience, and engages directly with the productivity claims in current consultancy framings of AI in utilities. The numbers in those framings support the argument developed here more than they are usually understood to. The workforce implication of the productivity claims is what the consultancy register declines to name. The translation from AI capability to utility workforce change runs through institutional defenses that hollow out from within rather than break from outside. Task content inside roles shifts first and most visibly. Role-level headcount follows on longer timelines that differ systematically between reliability and safety functions. The decade ahead looks slow until it does not. ### Command of the Interconnect: The Hidden Infrastructure War Beneath Artificial Intelligence URL: https://www.aixenergy.io/indium/ Last updated: 2026-07-29T19:42:44.000Z by [Macdonald Amoah](https://www.linkedin.com/in/macdonald-amoah-089bb7100/?ref=aixenergy.io), [Jahara Matisek](https://www.linkedin.com/in/jahara-franky-matisek-phd-9294a9a/?ref=aixenergy.io), and [Morgan Bazilian](https://www.linkedin.com/in/morganbazilian/?ref=aixenergy.io) The future of artificial intelligence may depend less on processors than on light itself. Inside hyperscale AI data centers, computation is no longer constrained primarily by transistor density or even electricity supply. The emerging bottleneck is the movement of information between processors at the speed required to hold giant AI systems together as a coherent machine. That shift has transformed a little-known semiconductor material into one of the most strategically important substances in the global technology economy: [indium phosphide](https://www.intelmarketresearch.com/ai-computing-power-indium-phosphide-market-22228?ref=aixenergy.io) (InP). 0:00 /1:01 1× Few policymakers understood its value until Beijing placed InP substrates under export licensing controls in 2025\. Yet, inside hyperscale data centers, InP is the critical photonic link, enabling the lasers that push data through optical interconnects, fusing tens of thousands of individual processors into a single computing fabric. Since 1993, America has been 100 percent [import reliant](https://pubs.usgs.gov/myb/vol1/2023/myb1-2023-indium.pdf?ref=aixenergy.io) for the refined indium metal required to produce these substrates. China [produces](https://www.reuters.com/business/energy/indium-hits-highest-prices-decade-chinese-speculation-supply-risks-2026-02-09/?ref=aixenergy.io) about 70 percent of global refined indium production, and as of late 2024, it directly supplied a quarter of U.S. imports. The remainder of the U.S. supply chain is similarly [concentrated](https://www.questmetals.com/blog/indium-prices-surge-to-decade-highs?ref=aixenergy.io), relying on the Republic of Korea (29 percent), Japan (18 percent), and Canada (14 percent). For the companies building AI and fifth-generation (5G) communications systems, the vulnerability gets even worse at the wafer stage. Two Japanese firms and one U.S.-headquartered, China-based AXT collectively [control](https://finance.biggo.com/news/xsiIxJ0Bh5an-7GhIww3?ref=aixenergy.io) 90 percent of InP wafers globally. ## **AI is Becoming a Materials Science Problem** Scaling AI requires transitioning from electrical signals to light; generating that light requires a continuous supply of InP. Securing this critical material is essential for America and its allies to maintain their economic strength and military power. Artificial intelligence is often described as a software revolution. In practice, scaling frontier AI increasingly resembles an industrial mobilization problem. Every major expansion in AI capability now carries a parallel expansion in physical infrastructure requirements: electricity generation, cooling systems, high-voltage transformers, transmission capacity, fiber optics, semiconductor fabrication, and critical mineral supply chains. As model sizes and GPU clusters continue growing, the constraints increasingly shift away from code and toward the physical systems capable of moving energy and information at extreme scale. ## **Optical Fiber** Training a large language model requires continuous, high-volume data exchange across thousands of processors operating in parallel. As GPU clusters expand from the thousands into the hundreds of thousands, the interconnect fabric becomes the primary limiting factor. Industry projections [suggest](https://blog.semtech.com/key-technology-trends-shaping-data-center-and-telecom-infrastructure-hardware-innovation-in-2025?ref=aixenergy.io) GPU clusters could reach one million units by 2026\. At this massive scale, the physical properties of copper wire become actively hostile to performance. At data rates of 224 gigabits per second per lane (the industry standard in late 2024), the effective reach of a passive copper cable [collapses](https://markets.financialcontent.com/wral/article/tokenring-2026-1-1-the-speed-of-light-silicon-photonics-shatters-the-ai-interconnect-bottleneck?ref=aixenergy.io) to less than one meter. Optical fiber bypasses these physical limits entirely. It transmits photons through glass at roughly 200,000 kilometers per second with losses [measured](https://blog.semtech.com/key-technology-trends-shaping-data-center-and-telecom-infrastructure-hardware-innovation-in-2025?ref=aixenergy.io) in fractions of a decibel per kilometer. A single rack of advanced AI hardware can now [draw](https://blog.semtech.com/key-technology-trends-shaping-data-center-and-telecom-infrastructure-hardware-innovation-in-2025?ref=aixenergy.io) up to 125 kilowatts, which is more than an average American household consumes in a month. Within that massive power budget, the transceivers converting electrical signals to optical pulses represent a disproportionate share of energy consumption. NVIDIA chief executive Jensen Huang quantified this concern at GTC 2025, noting that with six pluggable transceivers per GPU each consuming 30 watts, [scaling](https://ayarlabs.com/blog/co-packaged-optics-step-into-the-spotlight/?ref=aixenergy.io) to a million GPUs would devote enormous capacity solely to signal conversion. Transitioning to co-packaged optics (integrating photonic and electronic functions into a single package) [reduces](https://newsletter.semianalysis.com/p/co-packaged-optics-cpo-book-scaling?ref=aixenergy.io) this energy burden by approximately 73 percent. This cuts power usage from around 17 watts per 800-gigabit link down to under 5 watts. The commercial scale of this hardware transition is massive. The data center optical component market [exceeded](https://introl.com/blog/fiber-optics-data-center-state-of-art-optical-interconnect-2025?ref=aixenergy.io) $16 billion in 2025, growing by more than 60 percent year over year. Shipments of 800-gigabit optical transceiver modules doubled over the same period. The silicon photonics market serves as the integrated platform underlying much of this infrastructure. This specific sector stood at $3.1 billion in 2025 and is forecast to reach $10.4 billion by 2030, [advancing](https://introl.com/blog/fiber-optics-data-center-state-of-art-optical-interconnect-2025?ref=aixenergy.io) at a 27 percent compound annual growth rate. Today, over 80 percent of data center links in hyperscale environments now [rely](https://www.datamintelligence.com/research-report/optical-interconnect-in-ai-data-centers-market?ref=aixenergy.io) on optical interconnects. At sufficient scale, interconnect performance stops being a secondary engineering issue and becomes a hard computational limit. A modern AI training cluster cannot simply “slow down” when optical transmission bottlenecks emerge. GPUs begin sitting idle while waiting for data synchronization across the cluster. Training efficiency collapses even as electricity consumption remains elevated. In practical terms, photonic constraints can strand billions of dollars of AI infrastructure capacity. ## **Why Indium Phosphide Matters** Silicon has dominated semiconductor manufacturing for six decades because it is abundant, highly scalable, and compatible with mature fabrication infrastructure. However, silicon has a fundamental physical limitation since it [cannot](https://www.viksnewsletter.com/p/why-cpo-uses-external-lasers?ref=aixenergy.io) produce laser light. It is an indirect-bandgap material, meaning that when an electron transitions between energy states, it cannot efficiently emit a photon. Indium phosphide (InP) solves this exact problem. As a direct-bandgap compound semiconductor, electron-hole recombination within InP releases energy as photons with high probability. InP is the only major integrated-photonics [platform](https://www.photondelta.com/news/what-is-indium-phosphide-and-what-can-it-do-more-than-other-pic-platforms/?ref=aixenergy.io) that can both conduct and produce light. This makes it uniquely capable of generating the on-chip laser sources required by AI optical systems. Its emission characteristics are also perfectly matched to the telecommunications C-band [centered](https://www.sciencedirect.com/topics/engineering/silica-optical-fiber?ref=aixenergy.io) at 1,550 nanometers. This is the precise wavelength [window](https://andersstorm.substack.com/p/why-indium-phosphide-inp-is-becoming) at which single-mode optical fiber [exhibits](https://resources.l-p.com/1550nm-optical-transceiver-transmission-distances?ref=aixenergy.io) minimum attenuation of 0.2 decibels per kilometer. No other commercially mature material combines these properties at this wavelength with comparable performance. In earlier industrial eras, strategic power flowed through oil fields, pipelines, shipping lanes, and telecommunications cables. In the AI era, geopolitical leverage may increasingly depend on command of the interconnect: the materials and systems that allow information to move at light speed across giant computational fabrics. Consequently, the silicon photonics industry relies completely on InP to generate light. Silicon is used to fabricate the waveguides, modulators, and passive components in a modern optical transceiver. Manufacturers must then bond InP laser dies directly onto silicon substrates or use InP-based external cavity lasers to feed silicon chips with continuous-wave light. Leading research institutions (including UC Santa Barbara, TU Eindhoven, and IMEC) continue to advance heterogeneous integration by printing InP gain material onto silicon waveguides. Regardless of the integration method, the InP material itself remains indispensable. As Professor Martijn Heck of TU Eindhoven recently noted, there are [no substitutes](https://www.viksnewsletter.com/p/interview-indium-phosphide-lasers?ref=aixenergy.io) for III-V semiconductors like InP in their specific applications. ## **Weaponizing the Optical Supply Chain** The strategic problem with indium actually occurs before the mine. Indium is almost never targeted for primary mining. It occurs in trace concentrations (one to 100 parts per million) within zinc-sulfide ore deposits. Consequently, indium is [recovered](https://pubs.usgs.gov/periodicals/mcs2024/mcs2024-indium.pdf?ref=aixenergy.io) as a byproduct of zinc smelting. This means a highly inelastic supply curve. Spikes in indium demand do not equal more indium mines. Production only surges if global zinc economics justify expanding zinc smelting capacity or improving recovery rates at existing plants. Because Beijing [controls](https://www.bloomberg.com/news/articles/2025-11-19/chinese-smelters-ramp-up-zinc-exports-on-global-supply-squeeze?ref=aixenergy.io) half of global zinc smelting capacity, it commands the global indium supply, leaving Western markets exposed to a major electronics vulnerability. Beijing has steadily [weaponized](https://mwi.westpoint.edu/minerals-magnets-and-military-capability-chinas-rare-earth-weaponization-should-be-a-wake-up-call/?ref=aixenergy.io) these dependencies through a calibrated escalation of export controls. The [mineral control campaign](https://www.sipri.org/commentary/topical-backgrounder/2026/chinas-export-control-framework-domestic-developments-and-international-positioning?ref=aixenergy.io) began in mid-2023 with licensing requirements on gallium and germanium, followed closely by graphite and antimony. In December 2024, China escalated significantly by [banning](https://www.questmetals.com/blog/u-s-faces-ripple-effects-from-china-s-germanium-and-gallium-export-restrictions?ref=aixenergy.io) gallium, germanium, and antimony exports to the United States alongside a blanket prohibition on all dual-use item exports to U.S. military end users. In early 2025, China [imposed](https://www.iea.org/policies/26795-decision-to-implement-export-controls-on-tungsten-tellurium-bismuth-molybdenum-and-indium-related-items?ref=aixenergy.io) export controls over tungsten, tellurium, bismuth, molybdenum and indium. This caused indium prices to jump from $2,600 per kilogram to $3,000 in less than a month, and over time it has led to indium prices [peaking](https://tradingeconomics.com/commodity/indium?ref=aixenergy.io) around $4,700 per kilogram in March 2026. Price shocks are damaging, but Chinese licensing regimes are even more dangerous. They operate as a mandatory intelligence-gathering [mechanism](https://www.dhs.gov/sites/default/files/publications/20%5F1222%5Fdata-security-business-advisory.pdf?ref=aixenergy.io). To obtain export approval, Western applicants must submit exhaustive end-use information to Chinese authorities, serving as a dual-use intelligence [purpose](https://orfamerica.org/newresearch/chinas-critical-mineral-export-controls?ref=aixenergy.io). Furthermore, the early 2025 directive did not just target raw indium metal. It explicitly restricted indium phosphide substrates and the highly specialized precursor chemicals required to manufacture them. The distinction is critical. Producing a finished substrate requires advanced crystal growth infrastructure and decades of manufacturing expertise. It is the finished substrate, rather than the raw ore, that feeds directly into the optical transceivers powering AI data centers. Ultimately, this strategy injects severe uncertainty directly into the investment and procurement decisions underpinning the American tech sector. Beijing is doing more than driving up commodity costs. It is actively holding the hardware layer of Western AI development at risk. Strategic leverage does not require a full embargo. It only requires enough uncertainty to distort procurement decisions, delay infrastructure investments, and force competitors into defensive supply-chain behavior. In that sense, export licensing regimes function not only as trade restrictions, but as instruments of industrial pressure. ## **Breaking the Photonic Chokepoint over AI** The convergence between AI infrastructure and energy infrastructure is becoming impossible to separate. Optical interconnect efficiency now directly affects data center electricity demand, cooling loads, rack density, and transmission planning. In this sense, indium phosphide is not simply a semiconductor issue. It is an emerging energy systems issue. The AI competition has become a [materials](https://www.sfa-oxford.com/knowledge-and-insights/critical-minerals-in-low-carbon-and-future-technologies/critical-minerals-in-artificial-intelligence/?ref=aixenergy.io) race, stretching from [power transformers](https://www.whitehouse.gov/presidential-actions/2026/04/presidential-determination-pursuant-to-section-303-of-the-defense-production-act-of-1950-as-amended-on-grid-infrastructure-equipment-and-supply-chain-capacity/?ref=aixenergy.io) to obscure rare earths like [yttrium](https://nationalinterest.org/blog/techland/the-hidden-supply-chain-risk-in-ai-us-reliance-on-chinas-yttrium?ref=aixenergy.io). During the twentieth century, great powers competed for command of oil supply chains because fuel determined military mobility and industrial output. In the twenty-first century, command of the interconnect may play a comparable role for computational power. Nations that cannot secure the physical infrastructure required to move information at scale may find themselves structurally disadvantaged in AI development regardless of software sophistication. The global market for indium phosphide wafers is relatively small, [valued](https://www.mordorintelligence.com/industry-reports/indium-phosphide-wafer-market?ref=aixenergy.io) at just over $200 million in 2026\. Yet it operates as a massive strategic chokepoint. A tiny physical input governs the performance of trillion-dollar cloud computing and AI systems, and a [disruption](https://yianisz.substack.com/p/indium-phosphide-inp-the-quiet-bottleneck) to its supply can damage economies and degrade military operations. Expanding capacity means overcoming decades of highly rigid zinc smelting economics and building up specialized crystal growth expertise that does not yet exist at scale outside of East Asia. Silicon and other alternative materials cannot physically replicate InP's light-emitting properties at the required telecommunications wavelengths. The strategic importance of photonic materials is likely to increase rather than diminish. AI architectures are already moving toward co-packaged optics, photonic switching, and increasingly dense optical interconnect fabrics designed to overcome the physical limitations of electrical transmission. Future generations of AI infrastructure may become even more dependent on advanced photonic materials than current systems. Three lines of effort are essential. First, the full value chain must be mapped with operational precision. The Defense Logistics Agency is already [investigating](https://www.bhfs.com/insight/critical-update-usgs-expands-mineral-list/?ref=aixenergy.io) how best to stockpile indium, and the newly established [Economic Defense Unit](https://defensescoop.com/2026/04/10/dod-economic-defense-unit-edu-steve-feinberg-pentagon/?ref=aixenergy.io) should be tasked with completing the picture. Existing USGS mineral commodity surveys track raw material flows at a national level, but they do not trace the specific path from allied zinc concentrates through indium refining, precursor chemical synthesis, wafer crystal growth, and final optical transceiver assembly. An operational map must identify chokepoints at each stage: which smelters recover indium as a byproduct and which do not, where precursor trimethylindium is synthesized, which crystal growth facilities can produce substrate-quality boules, and how many qualified transceiver manufacturers source from non-Chinese wafer suppliers. Without this granularity, stockpiling raw indium addresses only one link in a multi-stage chain. Second, allied wafer and refining networks must be forged in lockstep. America cannot rebuild this entire photon-for-AI ecosystem domestically in the near term. Washington needs to co-invest with Japan, South Korea, and Canada to support non-Chinese refining, recycling, and wafer production capacity. These are niche material markets that private capital routinely ignores because individual investments appear too small to justify diligence costs. Government-directed subsidies and co-financing arrangements, modeled on the joint ventures that expanded allied semiconductor packaging capacity under the CHIPS Act, can close this gap. Coordination must extend beyond funding to include joint qualification standards, so that allied producers can be certified as drop-in alternatives without years of redundant testing by each end-user. Finally, there need to be bankable demand signals. Production capacity will not materialize if end-users refuse to qualify alternative suppliers, and suppliers will not invest in expensive crystal growth infrastructure against the risk that a single customer defects back to cheaper Chinese sources. The Pentagon and the broader federal government must use their procurement power to mandate secure InP supply chains for all future advanced communications and defense AI systems. Long-term purchase agreements that guarantee demand for allied producers would convert what is currently a geopolitical aspiration into a commercially investable proposition. No amount of software innovation or venture capital can substitute for the physical ability to generate photons at the wavelengths required to hold large AI systems together. The next great technological competition may not be decided solely by who builds the smartest algorithms or the largest data centers. It may ultimately be decided by who controls the obscure materials, optical systems, energy infrastructure, and industrial supply chains that allow artificial intelligence to move at the speed of light. ## Sign up for AIxEnergy AIxEnergy analyzes how artificial intelligence is reshaping electricity demand, grid operations, and energy infrastructure, with a focus on system risk, policy, and institutional decision-making. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### AEO 2026 in Review: A Structural Shift in U.S. Energy System Dynamics URL: https://www.aixenergy.io/aeo2026/ Last updated: 2026-07-10T17:29:58.000Z The Annual Energy Outlook has always carried authority. It does not predict; it structures. It does not claim certainty; it defines the bounds of it. For years, that distinction blurred in practice as the “Reference Case” came to stand in for a future that markets and policymakers could point to. Article Summary 0:00 /81.795193 1× AEO2026 removes that comfort. It does not replace it with a new forecast. It removes the idea of a center altogether and replaces it with a framework for testing how the system behaves when its assumptions move. In doing so, it reveals a deeper shift: the U.S. energy system is no longer organized around total energy demand or fuel substitution alone. It is reorganizing around electricity—its growth, its timing, and the forces now driving it. The comparison with AEO2025 makes that shift visible. The numbers evolve, but the structure changes more. What follows is a technical reassessment grounded in the narratives, assumptions, and reported data, with attention to where the outlook has moved in meaning as well as magnitude. ## The Baseline: From “Reference Case” to Experimental Control The most explicit methodological change between the two outlooks is definitional, and it sits at the center of how every result should be read. In AEO2025, the Reference Case is described as a policy-neutral representation of the system under laws and regulations as of December 2024, intended to reflect business-as-usual trends.¹ In AEO2026, the same construct is renamed the Counterfactual Baseline and is accompanied by a clear clarification that it should not be regarded as the most likely of the cases, but instead serves as an experimental control.² Nothing in the underlying machinery has changed. The model still uses NEMS, still holds policy constant at a cutoff date, and still evaluates side cases around that anchor. What has changed is the interpretive center. The 2025 framing allowed the Reference Case to function as an anchor that markets and analysts treated as a central trajectory, even if that was not its stated purpose. The 2026 framing removes that anchor and replaces it with a disciplined statement about function. This distinction reframes the entire output of the model. The results are no longer implicitly presented as a path forward, but as a structured environment for testing how the system responds when assumptions move. The outlook becomes less a projection and more a sensitivity surface, and the burden shifts to the reader to understand relationships rather than to select outcomes. ## Macroeconomic Assumptions: Narrow Drift Within a Constrained Envelope The macroeconomic assumptions across the two outlooks remain tightly bounded, but the narrative treatment in 2026 places greater emphasis on structural constraints. In AEO2025, GDP growth in the Reference Case averages 1.8 percent annually through 2050.³ In AEO2026, the Counterfactual Baseline assumes slightly lower growth at 1.7 percent annually, with side cases spanning a range from 1.2 percent to 2.2 percent.⁴ The more important shift lies in the explanation. The 2026 narrative explicitly ties slower long-term growth to demographic constraints, including population, labor force, and productivity limits, rather than treating it primarily as an economic uncertainty. Population projections reinforce this bounded outlook, ranging from 352 million to 382 million by 2050.⁵ This tighter macro envelope places greater analytical weight on technology and structural change within the energy system itself, because macroeconomic variability alone is not sufficient to drive large differences in outcomes. ## Total Energy Consumption: Stability as a Structural Feature Both outlooks converge on a consistent conclusion that total U.S. energy consumption remains effectively flat through 2050 across most scenarios. In AEO2026, total consumption trends range from declines of roughly 0.6 percent annually to near-zero growth, with only modest expansion in high-growth cases.⁶ This result is not an anomaly but an extension of a long-running pattern. Between 2000 and 2025, total energy consumption declined slightly, even as GDP expanded at more than two percent per year, with the system operating within a narrow band of approximately 88 to 99 quadrillion BTUs.⁷ AEO2025 presents this trend as context, while AEO2026 embeds it as a defining structural condition of the system. The implication is that efficiency gains and structural economic changes continue to offset growth across sectors. The United States is no longer an energy-growth economy in the traditional sense, and that constraint persists regardless of variation in policy or technology assumptions. ## Electricity Demand: The Break from Historical Behavior The most consequential divergence between AEO2025 and AEO2026 emerges in electricity demand, where the system departs from its prior period of stagnation. After more than a decade of flat demand, electricity consumption has increased by 2.1 percent annually over the past five years and is projected to grow between 0.9 percent and 1.6 percent per year through 2050.⁸ This renewed growth is not driven by traditional factors such as population or industrial expansion alone. AEO2026 isolates a new driver that is explicitly embedded in the model structure: data centers and AI-driven computation. The Commercial Demand Module separates data center server energy use from broader commercial consumption, allowing it to be tracked and scaled independently. The model assumes rapid expansion of installed AI server capacity, with exponential growth in the high-demand case. The resulting magnitude is substantial. Data center electricity consumption reaches 818 billion kilowatt-hours by 2050 in the High Electricity Demand case, representing more than a sixteenfold increase from 2020 levels.⁹ Even in the baseline case, the contribution is material, and data centers combined with electric vehicles account for 50 to 80 percent of incremental electricity demand growth while representing only 10 to 25 percent of total demand.¹⁰ This asymmetry indicates that growth is no longer diffuse across the economy, but concentrated in specific technologies that operate with distinct characteristics. ## Load Shape: A New Determinant of System Cost AEO2026 elevates the importance of load shape, moving beyond total demand to consider how electricity is consumed over time. Two distinct demand profiles emerge from the analysis, each with different system implications. Electric vehicles introduce variability by concentrating charging in specific hours, typically in the evening and overnight, which increases peak demand and drives investment in transmission and distribution infrastructure. Data centers introduce continuity by operating with a relatively constant load across all hours, raising the baseline level of demand and increasing the need for sustained generation capacity.¹¹ These differences translate directly into cost structures. Peak-driven demand requires infrastructure sized to meet short-duration maximum loads, while continuous demand requires capacity that operates at high utilization levels. In scenarios with lower EV penetration, distribution costs increase as peak demand remains high relative to total consumption. In scenarios with high data center growth, generation costs rise as capacity must expand to meet continuous load requirements. AEO2025 does not emphasize this distinction to the same degree. In AEO2026, load shape becomes a primary analytical variable, reflecting a shift in how system planning is conceptualized. ## Electricity Supply: Scaling a Continuous System The demand shifts described above are reflected in supply-side expansion. Across all cases in AEO2026, total electricity generation increases by 25 to 50 percent by 2050, while installed capacity expands by 50 to 90 percent.¹² These increases are required not only to meet higher total demand but also to accommodate the structural characteristics of that demand. The fuel mix evolves in a direction consistent with AEO2025 but with greater clarity in magnitude. Natural gas remains the largest single contributor at roughly 40 percent of generation in the baseline case, while wind and solar each rise to approximately 20 percent shares. Coal declines from about 16 percent in 2025 to less than one percent under policy-constrained scenarios or around five percent without those constraints. Nuclear generation remains relatively stable in absolute terms but declines in share as total generation increases.¹³ The key difference between the outlooks lies not in direction but in scale and context. Capacity expansion is increasingly driven by continuous, technology-specific demand rather than by broad-based growth across sectors. ## Transportation: Efficiency and Policy Sensitivity While electricity demand rises, transportation energy consumption continues to decline, reinforcing the decoupling between total energy use and economic activity. AEO2026 projects transportation energy use falling from 27 quadrillion BTUs in 2025 to between 21 and 25 quads by 2050, depending on policy assumptions.¹⁴ Electric vehicle adoption plays a central role in this decline. Under current policy assumptions, EVs account for 40 to 46 percent of the light-duty fleet by 2050, with zero-emission freight trucks reaching 21 to 24 percent. In the absence of those policies, EV penetration falls significantly, with light-duty share near 18 percent and freight electrification near five percent.¹⁵ These differences propagate through the system. Reduced electrification leads to increased liquid fuel demand, with potential increases of up to three million barrels per day by 2050, affecting refining activity, trade balances, and LNG exports.¹⁶ AEO2026 traces these interdependencies more explicitly than AEO2025, highlighting the system-wide implications of sector-specific changes. ## Model Evolution: From Expansion to Resolution The evolution of NEMS between the two outlooks supports these analytical shifts. AEO2025 introduced new modules for hydrogen markets, carbon capture, and hydrocarbon supply, expanding the model’s scope.¹⁷ AEO2026 focuses on improving resolution and performance, reducing runtime by more than half, decreasing disk usage, enhancing LNG modeling, and refining hourly aggregation to better capture variability in renewable generation and load.¹⁸ These improvements enable the model to represent temporal dynamics more accurately, which is necessary to analyze the effects of load shape and continuous demand introduced by data centers. ## Conceptual Synthesis: Toward a Cognitive Grid Taken together, these changes point beyond incremental system evolution toward a different organizing logic. The grid is no longer only a delivery system for energy; it is becoming the substrate for continuous, information-driven activity. The emergence of large, steady, computation-driven loads introduces a new form of system stress—one that is not defined by peaks alone but by persistent, high-baseline demand. This is where the concept of a *Cognitive Grid* becomes useful. In this framing, electricity infrastructure and computational systems are not parallel developments but integrated layers of a single architecture. The grid supplies power to computation, while computation increasingly informs how the grid is planned, operated, and optimized. The result is a feedback loop in which energy and information systems co-evolve. AEO2026 does not name this shift, but it provides the empirical basis for it. By isolating data center demand, quantifying its growth, and linking it to system costs through load shape, the outlook reveals the contours of a system that is no longer defined solely by energy flows, but by the interaction between energy and computation. ## Final Observation The most important change between AEO2025 and AEO2026 is not a single projection or a single case. It is the emergence of a new organizing principle in which the structure of electricity demand, including its magnitude, timing, and persistence, determines system outcomes as much as total energy consumption itself. This principle was present in earlier outlooks but remained implicit. In AEO2026, it becomes explicit, quantified, and embedded in the model. The system is no longer evolving along the same path, and the difference is visible not in where the model points, but in how it is constructed. ## Notes 1. U.S. Energy Information Administration, *Annual Energy Outlook 2025*, narrative section. 2. U.S. Energy Information Administration, *Annual Energy Outlook 2026*, narrative section. 3. AEO2025 narrative, macroeconomic assumptions. 4. AEO2026 narrative, macroeconomic assumptions. 5. AEO2026 narrative, population projections. 6. AEO2026 narrative, total energy consumption trends. 7. AEO2026 narrative, historical energy consumption context. 8. AEO2026 narrative, electricity demand growth. 9. AEO2026 narrative, data center electricity consumption. 10. AEO2026 narrative, share of demand growth. 11. AEO2026 narrative, load shape and hourly demand discussion. 12. AEO2026 narrative, generation and capacity growth. 13. AEO2026 narrative, electricity fuel mix. 14. AEO2026 narrative, transportation energy consumption. 15. AEO2026 narrative, EV adoption rates. 16. AEO2026 narrative, oil demand sensitivity. 17. AEO2025 narrative, NEMS enhancements. 18. AEO2026 narrative, NEMS performance and resolution improvements. ### The Power Sector's AI Decade Part I URL: https://www.aixenergy.io/the-power-sectors-ai-decade_part_i/ Last updated: 2026-07-29T19:41:52.000Z In June 2023, [Southern Company received one of the first conditions-based FAA waivers](https://dronelife.com/2023/06/26/southern-company-gets-bvlos-waiver-for-fully-remote-drone-operations-with-skydio-dock/?ref=aixenergy.io) authorizing remote-operated, beyond-visual-line-of-sight autonomous drone operations at critical infrastructure sites across its system, from Georgia to California. The hardware is the Skydio X2 paired with the Skydio Dock, a weatherproof drone garage that launches its aircraft, monitors the airspace, and recovers it without a pilot on site. The operational use case is the one [Skydio describes plainly](https://www.skydio.com/blog/skydio-dock-and-remote-ops?ref=aixenergy.io): after a weather event or equipment alarm, the drone flies a pre-planned route over a substation to inspect switches, transformers, and busbars, and streams the result back to a technician in an office somewhere else. 0:00 /2:33 1× Two things happen in that single deployment. A task that a substation technician or a contractor pilot would have performed in person, often hours after the alarm, is now performed remotely in minutes. And a capability that did not exist at all in most substation operations is now routine: thermal scans that previously required a scheduled outage can be flown on demand, on a schedule, or automatically in response to a sensor trip. The data is continuous where it was episodic. The inspection is remote where it was physical. The technician has become a supervisor of a system, not a visitor to an asset. Many of the AI and robotics deployments visible in US utility operations in 2026 share this two-part structure. Some part of a human task moves to a machine. Some new capability becomes possible that no human was performing before. The interesting question is not whether the pattern will scale across the industry. It is how long it takes to show up as measurable workforce change, where it shows up first, and what utilities, regulators, investors, and workers should do about it now. Three distinctions and a twelve-role sample are a workable place to start. The [Cognitive Grid](https://www.aixenergy.io/cognitive-grid-articles/) work explored on [AIxEnergy](https://www.aixenergy.io/) offers the nearest analytical neighbor to the question this series takes up. [Brandon Owens](https://www.aixenergy.io/author/brandon/)'s argument is that operational judgment in critical infrastructure is quietly migrating into automated systems faster than the governance frameworks around that authority can adapt, a phenomenon he calls the latency gap between machine-speed execution and human-speed oversight. His [EthosGrid Open Standard](https://www.aixenergy.io/content/files/2026/01/Ethos%5FGrid%5FOpen%5FStandard.pdf) proposes a constitutional response, separating machine proposal from human authorization so that capability does not silently become permission. This series runs the same institutional diagnosis through the workforce rather than through governance: when operational judgment migrates into software, the humans whose roles were defined around that judgment are the other side of the erosion. ## **Three distinctions to start with** Most attempts to read the workforce implications of AI in utilities go wrong on one of three distinctions. All three deserve to be put on the table early, in plain language, before any of the analysis here is asked to do work. The first distinction is between vendor capability and utility adoption. What an AI system or a robotic platform can do in a controlled demonstration is bounded by engineering: model performance, sensor precision, data availability, hardware reliability. What a regulated utility actually deploys in production is bounded by something else entirely: data architecture, operational trust, liability allocation, licensing, union contracts, regulatory staffing minimums, insurance coverage, and capital stock turnover. Vendor demonstrations move on the first clock. Utility production deployments move on the second. The two clocks tick at very different rates. The technical shorthand used here, and throughout the series, is that vendor capability sets the **capability horizon** and utility deployment sets the **adoption horizon**. Capability horizons for the systems considered here are short. Adoption horizons are longer, more variable, and more dependent on institutional context than on technological readiness. The Censys Technologies BVLOS demonstration in February 2026 makes the gap concrete. [A 79-mile dual-leg drone mission flown between Daytona Beach and Mims, Florida](https://censystech.com/press-release-censys-technologies-to-host-live-public-demonstration-of-long-range-bvlos-drone-mission-across-critical-infrastructure/?ref=aixenergy.io), launched from an automated dock and transiting Class C controlled airspace, was completed with a fixed-wing VTOL aircraft. It was not flown under a special-purpose research authority. It was flown under existing FAA Part 107 waivers, of which Censys [holds more than 85](https://censystech.com/about/?ref=aixenergy.io). The capability to inspect transmission and gas-pipeline corridors at scale is in hand, and has been demonstrated commercially over populated areas. What is not in hand is routine legal authority to operate at corridor scale without applying for a per-mission waiver each time. [The FAA's Part 108 Notice of Proposed Rulemaking](https://www.federalregister.gov/documents/2025/08/07/2025-14992/normalizing-unmanned-aircraft-systems-beyond-visual-line-of-sight-operations?ref=aixenergy.io), published August 7, 2025 and now in final-rule drafting, is the proposed unlock. Until it lands, the binding constraint on corridor-scale autonomous line patrol is rulemaking throughput, not vendor capability. The capability and adoption horizons in this case are months apart in the lab and several years apart in production. That gap is not exotic. It is typical. The second distinction is between two kinds of work inside a utility, which run on two very different institutional clocks. The first kind is **reliability work**: dispatch, load forecasting, market analysis, customer service, billing, predictive maintenance, meter reading, most regulatory compliance analysis. Failure in reliability work produces consequences that are bounded and recoverable: imbalance charges, customer-minutes-out, modest SAIDI or SAIFI degradation, complaints to a regulator, the occasional rate case finding. The harms are economic and serviceable; insurers price them, regulators monitor them, and utility boards are comfortable authorizing AI deployment because the downside is circumscribed. The second kind is **safety work**: nuclear reactor operations, protection relay coordination, energized switching, live-line work, gas-pipeline pressure management, dam operations. Failure in safety work produces immediate physical harm, sometimes catastrophic. The human-in-the-loop requirement in safety work is not a managerial preference. It is attached to NRC, FERC, and state commission licensing regimes. It is reinforced by criminal and civil liability frameworks built around specific past failures, including Three Mile Island, the 2003 Northeast blackout, the San Bruno gas explosion, and the cascade of utility-caused wildfires across the western United States in the past decade. It is codified in union contracts that specify staffing patterns around high-hazard work. And it is shaped by the institutional memory of the regulators, operator communities, and union leaders who have lived through the events that the licensing regime was built to prevent. These two kinds of work coexist inside a single utility, often inside a single operations center. They are not separated by job title in any clean way. But they run on very different clocks. Regulatory and institutional comfort with AI deployment in reliability work builds in years. Regulatory and institutional comfort with AI deployment in safety work builds in decades. That difference is not a matter of technological maturity; the same machine learning techniques that have already moved into reliability work could in principle inform safety work. It is a matter of institutional architecture, and most of the heterogeneity in how AI metabolizes through the utility workforce can be traced to it. The third distinction is between AI's effect on a task and its effect on a role. A task is a specific activity performed within a job: climbing a transmission tower to torque a bolt, writing the first draft of a rate case filing, reviewing a thermal image for equipment defects, making a dispatch adjustment during a frequency excursion. A role is a bundle of tasks that a single worker performs, usually under a single job title and under a specific set of institutional arrangements around licensing, pay, and supervision. AI can compress tasks without eliminating roles, and it usually does. A wind technician still climbs the tower to torque the bolt; what AI changes is the inspection flight that preceded the climb, the work-order generation that followed it, and the condition-assessment judgment that triaged it in the first place. A lineman still makes the physical connection on an energized conductor; what AI changes is the patrol that identified the repair need. The physical work in both cases is robust. The cognitive scaffolding around it is what erodes first. The task-versus-role distinction matters because AI effects on task content can run far ahead of AI effects on role headcount, especially in fields with significant physical-dexterity requirements or strong institutional defenses around the role itself. When the two are conflated, the analysis either overstates near-term displacement (by assuming that task compression means role elimination) or understates it (by assuming that a persistent role means the work inside it is unchanged). Neither is right. The matrix below, and the mechanism developed in Part 2 of this series, both depend on keeping task and role effects analytically separate. ## **A first look at the workforce** The three distinctions above are most useful when they are mapped onto specific jobs. Below is a sampling of twelve roles drawn from across the US electricity industry. Each row carries the role title, an approximate national headcount drawn from BLS Occupational Employment and Wage Statistics for [NAICS 221100, Electric Power Generation, Transmission and Distribution, May 2024](https://www.bls.gov/oes/current/naics4%5F221100.htm?ref=aixenergy.io), a tag identifying whether the role's institutional clock is reliability or safety, the dominant AI system type expected to drive change, an arrow estimating the medium-term direction of role-level headcount, and the time horizon over which the principal change becomes visible. Horizons used here and throughout the series are: **near-term**, three to five years; **medium-term**, five to fifteen years; **long-term**, fifteen to thirty years. Headcount arrows are: ↑ growth, ↔ steady, ↓ contraction, ↓↓ sharp contraction. The arrows describe role-level employment direction. Separately, task content inside every row in this matrix is already shifting, and in some rows shifts substantially even as the role-level arrow points up or stays flat. The full role-by-role matrix, covering roughly forty positions across generation, transmission, distribution, markets, and the regulatory infrastructure, will be developed in Part 3. | **Role** | **Approx US employment** | **Reliability or safety** | **Dominant AI system** | **Role-level headcount** | **Horizon** | | --------------------------------------- | --------------------------------- | ------------------------- | -------------------------------------------------------------- | ------------------------ | ----------- | | **Power dispatcher / system operator** | 5,800 | Reliability | Reinforcement learning, time-series ML | ↓ | Medium | | **Load forecaster / market analyst** | (split across NAICS 523, 221100) | Reliability | Time-series ML, LLMs | ↓ | Medium | | **Customer service representative** | 17,600 | Reliability | Large language models | ↓ | Near-medium | | **Meter reader** | 3,600 | Reliability | AMI plus computer vision | ↓↓ | Near | | **Compliance / regulatory analyst** | 2,400 | Reliability | LLMs with retrieval augmentation | ↓ | Medium | | **Substation technician** | (subset of SOC 49-2095, \~16,000) | Reliability + Safety | Computer vision on aerial drones | ↓ | Medium | | **Distribution lineman** | 57,800 utility-direct | Safety | Aerial drones for inspection; limited physical displacement | ↔ | Medium-long | | **Wind technician** | 4,600 | Reliability | Computer vision, predictive ML; task-level only, role persists | ↑ (industry growth) | Near | | **Solar PV installer / technician** | 3,400 | Reliability | Computer vision, optimization; task-level only, role persists | ↑ (industry growth) | Near | | **Protection engineer** | (subset of SOC 17-2071, \~20,000) | Safety | Physics-informed ML | ↔ → ↓ | Long | | **Nuclear reactor operator** | 4,300 | Safety | Physics-informed ML, RL | ↔ | Long | | **Project developer / interconnection** | (outside NAICS 221100) | Reliability | LLMs, optimization | ↑ | Near-medium | Three patterns in the table are noteworthy. The first is the heterogeneity of the role-level arrows. They do not point in one direction across the industry. Some roles grow, some hold steady, some contract, some contract sharply. The growth roles concentrate in renewables and project development, where physical buildout expands total employment even as AI augments the cognitive work around it. The contraction roles concentrate in coordination-layer work, where AI substitutes more than it augments. This is not a story of uniform displacement. It is a story of shifting composition, and the composition shift produces both winners and losers within the same workforce. The second is that growth arrows are not the same as low AI exposure. Wind technicians and solar installers are flagged up at the role level because the industries are expanding, but the task content inside those roles is already changing significantly. AI-driven inspection, diagnostics, and work-order generation are compressing the cognitive tasks that once filled the technician's shift between climbs. The physical dexterity requirement is robust; the surrounding cognitive scaffolding is not. That asymmetry is the task-versus-role distinction at work, and it generalizes beyond renewables to much of the field-physical workforce. The third is the systematic difference in horizons along the reliability-safety axis. Reliability roles change on near and medium horizons. Safety roles change on medium and long horizons. The difference is not random and is not a coincidence. It is the institutional-clock difference made visible at the role level. ## **What produces the timing, and what to do about it** Three patterns out of a twelve-row sample are suggestive rather than conclusive. They raise a question they do not answer: what is the actual mechanism by which AI capability translates into workforce change in a regulated industry, and why does it run so much faster on the reliability clock than on the safety clock? Part 2 of this series takes up that question directly. The short answer is that the translation runs through a four-stage process of capability maturation, task-content erosion, institutional-defense attenuation, and eventual role-level headcount change. Each stage has its own timeline, and the timelines differ systematically across the reliability-safety axis. The mechanism is visible in three other regulated industries with safety-critical workforces (aviation, freight rail, and maritime), each of which has moved through some version of the same arc on its own timeline. It is beginning to be visible in the US power sector now, at stage one for safety work and deeper into stage two for most reliability work. Parts of the US utility workforce have already completed the arc. Meter reading and utility customer service are the clearest examples, and they are not the last ones. ## **What comes in the rest of the series** **Part 2** develops the four-stage mechanism by which AI capability translates into workforce change in regulated industries, traces the mechanism through aviation, freight rail, and maritime precedents, maps the mechanism onto the US power sector workforce, and draws out the implications for utilities, regulators, investors, and workers. **Part 3** builds the full role-by-role matrix across the major positions in NAICS 221100 and the institutional infrastructure that surrounds it: FERC, NERC, the seven ISOs and RTOs, and the fifty state public utility commissions. Roughly forty roles, each assessed on the same axes as the twelve-role sample above. **Part 4** takes up the half of the AI-and-robotics story that gets less attention than displacement: the new capabilities that AI and robotics make possible, and that humans were never able to perform at all. Continuous inspection of all 200,000 miles of US transmission line. Energized-line close-range work. Inside-transformer monitoring. Post-storm damage pre-assessment. The capabilities that expand what a utility can do, rather than compressing what its workforce must do. **Part 5** returns to strategy, with specific recommendations for utilities, ISOs and RTOs, regulators, capital allocators, and the workforce itself, including a direct engagement with the productivity claims in current consultancy framings of AI in utilities. The translation from AI capability to utility workforce change runs through institutional defenses that hollow out from within rather than break from outside. Task content inside roles shifts first and most visibly. Role-level headcount follows on longer timelines that differ systematically between reliability and safety functions. The decade ahead looks slow until it does not. ## Sign up for AIxEnergy AIxEnergy analyzes how artificial intelligence is reshaping electricity demand, grid operations, and energy infrastructure, with a focus on system risk, policy, and institutional decision-making. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### The Grid Meets the Machine URL: https://www.aixenergy.io/the-grid-meets-the-machine/ Last updated: 2026-07-10T17:30:07.000Z There are moments when a technical report does more than inform. It reframes the problem so completely that everything written before it feels slightly misaligned. The International Energy Agency’s latest report on energy and artificial intelligence is one of those moments. For the past year, the conversation has been dominated by velocity. The rise of AI has been described in the language of exponential curves and runaway demand: data centers scaling to gigawatts, electricity systems under siege, a digital revolution outpacing the physical world. These claims are not wrong, but they are incomplete. They describe motion without structure. 0:00 /1:40 1× The IEA provides that structure. It shifts the discussion away from abstraction and into the physical system that must carry AI forward. It forces a confrontation with the reality that computation does not exist in isolation. Every model run, every inference, every generated image or autonomous task must ultimately resolve into electrons moving through wires, heat dissipated through cooling systems, and equipment manufactured, delivered, and installed over years—not milliseconds. The central insight of the report is therefore simple, but profound. The AI revolution is not just a story of demand growth. It is a story of temporal mismatch between two systems that were never designed to move together. AI operates at the speed of code, while the grid operates at the speed of infrastructure, and everything that follows flows from that tension. ## The Scale Is Undeniable—But It Is Not Unconstrained The scale of investment is staggering. Technology companies deployed more than $400 billion in capital expenditure in 2025, with expectations of another sharp increase in 2026\. A small cluster of firms now invests more annually than the entire global oil and gas upstream sector. At the same time, the IEA’s own tracking shows that specialized AI data centers have tripled in capacity in less than two years. It is tempting to read these figures as inevitability—as if the future is already written in capital allocation. The IEA resists that temptation. It insists on inserting friction back into the narrative. The report catalogues the limits of the physical world with quiet precision. Transformers require years to procure. Gas turbines face multi-year delivery timelines. Interconnection queues stretch toward a decade in some regions. Semiconductor supply chains are constrained not just by fabrication, but by memory, packaging, and even industrial gases such as helium, whose disruption reverberates through the entire system. These constraints do not stop the buildout. They shape it. They bend the trajectory of AI growth into something more uneven, more contested, and more uncertain than the clean curves of projection models suggest. The future is not simply a question of how much demand exists, but how much can be built, where, and when. ## A New Kind of Load Enters the System Electricity demand has always grown, but it has rarely changed character as quickly as it is now. The IEA’s projections show data-center electricity consumption roughly doubling by 2030, with AI-specific facilities growing even faster. Yet the deeper shift lies not in the totals, but in the form of the load itself. AI compresses energy into density. A server rack the size of a refrigerator can now draw power comparable to dozens of homes. This is not incremental demand layered onto the grid; it is concentrated demand that arrives in discrete, high-impact bursts. It also arrives with uncertainty. Data centers are often planned at full capacity but filled gradually, creating a gap between contracted load and realized demand. The grid must prepare for the maximum while operating within the reality of the partial. This breaks the logic of traditional planning. The system was built to respond to gradual change—population growth, industrial expansion, predictable cycles of use. It was not built for demand that behaves like a capital market, appearing in large, speculative blocks ahead of certainty. The challenge is no longer how to serve demand, but how to interpret it. ## Efficiency and Expansion: A System in Tension The argument that efficiency will neutralize AI’s energy impact has become a reflex. The IEA complicates that reflex without dismissing its underlying truth. On a per-task basis, AI is becoming radically more efficient. Improvements in algorithms, model design, and hardware have driven down energy use per query at a pace rarely seen in industrial systems. Tasks that once required significant compute now require a fraction of that energy, and the trajectory of improvement remains steep. But efficiency does not operate in isolation. It changes behavior. As costs fall, new use cases emerge. As capabilities expand, expectations rise. The system does not stabilize; it evolves. Video generation replaces text. Multi-step reasoning replaces simple queries. Autonomous agents replace discrete interactions. Each step increases the energy intensity of what counts as “normal.” The result is a dynamic equilibrium. Efficiency reduces the cost of computation, and lower cost expands the scope of computation. Demand grows not despite efficiency, but partly because of it. This is not a paradox to be solved, but a system to be understood. ## The Quiet Transformation of the Power Customer The technology sector is undergoing a transformation that is easy to overlook because it unfolds incrementally. It is no longer just buying electricity. It is beginning to shape the system that produces it. The IEA shows that data-center operators are among the largest purchasers of renewable energy globally. But procurement is expanding beyond renewables into a broader portfolio that includes nuclear, geothermal, and large-scale storage. This is not ideological diversification. It is operational necessity. AI workloads introduce volatility into systems that were optimized for steadiness. Training cycles, inference bursts, and distributed computation create patterns of demand that require new forms of balancing and resilience. The response is a reconfiguration of how energy is sourced, stored, and managed. The boundary between consumer and system actor begins to dissolve. What emerges is not simply a new class of customer, but a new class of infrastructure participant. ## The Mirage of Independence In response to grid constraints, some developers have turned toward onsite generation, particularly natural gas, as a way to bypass interconnection delays. The logic is intuitive: if the grid is slow, build around it. The IEA’s analysis exposes the limits of that logic. Onsite systems must be oversized to ensure reliability, often by a wide margin. They depend on the same constrained supply chains as centralized infrastructure. They require fuel, maintenance, and integration into broader operational frameworks. They do not eliminate dependency, but reconfigure it. The idea of an AI economy detached from the grid is therefore more narrative than reality. The system remains interconnected, whether that connection is acknowledged or not. ## The Politics of Visibility Electricity prices sit at the intersection of economics and perception, and the IEA treats them with appropriate caution. Data centers do not universally raise prices, but they create conditions under which price pressure becomes more likely. The problem is not simply scale. It is timing and uncertainty. Large loads can trigger investment before their full shape is known, and those investments must be paid for regardless of how demand evolves. This is where visibility matters. Data centers are not abstract additions to the system. They are physical structures, located in communities, drawing power in ways that are both concentrated and conspicuous. They become symbols of broader concerns about fairness, cost allocation, and environmental impact. In that context, perception becomes a form of infrastructure, shaping policy as directly as engineering constraints. ## The Imbalance at the Heart of the System Perhaps the most consequential insight in the report is not about data centers at all, but about the energy sector’s own adoption of AI. The technology that is driving new demand has not yet been fully deployed to manage that demand. Barriers are practical and persistent: limited digital expertise, fragmented data systems, cybersecurity concerns, and policy frameworks that lag behind technological capability. The result is an asymmetry in which the system absorbs the cost of AI expansion before it captures its efficiency gains. This imbalance is temporary, but it is not trivial. It shapes how the transition is experienced, and therefore how it is governed. ## Synchronization as the Defining Challenge The IEA does not name it explicitly, but the concept that emerges from its analysis is synchronization. Two systems—digital and physical—are evolving along different timelines, governed by different constraints, and optimized for different objectives. When those timelines diverge, friction appears as congestion, cost escalation, and planning failure. The task is not to accelerate one system to match the other, but to coordinate them. This requires new forms of visibility into demand, new mechanisms for flexibility, and new frameworks for allocating cost and risk. It requires a shift from reactive planning to anticipatory design. Above all, it requires acknowledging that the problem is not one of scale alone, but of alignment. ## Conclusion Artificial intelligence has introduced a new tempo into the economy, one defined by iteration, deployment, and rapid scaling. The energy system operates on a different tempo, one defined by construction, regulation, and long-lived assets. The IEA’s report reveals what happens when those tempos intersect. The future of AI will not be determined solely by advances in computation. It will be determined by the capacity of the physical system to absorb, adapt, and synchronize with that computation. Infrastructure, not imagination, becomes the limiting factor. The electricity surge now underway is not simply a rise in demand, but a collision between two ways of building the world. One moves in milliseconds, while the other moves in years, and the outcome depends on how—and whether—they learn to move together. ## References Associated Press. “Impact of Middle East Conflict on Industrial Gas Supply.” 2026. Bloomberg News. “AI Companies Turn to Debt Markets as Capex Surges.” 2025. Data Center Watch. “US Data Center Project Delays and Cancellations.” 2025. International Energy Agency. *Energy and AI: Key Questions on Energy and Artificial Intelligence*. World Energy Outlook Special Report. Paris: International Energy Agency, 2026. IDC. *High-Bandwidth Memory Supply Constraints and Outlook*. 2025. Micron Technology. *Memory Manufacturing and Capacity Trends*. 2024. Reuters. “Semiconductor Supply Chain Disruptions and Market Response.” 2025. SemiAnalysis. “AI Server Architecture and Memory Scaling.” 2025. U.S. Geological Survey. “Global Helium Supply Disruptions.” 2026. Wood Mackenzie. *Gas Turbine Supply Chain Outlook*. 2025. ### Speed to Power Is Not a Transmission Ownership Problem URL: https://www.aixenergy.io/speed-to-power-is-not-a-transmission-ownership-problem/ Last updated: 2026-06-30T02:31:35.000Z The Constraint Has Shifted—and Most of the Industry Is Still Solving the Old One _This post is for paying subscribers only._ ### The Grid Wasn’t Built for AI URL: https://www.aixenergy.io/the-grid-wasnt-built-for-ai/ Last updated: 2026-06-30T02:31:27.000Z The real problem isn’t how much power data centers use—it’s how their demand behaves, and why the system doesn’t price it. _This post is for paying subscribers only._ ### The Grid Is Losing Its Memory URL: https://www.aixenergy.io/the-grid-is-losing-its-memory/ Last updated: 2026-06-30T02:31:16.000Z As AI-driven demand surges, a deeper shift is unfolding inside the power system _This post is for paying subscribers only._ ### The New Geography of Power: How States Are Shaping the AI Infrastructure Era URL: https://www.aixenergy.io/the-new-geography-of-power-how-states-are-shaping-the-ai-infrastructure-era/ Last updated: 2026-07-29T19:40:30.000Z States are competing to shape AI as physical infrastructure. Diverging archetypes—from hyperscale clusters to energy-led expansion—are creating a new hierarchy. Regions that align compute, energy, and institutions will lead; others will operate within systems designed elsewhere. _This post is for paying subscribers only._ ### Shadow Grid: The Grid Is Splitting in Two URL: https://www.aixenergy.io/the-darker-shadows-of-the-shadow-grid/ Last updated: 2026-07-29T19:41:57.000Z The S[hadow Grid](https://www.aixenergy.io/shadow-grid-articles/) is not emerging from legislation or long-range planning. It is being built, contract by contract, by data center developers solving a simpler problem: how to secure power on their own terms. Faced with interconnection delays, uncertain timelines, and rising costs, they are assembling a parallel system of generation, storage, and control directly on or adjacent to their campuses. It runs alongside the traditional grid but is not governed by it in the same way. It is built for certainty, speed, and autonomy—not participation. Article Summary 0:00 /206.326939 1× On the ground, this shift is already visible. In Ohio and New Mexico, in Tennessee and Utah, natural gas turbines, solar arrays, and battery systems are being designed into data center campuses from the outset. These are not backup resources waiting for an outage. They are primary infrastructure, sized and configured for continuous operation. Some sites maintain selective ties to the grid, drawing power when it is advantageous. Others are engineered to operate largely on their own. What is taking shape is a distributed, privately controlled layer of energy infrastructure that mirrors the grid in function but not in form. To understand the magnitude of this buildout, AIxEnergy assembled a national dataset from state air permits, siting dockets, county filings, corporate disclosures, earnings calls, and equipment procurement signals, combined with probability-weighted commercial operation timelines. The numbers tell a precise story. Between 2026 and 2028, expected shadow gas additions—47 gigawatts—closely track the 46 gigawatts of gas planned for grid connection. But the symmetry does not hold. In 2027, probability-weighted shadow capacity reaches 21.7 gigawatts, exceeding the 12.3 gigawatts slated for the grid. Across three years, these projects represent 186 gigawatts of privately developed capacity exposure—generation that exists, but does not fully register in the models planners rely on. [AIxEnergy Proprietary DataDownload structured datasets that reveal how AI is reshaping electricity demand, infrastructure constraints, and the evolving architecture of the energy system.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-183.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/ChatGPT-Image-Mar-28--2026--02_40_27-PM-1.png)](https://www.aixenergy.io/data-2/) Most of this system remains difficult to see. It does not pass cleanly through interconnection queues. It does not fully appear in resource adequacy plans. And it sits only partially inside the datasets used to forecast how the grid will evolve. In a previous AIxEnergy article, [The Shadow Grid: How Regulatory Seams Are Reshaping American Power](https://www.aixenergy.io/shadowgridseams/), [Brandon Owens](https://www.aixenergy.io/author/brandon/) grounds the governance problem in a simple observation: grid regulators and operators depend on interconnection queues, resource adequacy plans, and capacity forecasts to understand the system they are responsible for managing. The shadow grid introduces blind spots in all three. The Federal Energy Regulatory Commission’s December 2025 order is a direct response to that problem. In important respects, it is the right response. But it carries an unintended consequence that has not been fully examined. By raising the cost and complexity of grid participation for large operators, it creates a financial incentive to bypass the system altogether. The risk is not simply more private generation. It is a version of the shadow grid that becomes structurally invisible—outside planning processes, outside reliability frameworks, and increasingly outside coordination. To make that risk clear, it is necessary to be precise about what the shadow grid actually is. Not all behind-the-meter configurations are the same. And the policy consequences depend on which version of the system is actually emerging. # The shadow grid is not one thing When people talk about the shadow grid, they typically mean data centers that generate their own power behind the meter rather than drawing it from the grid. But the FERC order and PJM’s compliance filings reveal at least four distinct configurations, each with a different regulatory treatment, a different resource adequacy impact, and a different risk profile for investors. ## Truly islanded generation The first and most extreme configuration is zero grid connection: no transmission service taken, no Network Load designation, no PJM obligation to serve the load. PJM has no visibility into this load at all. It does not exist in the planning framework, and because PJM cannot see it, there is no PJM-imposed cap on it. ## “Option 5”: grid-connected with protections The second configuration is what PJM calls “Option 5” in its co-location filings: a grid-connected arrangement where physical protections prevent grid energy from flowing to the load under normal conditions, but where PJM backup service remains available and the load retains its Network Load designation. PJM does plan for this load, and the operator carries obligations. Because this configuration does not carry reserves equal to the installed reserve margin, allowing it to grow unchecked degrades PJM’s loss of load expectation (LOLE), the standard measure of the probability the grid will be unable to serve demand. PJM [accommodates “Option 5” through a 2006 settlement agreement](https://www.pjm.com/-/media/DotCom/documents/ferc/filings/2025/20250324-el25-49-000.pdf?ref=aixenergy.io) that relaxed the LOLE standard from 1-in-10 to 1-in-9.5 years, and caps this configuration at approximately 2,000 MW system-wide, of which roughly 1,300 MW are already in use. This cap is relevant to the perverse incentive argument below. ## Large BTM with grid connection The third configuration is the primary target of FERC’s December 2025 order: large behind-the-meter generation with a full grid connection. Here, a data center generates some of its own power but takes Network Integration Transmission Service (NITS) from the grid and carries a full Network Load designation, meaning PJM has a legal obligation to serve it. Under existing rules, PJM plans for the net load. If a 500 MW data center has a 500 MW behind-the-meter generator, it appears in PJM’s planning as zero net load. PJM must serve it but cannot plan the resources to do so. [FERC found](https://www.bakerbotts.com/thought-leadership/publications/2025/december/ferc-issues-order-providing-guidance-for-co-locating-power-plants-with-data-centers-within-pjm?ref=aixenergy.io) that “loads with BTMG are not fully accounted for in resource adequacy planning,” and the behind-the-meter generator that offsets the capacity obligation cannot participate in PJM’s capacity market to fill that planning gap. [Manual 14G states this explicitly](https://www.pjm.com/-/media/DotCom/documents/manuals/m14g.pdf?ref=aixenergy.io). FERC’s proposed remedy is a materiality threshold: facilities above a certain size lose the ability to net their behind-the-meter generation against their load obligation. The full load reappears in PJM’s resource adequacy planning, and PJM can plan for what it must serve. That is the right outcome, if the threshold holds. ## Full co-location: the reference point Before introducing the fourth configuration, it is worth establishing what full grid participation looks like. In a true co-location arrangement, the generator interconnects directly to the grid as a capacity resource, the load takes full NITS service, and both sides are fully visible to PJM’s planning and capacity market. The load-serving entity procures capacity for the full gross load. There is no governance gap. This is the reference point against which the intermediate configurations should be evaluated. ## BYOG: the proposed bridge The fourth configuration sits between large BTM and full co-location. Bring Your Own Generation, or BYOG, was introduced in PJM’s [January 2026 Board Decisional Letter](https://www.pjm.com/-/media/DotCom/about-pjm/who-we-are/public-disclosures/2026/20260116-pjm-board-letter-re-results-of-the-cifp-process-large-load-additions.pdf?ref=aixenergy.io) as an expedited interconnection path. Under BYOG, a data center commits to bring new generation capacity to the system (through new-build facilities, uprates to existing generators, or PPAs) in exchange for faster interconnection and acceptance of curtailment risk rather than full firm service. The generation does not have to be physically behind the meter or on-site; the defining trade is speed in exchange for interruptibility. *(The formula of speed in exchange for a commitment is not new. The 2025 uncertainty framework published on this platform proposed using expedited interconnection as an incentive for tangible carbon reduction commitments. BYOG uses the same mechanism but substitutes curtailment acceptance for environmental performance.)* BYOG differs from full co-location in an important respect. In a true co-location arrangement, the generator interconnects directly to the grid as a capacity resource and the load takes full NITS service, with both sides visible to PJM’s planning. BYOG accepts interruptibility in place of firm service and does not require the generator to offer into PJM’s capacity market, at least not yet. The governance question BYOG raises is this: without a requirement to offer into PJM’s capacity market, the generation a data center “brings” may never become visible to PJM’s supply-side planning. The facility remains Network Load, PJM retains its service obligation, and the threshold change makes the load visible on the demand side. But if the BYOG generation stays off PJM’s books, the supply side remains in the shadows. BYOG could, without the right design, replicate the very governance problem it was meant to solve. The table below summarizes the configurations from least to most grid engagement. | Configuration | PJM Obligation | What It Means | | ------------------------------- | --------------------- | --------------------------------------------------------------------------------------------------------------------------- | | **Truly islanded** | None | No grid connection. Invisible to PJM. No planning visibility and no system cap applies. | | **“Option 5”** | Limited (backup only) | Grid-connected with protections. PJM sees the load but only as backup. Capped at roughly 2,000 MW system-wide. | | **Large BTM (above threshold)** | Full | Netting eliminated. PJM must plan for the full gross load. Primary target of the FERC remedy. | | **BYOG** | Full | Faster interconnection in exchange for interruptibility. Load is visible, but supply-side visibility may remain unresolved. | | **Full NITS / co-location** | Full | Both load and supply are fully visible to PJM. Fully integrated into planning and capacity markets. | # A threshold that may not hold FERC’s December 2025 order directed PJM to propose a materiality threshold above which BTM netting no longer applies. [PJM’s February 23, 2026 compliance filing proposed 50 MW](https://www.ferc.gov/news-events/news/ferc-directs-nations-largest-grid-operator-create-new-rules-embrace-innovation-and?ref=aixenergy.io). FERC has not yet accepted or rejected that level, and the outcome is far from certain. If FERC sets the threshold significantly higher than 50 MW, or rejects PJM’s proposal, a large portion of the hyperscale buildout remains in the netting-makes-both-invisible situation indefinitely. The resource adequacy gap Owens documented and FERC correctly diagnosed does not get fixed. Most coverage of this rulemaking treats the threshold as settled. It is not, and investors underwriting BTM projects on the assumption that the regulatory framework will stabilize around 50 MW should treat that assumption with care. # FERC is right, but creates a new risk Assuming the threshold takes effect near 50 MW, the order correctly addresses the resource adequacy and cost-causation problems created by large BTM loads operating outside the planning framework. The risk is not that FERC got the diagnosis wrong. It is that the remedy creates a financial incentive to exit the grid entirely. When netting is eliminated for a large BTM facility, the load-serving entity (LSE) responsible for that customer must now procure capacity for the full gross load rather than the net. The cost of that additional procurement is allocated across LSEs through the Locational Reliability Charge, PJM’s mechanism for recovering capacity commitments from load-serving entities based on their load obligations. Whether and how that cost ultimately reaches the specific data center customer depends on the retail tariff structure and the contract between the LSE and its customer. In competitive retail markets, LSEs typically pass incremental costs through. At PJM’s current capacity price of $329/MW-day, the order of magnitude of that additional cost for a 500 MW facility is approximately $60 million annually. For investors who underwrote the BTM structure specifically to avoid capacity costs, this is not a routine regulatory adjustment. It is a fundamental repricing of the investment thesis. For an operator who can make the physics work (who can genuinely island the facility with zero grid connection and no Network Load designation), exiting the framework entirely may be the rational response. The result would be no capacity charges, no NITS fees, no ancillary service obligations. The FERC order, by raising the cost of remaining in the framework, inadvertently creates a financial incentive to become exactly what Owens’ series identifies as the core governance problem: a facility genuinely invisible to the institutions designed to plan around it. This is where the 2,000 MW cap under PJM’s 2006 settlement becomes relevant. That cap applies to “Option 5,” the grid-connected-with-protections configuration where PJM retains a backup service obligation and can still see the load. If the financial pressure of the FERC order pushes operators toward genuine islanding with zero grid connection (outside even “Option 5”), that cap provides no constraint whatsoever. The reliability framework it represents would face pressure it was never designed to absorb. ## Two consequences of scaled full islanding The first is carbon accountability. Truly islanded gas generation is structurally exempt from every decarbonization mechanism: utility RPS compliance, state clean energy accounting, carbon inventories. It is not interconnected, not metered by utilities, and not visible to any regulatory framework designed to track and reduce emissions. Owens’ bifurcated carbon ledger does not just persist under this scenario. It expands. The second is gas-electric coordination failure, the piece of the shadow grid story that has not been told. Today, gas and electric market coordination happens through the dispatch stack and firm transportation contracts. PJM knows which gas-fired generators need fuel and can dispatch alternatives when they lose it. The central lesson of Winter Storm Uri was that this coordination can fail catastrophically, but at least the failure was visible and could be diagnosed. Truly islanded data center power plants break this coordination in a way that is fundamentally harder to diagnose. They are invisible to PJM, so it cannot dispatch alternatives when an islanded generator loses fuel. They may not hold firm gas transportation contracts, since there is no regulatory pressure requiring them to do so. Unlike grid-connected generators, they have no place in the coordination framework that allocates fuel supply during stress events. The coordination failure is acute precisely when coordination matters most. # The April 17 fork BYOG is designed as the alternative to full islanding: commit to bring generation, accept interruptibility, and receive faster interconnection. Whether it actually closes the resource adequacy gap depends on the FERC paper hearing reply brief deadline of April 17, 2026 (docket EL25-49-000), which will determine whether large BYOG generators must offer into PJM’s capacity market. If FERC requires capacity market participation, the generation committed through BYOG becomes visible to PJM’s supply-side planning alongside the load it already sees. The resource adequacy gap closes. This moves BYOG meaningfully toward the full co-location model: visible load, visible supply. If FERC does not require it, the threshold change creates a problem that is in one respect worse than today. Currently, PJM is blind to both the load and the supply for large BTM configurations. Under the threshold change without capacity market participation requirements, PJM sees the full load obligation but the supply covering it remains invisible. Planning is blind on only one side, and it is the wrong side. Most investors are treating the April 17 outcome as settled. It is not. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/03/shadow_grid_decision_tree.png) # The standby charge gap The actual ratepayer protection mechanism, where it exists, is state utility commission standby charge tariffs. When a large customer self-supplies but remains connected to the grid for backup or supplemental power, a properly designed standby charge requires payment toward grid infrastructure costs even when the customer is not actively drawing power. The charge has two components. The first is availability: the grid must be built and maintained to serve that customer at full load whenever its private generation fails. The second is system stability: large BTM generators affect grid frequency, voltage support, and reactive power balance continuously, whether or not they are drawing power. These are real costs that neighboring customers currently bear without recovery. Virginia is the most consequential near-term jurisdiction. In November 2025, the State Corporation Commission approved Dominion Energy’s new [GS-5 rate class](https://www.scc.virginia.gov/docketsearch?ref=aixenergy.io#/caseDetails/14) for customers consuming over 25 MW and required minimum demand charges to protect ratepayers from infrastructure costs. But the SCC explicitly [deferred the harder question](https://virginiamercury.com/2025/11/25/scc-approves-chesterfield-gas-plant-and-dominion-rate-hike-creates-new-rate-class-for-data-centers/?ref=aixenergy.io) of how to assign generation system costs to data centers, asking Dominion to return with a proposal in its 2027 rate case. That deferral is where the standby charge gap lives. For investors evaluating BTM projects in Virginia and similar jurisdictions, how the 2027 case resolves is a material open question, and the direction of resolution is not favorable to the current cost structure. # The answer that isn’t winning yet The debate over the shadow grid has been conducted almost entirely in terms of what is wrong with BTM gas. Less attention has gone to a constructive model that already exists and resolves the carbon, reliability, ratepayer, and coordination problems simultaneously while meeting the speed requirements of the AI buildout. The Power Couples framework, developed by Uday Varadarajan and colleagues at [RMI](https://rmi.org/how-power-couples-can-help-the-united-states-win-the-global-ai-race/?ref=aixenergy.io) and now being commercialized through [Colectric](https://colectric.com/?ref=aixenergy.io), pairs a large electricity consumer with new-build solar, wind, and storage resources sized to meet the on-site load, sited near an existing generator with approved grid interconnection. The existing plant’s surplus interconnection capacity provides grid access in months rather than years. Physical safeguards ensure no net grid power withdrawal, eliminating the resource adequacy concern entirely. The arrangement stays inside the institutional framework, visible to planners, countable in resource adequacy studies, and subject to the regulatory processes that BTM gas is bypassing. RMI’s analysis finds that over 50 GW of data center load could be served this way at 88% carbon-free energy on average, with over 30 GW available below $100/MWh. Colectric lists Apple, Meta, and Akamai among its incubation partners. The hyperscalers driving the BTM gas buildout are, in some cases, the same companies that are supporting the alternative. One honest constraint: the Power Couples model depends on the geographic distribution of existing generators with surplus interconnection, and the AI buildout’s concentration in climate-stressed Southern geographies creates real siting limits that any co-location model must navigate. Power Couples is not currently winning primarily because the regulatory framework has not yet made the full costs of BTM generation visible to the operators who choose it. When the materiality threshold takes effect, when standby charges reflect grid stability costs, and when gas price assumptions embedded in long-lived BTM capital assets are stress-tested against a globalizing gas market, the Power Couples economics improve materially relative to every segment of the shadow grid. The FERC order is a step in that direction. Its blind spot is the fully islanded model it may inadvertently accelerate. # Three near-term levers **April 17, 2026** is the most time-sensitive decision point. The FERC paper hearing (docket EL25-49-000) will determine whether BYOG generators must participate in PJM’s capacity market. A ruling requiring participation would close the visible-load-invisible-supply problem the threshold change otherwise creates, and would signal that grid engagement is the viable path rather than a cost to be avoided. **Virginia’s 2027 SCC rate case** is the second lever. The SCC’s November 2025 order deferred the question of how to assign generation system costs to data centers. When Dominion returns with a proposal, the outcome will determine whether standby charge-equivalent obligations are imposed on large BTM customers in the most important data center market in the country, and will set a template other states follow. **Disclosure** is the third lever. The [data center uncertainty framework](https://www.aixenergy.io/) published on this platform in November 2025 proposed granular behind-the-meter generation and storage plans filed with utilities and ISOs. That requirement would make the shadow grid visible to the institutions currently blind to it, and give gas pipeline operators the information they need to coordinate fuel supply for loads they cannot currently see. Owens’ series calls for enhanced reporting requirements, expanded regional planning roles, and AI transparency mandates for grid management algorithms, the right long-run responses. The three levers above are more immediately tractable and are where the most consequential decisions are being made right now. The shadow grid is not an accident of technology. It is the outcome of rules that stopped at the edge of the grid while the system itself kept expanding. Those edges are now visible. Some are being closed. Others remain open in ways that matter more with each new facility that steps outside the frame. The risk is not that power moves off the grid. It is that governance does not follow it. If the incentives continue to favor full islanding, if fuel coordination remains fragmented, if cost recovery lags behind physical reality, then the system will evolve into something planners can no longer fully see—much less manage. The next failure will not begin as a shortage of energy. It will begin as a gap between what the system is responsible for and what it can observe. And by the time that gap is visible, it will already be structural. [*Michael Leifman*](https://www.aixenergy.io/author/michael/) *is the founder of Tenley Energy Innovation LLC and writes about energy markets and AI infrastructure at* [*MichaelLeifman.Substack.com*](https://michaelleifman.substack.com/)*.* ### The Age of Electricity Meets the Age of Constraint: IEA’s Energy Technology Perspectives 2026 Describes a System in Transition URL: https://www.aixenergy.io/the-age-of-electricity-meets-the-age-of-constraint-ieas-energy-technology-perspectives-2026-describes-a-system-in-transition/ Last updated: 2026-07-10T17:30:16.000Z The International Energy Agency’s *Energy Technology Perspectives 2026* lands at a moment when the energy system is no longer defined by a single variable. For most of the twentieth century, the central problem was scale: how to produce more energy, more cheaply, and deliver it reliably. The early twenty-first century added a second axis—carbon—forcing the system to decarbonize while continuing to grow. Article Summary 0:00 /165.672063 1× Now a third dimension has emerged. The system is becoming constrained—not by any one factor, but by many at once. Electricity, materials, manufacturing capacity, trade policy, infrastructure timelines, and physical resources are beginning to bind together. The result is a system that is expanding and tightening at the same time. The IEA’s report does not state this outright. But read carefully, it documents the transition from a single-constraint system to a multi-constrained one. ## From Policy-Driven Growth to Market Momentum Beneath the market expansion is a set of reinforcing technical dynamics that the report captures but does not foreground. Solar PV’s cost decline is not only a function of scale, but of incremental efficiency gains in cell architectures—TOPCon and heterojunction designs pushing module efficiencies beyond 23–25 percent in commercial deployments. Manufacturing improvements, including wafer thinning and reduced silver content, are lowering material intensity per watt even as total output rises. Battery systems are undergoing a similar evolution. Lithium iron phosphate (LFP) chemistries have displaced nickel-rich cathodes in many applications due to cost, safety, and cycle life advantages. At the same time, cell-to-pack and cell-to-chassis integration are reducing inactive material, improving energy density at the system level rather than the cell level alone. These engineering shifts are as important as raw cost declines in enabling deployment. Wind technologies are scaling through rotor diameter expansion and higher hub heights, increasing capacity factors rather than relying solely on cost reductions. Offshore wind, in particular, is moving toward turbines exceeding 15 MW, introducing new challenges in installation logistics, transmission integration, and reliability. These technology-level improvements explain why cost curves have continued to fall even as input prices fluctuate. The system is learning not just how to build more, but how to build differently. The report opens with a fact that would have been controversial even five years ago: clean energy technologies now represent a market approaching $1.2 trillion.¹ Solar, wind, batteries, and electric vehicles are no longer emerging sectors. They are industrial systems with global scale. The pace of growth is equally striking. Over the past decade, these technologies have expanded at roughly 20 percent per year.² Even under conservative policy assumptions, the IEA projects that their market value will double again by 2035. What matters is not only the magnitude of this growth, but its source. For most of the transition, policy created the market. Subsidies, mandates, and carbon targets drove deployment. That relationship is now shifting. Costs have fallen to the point where economics, not policy, increasingly drives adoption. Around 80 percent of global solar and wind generation now occurs at lower cost than coal or gas.³ Battery prices have declined by roughly 75 percent over the past decade.⁴ In some markets, electric vehicles are approaching or achieving cost parity with internal combustion engines. This is the quiet turning point. Technologies that once required justification now propagate through the system on their own terms. And yet, the system is not stabilizing. It is becoming more complex. ## The Structure Beneath the Growth Curve The report’s most consequential insight lies beneath its market projections: the clean energy transition is structurally dependent on a narrow set of supply chains. China’s position is not simply dominant; it is foundational. Across key technologies, it controls between 60 and 85 percent of manufacturing capacity—and in some stages, more than 95 percent.⁵ These concentrations are especially acute in midstream processing and critical minerals, where substitution is difficult and lead times are long. The implications are measurable. The IEA estimates that a one-month disruption in Chinese battery exports would remove approximately $17 billion in output from global electric vehicle production.⁶ A similar disruption in solar supply chains would erase roughly $1 billion in monthly output from manufacturing outside China.⁷ These figures describe more than exposure. They describe system behavior. A localized disruption propagates globally because the system has become tightly coupled. The modern energy system is not just electrifying. It is becoming interdependent in ways that amplify both efficiency and fragility. ## Trade Policy Enters the System Core For decades, trade policy sat at the periphery of energy analysis. In *Energy Technology Perspectives 2026*, it moves to the center. The number of restrictive trade measures worldwide has increased nearly sixfold since 2015.⁸ Tariffs, export controls, and industrial subsidies are no longer episodic responses. They are structural features of the system. Governments are acting with clear intent: to secure domestic industries, reduce dependence, and capture economic value from the energy transition. But the outcomes are more ambiguous than the intent. Tariffs raise costs, but often less than expected. In many cases, their impact is diluted by falling commodity prices, shifts in trade flows, and increased domestic production.⁹ Solar modules, for example, account for only 10 to 15 percent of the total cost of a residential installation in many advanced economies, limiting the effect of tariffs on end users.¹⁰ At the same time, trade continues to expand. The IEA projects that the global value of clean energy technology trade will more than double by 2035, even in a more fragmented geopolitical environment.¹¹ The system is fragmenting politically while remaining integrated economically. That tension is not temporary. It is structural. ## A System Moving at Multiple Speeds The divergence in technology maturity is rooted in underlying engineering realities. Hydrogen production, for example, is constrained not only by cost but by electrolyser performance characteristics. Alkaline electrolysers offer durability and lower cost but limited dynamic response, while proton exchange membrane (PEM) systems provide flexibility suited to variable renewables but at higher capital cost due to reliance on scarce materials such as iridium. Solid oxide electrolysers promise higher efficiencies but remain in early stages of deployment, with durability challenges under real-world cycling conditions. Carbon capture systems face a different set of constraints. Post-combustion capture using amine solvents is commercially available, but energy penalties—often 15–25 percent of plant output—remain a barrier. Emerging approaches, including solid sorbents and membrane-based separation, aim to reduce these penalties but have yet to scale. Transport and storage infrastructure further complicate deployment, requiring coordinated development of pipelines and geological storage sites. In heavy industry, near-zero emissions steel production using direct reduced iron (DRI) with hydrogen introduces process integration challenges. Hydrogen purity, temperature control, and reactor design all affect metallurgical quality. Similarly, low-carbon cement pathways—such as calcination with carbon capture or alternative clinker chemistries—must balance cost, durability, and compatibility with existing construction standards. At the frontier, nuclear fusion and electrochemical processes represent a different class of innovation. Advances in high-temperature superconducting magnets have enabled recent fusion milestones, but confinement stability and net energy gain remain unresolved. Electrochemical pathways for ammonia and iron production offer theoretical efficiency gains but face scaling challenges in current density, catalyst durability, and system integration. These are not marginal technical hurdles. They define the pace at which each technology can move from concept to system integration. The energy transition is therefore governed not only by economics and policy, but by the physics and engineering limits embedded in each pathway. One of the report’s strengths is its refusal to treat clean energy as a single category. Instead, it maps technologies along a continuum of maturity. At one end are industrialized systems—solar, wind, batteries, electric vehicles—where scale and cost reductions reinforce each other. In the middle are technologies such as hydrogen and carbon capture. Investment is rising—global spending on low-emissions hydrogen reached nearly $8 billion in 2025, and CCUS investment has increased more than fifteen-fold since 2020—but deployment remains uneven.¹² Many projects have yet to reach final investment decision. At the far end are technologies still in early development: fusion, advanced materials, electrochemical processes. These attract capital and attention, but their timelines remain uncertain. The implication is straightforward. The energy transition is not a single curve. It is a layered system evolving at different speeds. That layering introduces friction. Technologies do not integrate smoothly when their timelines diverge. ## Industrial Competitiveness as a Binding Constraint Industrial competitiveness is ultimately anchored in manufacturing detail. In battery production, yield rates, electrode coating uniformity, and formation cycling protocols determine both cost and performance. Small improvements in yield—moving from 90 to 95 percent—can materially shift cost structures at scale. Similarly, advances in dry electrode processing have the potential to reduce energy consumption and capital intensity, though they are not yet widely commercialized. Solar manufacturing competitiveness depends on upstream integration. Control over polysilicon purification, ingot growth, and wafer slicing enables cost reductions that downstream assembly alone cannot achieve. Diamond wire sawing and kerfless wafer technologies are reducing material waste, while automation is driving labor cost differentials. Wind turbine manufacturing reflects a different set of constraints. Blade production is limited by transport logistics as much as by fabrication cost, driving interest in modular blade designs and on-site manufacturing. Gearbox reliability, power electronics, and condition monitoring systems increasingly determine lifecycle cost rather than upfront capital alone. In electrolysers, manufacturing scale is still emerging. Stack design, catalyst loading, and balance-of-plant integration all influence cost trajectories. Unlike solar or batteries, there is not yet a dominant design paradigm, leaving room for regional differentiation—but also increasing uncertainty. These details matter because they accumulate. Competitiveness is not determined by a single breakthrough, but by thousands of incremental improvements across the production chain. That accumulation is what currently separates leading manufacturing systems from those attempting to catch up. Perhaps the most important shift in the report is its emphasis on industrial competitiveness. Energy transitions were once framed in terms of resources and emissions. Increasingly, they are framed in terms of manufacturing advantage. China’s cost position reflects decades of accumulated scale, integrated supply chains, workforce development, and sustained policy support.¹³ The IEA disaggregates this advantage with unusual clarity. In battery manufacturing, efficiency accounts for more than 40 percent of the cost gap between China and Europe.¹⁴ In upstream solar production, energy and labor costs account for roughly 65 percent of the difference.¹⁵ These are structural gaps. They do not close quickly. At the same time, energy costs themselves are reasserting their influence. In energy-intensive industries, they can account for more than two-thirds of total production cost.¹⁶ This creates a new geography of advantage. Regions with access to low-cost energy—parts of the Middle East or North Africa—may outcompete traditional industrial centers in producing next-generation materials. The transition is reorganizing industrial geography as much as it is decarbonizing energy. ## Closing *Energy Technology Perspectives 2026* is disciplined in tone. It avoids exaggeration and resists narrative overreach. Its value lies in its clarity. But its most important contribution may be indirect. It shows a system that is expanding rapidly while becoming structurally tighter. It documents growth that is real, but increasingly conditional. It reveals a transition that is no longer governed by a single dominant constraint, but by the interaction of many. What it leaves largely unstated is the implication. As energy systems become more electrified, more digital, and more interdependent, the central challenge shifts. It is no longer sufficient to deploy technology at scale. The task becomes governing a system where constraints interact, propagate, and occasionally conflict. That is the next phase of the transition. The IEA’s report captures a system in motion—past the point of proof, but not yet in equilibrium. The technologies are scaling. The markets are expanding. The capital is flowing. But the system is tightening. And within that tightening—across supply chains, trade, manufacturing, and infrastructure—the next set of risks, and the next opportunities, are already forming. ## Sign up for AIxEnergy AIxEnergy analyzes how artificial intelligence is reshaping electricity demand, grid operations, and energy infrastructure, with a focus on system risk, policy, and institutional decision-making. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## Notes 1. International Energy Agency, *Energy Technology Perspectives 2026* (Paris: IEA, 2026), Executive Summary. 2. Ibid. 3. Ibid. 4. Ibid. 5. Ibid. 6. Ibid. 7. Ibid. 8. IEA, *Energy Technology Perspectives 2026*, Chapter 1. 9. Ibid. 10. Ibid. 11. IEA, *Energy Technology Perspectives 2026*, Executive Summary. 12. Ibid. 13. Ibid. 14. Ibid. 15. Ibid. 16. Ibid. ## Bibliography International Energy Agency (IEA). *Energy Technology Perspectives 2026*. Paris: IEA, 2026. ### Why Energy Infrastructure Is Now Multi-Constrained URL: https://www.aixenergy.io/multi-constraint/ Last updated: 2026-06-30T02:30:40.000Z Water, cooling, and permitting are emerging as binding constraints alongside electricity—reshaping where AI and large-scale digital infrastructure can be built, how quickly, and at what risk. _This post is for paying subscribers only._ ### The Shadow Grid Doctrine: The United States Is Building a Second Energy System for Artificial Intelligence URL: https://www.aixenergy.io/the-shadow-grid-doctrine/ Last updated: 2026-07-10T17:30:41.000Z On March 20th, 2026, the White House released its National Policy Framework for Artificial Intelligence. Most policy documents fade quickly into the background noise of Washington—summarized, debated, and then absorbed into the slow churn of implementation. This one will not. Article Summary 0:00 /122.565782 1× Because this is not, in substance, an AI policy document. It is an infrastructure document. [03.20.26-National-Policy-Framework-for-Artificial-Intelligence-Legislative-Recommendations03.20.26-National-Policy-Framework-for-Artificial-Intelligence-Legislative-Recommendations.pdf234 KBdownload-circle](https://www.aixenergy.io/content/files/2026/03/03.20.26-National-Policy-Framework-for-Artificial-Intelligence-Legislative-Recommendations.pdf "Download") Read carefully, its logic becomes unmistakable. It does not ask what artificial intelligence should be allowed to do. It asks how fast it can be built. It does not begin with risk, or rights, or system coordination. It begins with capacity—compute, energy, and physical deployment at scale. The organizing principle is acceleration. That choice places the United States on a distinct trajectory. Not merely toward leadership in artificial intelligence, but toward the rapid construction of a new layer of physical and economic infrastructure—one that is only partially visible within the institutions designed to govern it. Artificial intelligence is no longer just software running on distant servers. It is becoming one of the fastest-growing sources of electricity demand in modern history. It is reshaping capital allocation across energy and industrial systems. It is embedding decision-making into the operational fabric of infrastructure itself. The framework recognizes the opportunity. It does not yet fully account for the consequences. And those consequences will not unfold gradually. They are already beginning. ## What the Framework Accelerates—and What It Leaves Exposed The White House framework will be described, in familiar terms, as pro-innovation and pro-growth. That description is accurate but incomplete. The more precise characterization is that it elevates infrastructure development above all competing priorities. Speed becomes the governing objective. Friction is treated as a liability. Coordination is assumed to follow. That ordering matters. Because infrastructure, once built, does not wait for governance to catch up. What follows are not hypothetical risks. They are structural conditions already set in motion by the framework’s design. ### 1\. Infrastructure First, Governance Later The framework accelerates physical buildout—data centers, energy systems, and supporting assets—without establishing a corresponding system of control. This sequencing is not neutral. Infrastructure defines behavior. It locks in incentives, patterns of use, and pathways of dependency. By the time governance adapts, the system it seeks to govern is already operating at scale. The risk is not that governance fails to emerge, but that it arrives too late to shape outcomes in a meaningful way. ### 2\. The Formal Emergence of the Shadow Grid In endorsing behind-the-meter generation and private power procurement, the framework quietly authorizes a second energy system to take shape. Data centers will increasingly co-locate with dedicated generation, optimizing for uptime and cost rather than system-wide coordination. This is not an incremental evolution of the grid. It is a structural divergence. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/03/ChatGPT-Image-Mar-21--2026--02_38_44-PM.png) What emerges is a layered architecture: a privately controlled, rapidly expanding energy network operating alongside the traditional grid, but not fully governed by it. Visibility declines. Coordination weakens. The system becomes harder to manage precisely as it becomes more critical. ### 3\. Ratepayer Protection Without System Mechanics The commitment to shield residential ratepayers from rising costs reflects a political reality. But the framework does not resolve the underlying system dynamics required to achieve it. As large, high-value loads migrate toward private supply, the shared cost base erodes. Transmission, distribution, and legacy generation costs remain. They do not vanish when demand leaves. Over time, the gap must close. Through rates, through reliability, or through policy intervention. The promise holds at the level of rhetoric. The system behaves according to physics and economics. ### 4\. Sovereignty Shifts from Public Systems to Private Infrastructure Energy infrastructure has historically existed within a framework of public oversight, even when privately owned. The expansion of AI-linked, privately controlled generation introduces a different model. Strategic assets begin to operate with reduced regulatory visibility, guided by internal optimization rather than system-wide objectives. This is not simply deregulation. It is a redistribution of control over critical infrastructure. Decisions that shape energy supply, system stability, and regional development migrate toward actors whose incentives are not aligned with public coordination. ### 5\. Environmental Exposure Without Integrated Accounting The framework emphasizes speed and scale, but does not integrate environmental accounting into its infrastructure model. Behind-the-meter generation can expand rapidly, particularly where it offers cost or reliability advantages. Yet these systems may not be fully captured within existing planning or reporting frameworks. The result is a widening gap between policy intent and system reality. Emissions, water use, and land impacts accumulate in ways that are difficult to measure and even harder to govern after the fact. ### 6\. A New Class of Load Without a Model to Understand It Artificial intelligence introduces a demand profile that does not resemble anything the grid has managed before. It is concentrated, dynamic, and driven by computational cycles rather than human behavior. Training runs, inference bursts, and rapid scaling events create patterns that are invisible to traditional planning models. Yet the framework treats demand as a question of volume, not behavior. Capacity can be added. Behavior must be understood. Without that understanding, the system operates with a blind spot at its center. ### 7\. Reliability as an Emergent Property, Not a Managed Outcome Grid reliability depends on coordination. Generation, transmission, and load must operate in balance, continuously and precisely. As more infrastructure moves outside coordinated frameworks, optimizing locally rather than systemically, that balance becomes harder to maintain. Reliability does not disappear. It becomes unpredictable. It emerges from the interaction of independently optimized systems rather than from centralized planning. That shift introduces a form of risk that is difficult to model and even harder to control. ### 8\. Centralized Authority Without Operational Reach The framework moves to preempt state-level regulation, consolidating authority at the federal level. This resolves one form of fragmentation. It creates another. Authority is centralized. Capability is not. No single institution is equipped to manage the real-time, cross-sector dynamics of AI embedded in infrastructure. Energy regulators oversee markets. Standards bodies define specifications. Security agencies monitor threats. None operate at the level of the integrated system now emerging. The result is a gap between jurisdiction and execution. The system becomes national in scope, but remains fragmented in operation. ### 9\. Markets Designed for a Different Era Electricity markets were built on assumptions that are no longer fully valid: predictable demand, centralized generation, and clear price signals. AI disrupts each of these. Demand becomes flexible, strategic, and in some cases self-supplied. Price signals lose influence when large actors can operate outside them. Investment decisions become harder to align with system needs. Markets do not fail immediately. They drift. And in that drift, inefficiencies accumulate, resilience declines, and the system becomes more difficult to stabilize. ### 10\. National Security Framed at the Wrong Layer The framework acknowledges risks associated with advanced AI models—misuse, strategic competition, and technological leadership. These are real concerns. But they are incomplete. The more immediate vulnerability lies in the infrastructure that supports those models. Concentrated clusters of compute and energy represent high-value targets. Their disruption carries consequences that extend beyond any single application. Security, in this context, is not just about controlling information. It is about ensuring the continuity of systems that have become essential to economic and national function. ## The Deeper Pattern Taken together, these risks point to a single underlying condition. The framework treats artificial intelligence as something to be accelerated within existing system logic. In reality, AI is altering that logic. It is changing how electricity is consumed, how infrastructure is financed, how decisions are made within systems, and how control is distributed across institutions. It is not simply another sector to be enabled. It is a force that reorganizes the system itself. And systems do not reorganize without consequence. ## What Comes Next None of these risks are inevitable. They are the product of sequencing—of building before governing, of scaling before coordinating. The solution is not to slow development or constrain innovation. It is to match the pace of infrastructure expansion with a corresponding evolution in governance. That governance cannot exist solely at the level of policy or regulation. It must operate within the system itself—capable of observing, constraining, and coordinating behavior in real time. It must integrate energy, computation, and decision-making into a coherent framework. What is emerging is not simply a larger grid, nor a more digital economy. It is a new form of infrastructure—one in which electricity and intelligence are no longer separate domains, but components of a single, interdependent system. The frameworks that governed the industrial energy era were designed for physical flows and human-driven demand. They are not sufficient for systems in which computation drives load, algorithms shape behavior, and infrastructure responds dynamically to machine-generated signals. A new layer is required—one that sits above markets and below policy, translating intent into operational constraint. A governance architecture designed not just to regulate outcomes, but to shape system behavior itself. This is the work ahead. In the twentieth century, energy defined the boundaries of economic power. Oil fields, pipelines, and geopolitics determined the structure of the global system. In the twenty-first, that foundation is shifting. Energy is no longer only a fuel. It is the substrate of intelligence. And intelligence, in turn, is beginning to reorganize how energy is produced, consumed, and controlled. The White House framework marks the moment the United States chose to accelerate into that future. But acceleration alone is not strategy. The systems we are building will determine not just economic outcomes, but the distribution of control, the resilience of infrastructure, and the integrity of the social contract that underpins it. The question is no longer whether artificial intelligence will reshape the energy system. It already is. The question is who will define the architecture of that system—and whether it will be governed by design, or left to emerge by consequence. Because the risk is not that artificial intelligence fails. The risk is that it succeeds—faster than the systems designed to guide it. *Brandon N. Owens is the founder of AIxEnergy, a research platform focused on artificial intelligence and energy systems. He has worked across energy markets, infrastructure and policy, including roles at NREL, GE Energy, S&P Global and NYSERDA.* ## Sign up for AIxEnergy AIxEnergy analyzes how artificial intelligence is changing electricity demand, grid operations, and energy infrastructure, with a focus on system risk, policy, and institutional decision-making. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### From Automation to Control: Architecting the Software-Defined Grid URL: https://www.aixenergy.io/from-automation-to-control/ Last updated: 2026-06-30T02:30:24.000Z The grid is shifting from hardware-based control to software-defined systems. AI, electrification, and distributed energy are driving this change, introducing new risks and requiring integrated, adaptive control across sensing, communication, and computation layers. _This post is for paying subscribers only._ ### The Shadow Grid: AI’s Hidden Energy Network and the Crisis of Infrastructure Visibility URL: https://www.aixenergy.io/the-shadow-grid-ais-hidden-energy-network-and-the-crisis-of-infrastructure-visibility/ Last updated: 2026-06-30T02:30:12.000Z AI data centers are driving a surge in electricity demand and spawning a “shadow grid” of private power plants. Built outside traditional planning, this hidden infrastructure erodes visibility over the energy system and creates new challenges for governance and reliability. _This post is for paying subscribers only._ ### The Architecture Problem Holding Back AI on the Electric Grid URL: https://www.aixenergy.io/the-architecture-problem-holding-back-ai-on-the-electric-grid/ Last updated: 2026-06-30T02:30:03.000Z AI is spreading across electric utilities, but most projects stall. The problem isn’t the algorithms—it’s the data. Grid systems operate on different clocks, from milliseconds to days. Until utilities build time-synchronized data architectures, AI will remain experimental rather than operational. _This post is for paying subscribers only._ ### A Blueprint for Enterprise AI Transformation in the Energy Sector URL: https://www.aixenergy.io/blueprint/ Last updated: 2026-06-30T02:29:54.000Z AI is reshaping energy companies by embedding prediction into infrastructure operations. A five-part blueprint guides transformation: identify where AI creates value, build unified data architecture, manage AI as a portfolio, govern model risk, and develop strategic technology ecosystems. _This post is for paying subscribers only._ ### Artificial Intelligence, Electric Infrastructure, and the Question of Who Pays URL: https://www.aixenergy.io/shadowgrid_whitehouse/ Last updated: 2026-06-30T02:29:43.000Z AI companies pledged at the White House to finance power for data centers to protect ratepayers. But AI’s massive electricity demand is reshaping grid architecture—driving new generation, transmission, and private “shadow grid” systems that could shift energy infrastructure costs across the economy. _This post is for paying subscribers only._ ### How Regulatory Seams Are Reshaping American Power URL: https://www.aixenergy.io/shadowgridseams/ Last updated: 2026-07-29T19:38:07.000Z State and federal policies are opening regulatory seams that let large AI data centers build off-grid power, often gas-fueled, outside traditional utility oversight. Clean-energy mandates bind the grid, not private systems—creating a parallel “shadow grid” with emissions and ratepayer risk. _This post is for subscribers only._ ### America’s Record Clean Energy Buildout — and the Parallel System Emerging Beside It URL: https://www.aixenergy.io/shadowgrid/ Last updated: 2026-06-30T02:29:23.000Z The U.S. plans to add 86 GW of utility-scale power in 2026—the largest annual expansion in decades—led by solar and battery storage. But a parallel wave of privately built, gas-fired generation for AI data centers is emerging beyond traditional grid reporting. _This post is for paying subscribers only._ ### Energy Innovation at the Edge of Power: A Review of the IEA’s The State of Energy Innovation 2026 URL: https://www.aixenergy.io/energyinnovation/ Last updated: 2026-07-10T17:30:49.000Z When the [International Energy Agency](https://www.iea.org/?ref=aixenergy.io) released [The State of Energy Innovation 2026](https://www.iea.org/reports/the-state-of-energy-innovation-2026?ref=aixenergy.io) on February 17, 2026, it did more than publish a report. It issued a strategic barometer for the global energy system at a moment of structural inflection. The atmosphere surrounding energy is no longer defined solely by decarbonization targets or emissions trajectories. It is charged by geopolitical fracture, industrial policy resurgence, supply chain recalibration, electrification at scale, and the accelerating physics of climate change. The IEA’s assessment makes it unmistakably clear that innovation now occupies the center of this convergence. Author's Summary 0:00 /46.479297 1× The report’s most consequential contribution is not a single statistic, but a reframing. Energy innovation is no longer presented primarily as an environmental obligation. It is positioned as a pillar of economic competitiveness and national security. In the IEA’s survey of practitioners and experts, 80 percent ranked energy security among the top three drivers of innovation in 2025, ahead of affordability and greenhouse gas mitigation.¹ That hierarchy speaks volumes. It reflects a world that has absorbed the lessons of fuel price volatility, fragile supply chains, and concentrated critical mineral markets. Clean energy, in this context, is not merely about carbon. It is about control, continuity, and sovereignty. The scale of the transformation is extraordinary. Global markets for energy technologies such as batteries, turbines, motors, and transformers are measured in trillions of dollars.² Energy expenditures represent a significant share of global economic output. One in ten patents now relates to energy technologies, a higher proportion than in chemicals, pharmaceuticals, or transport.³ Energy has become innovation-intensive in a manner once reserved for semiconductors at the dawn of the digital era or aerospace during the space race. The laboratory, the manufacturing line, and the grid control room now share a common frontier. As an energy technology historian, I read these figures with a long memory. A century ago, energy innovation advanced in isolated workshops and test fields, often propelled by individuals working against economic headwinds. Today it unfolds within coordinated national strategies, venture capital ecosystems, public research institutions, and multinational supply chains. The IEA documents more than 150 significant innovation highlights in 2025 and 50 upgrades in technology readiness levels among tracked emerging technologies.⁴ The ecosystem is not dormant. It is kinetic. Yet innovation does not self-organize at scale. The report underscores the decisive role of public investment in catalyzing long-term technological transformation. Retrospective evaluations of United States energy R&D programs show benefits at least three times greater than their costs, and in some cases several hundreds of dollars in benefits for every dollar invested.⁵ These returns are not abstract. They include fuel savings, equipment cost reductions, expanded markets, and technological spillovers that compound across decades. Lithium-ion batteries illustrate the pattern. Their lineage traces to publicly funded research in the 1970s, including early government-supported patent activity, long before they powered electric vehicles and grid storage systems.⁶ Floating liquefied natural gas projects benefited from early public risk-sharing before achieving commercial scale.⁷ Next-generation geothermal technologies depended for decades on government-backed research when private capital judged the risks prohibitive.⁸ Markets harvest what public policy plants. For public institutions, the implications are profound. Energy innovation can no longer be compartmentalized within environmental agencies or narrow R&D portfolios. Industrial strategy and climate strategy have fused. Ministries of energy, finance, trade, defense, and education must coordinate around shared objectives. Long-term funding frameworks must be insulated from short-term political cycles. Demonstration programs must be paired with procurement mechanisms that create credible and durable demand signals. Durability, not spectacle, defines successful innovation systems. The IEA documents more than 80 new energy innovation policies introduced across 32 jurisdictions in 2025 alone.⁹ Policy dynamism reflects engagement, yet volatility undermines capital formation. Investors require predictability. Innovators require transparent regulatory pathways. A boom-and-bust cycle in clean energy erodes confidence more swiftly than technological setbacks. Regulators occupy a pivotal position in this transition. The twentieth-century electricity system was built on centralized generation, vertically integrated monopolies, and cost-of-service oversight. The twenty-first century grid is distributed, digital, and data-rich. The IEA’s dedicated focus on innovation for electricity grid resilience underscores the magnitude of this evolution.¹⁰ Advanced inverters, grid-forming technologies, long-duration storage, real-time monitoring systems, and cybersecurity architectures are not peripheral upgrades. They are foundational. Electrification without resilience is fragility in disguise. As transport and buildings electrify, system complexity intensifies. Regulators must therefore expand their analytical frameworks beyond static rate impacts to incorporate dynamic efficiency, resilience valuation, and long-term system optimization. Performance-based regulation, regulatory sandboxes, and structured pilot programs can accelerate learning while preserving consumer protections. The regulator of this century is not merely a referee. It is a system architect. For private sector institutions, the message is equally unambiguous. Innovation is not episodic. It is structural. Corporate R&D strategies must integrate policy intelligence, supply chain diversification, and geopolitical risk management. Firms that treat regulation as an external constraint will lag those that engage constructively with policymakers and research institutions. Alignment with long-term transition trajectories is not optional. It is a strategic necessity. The IEA’s analysis of revealed technology advantages offers both caution and encouragement. Countries that invested early in wind power now demonstrate strong specialization in wind patenting. Fossil fuel producers exhibit specialization in fossil technologies.¹¹ Innovation strengths and industrial strengths reinforce one another, yet technical specialization does not automatically translate into manufacturing dominance.¹² Workforce development, logistics infrastructure, trade strategy, and capital access determine whether intellectual property becomes industrial leadership. Public institutions must therefore cultivate full-stack ecosystems: from university laboratories to apprenticeship pipelines, from demonstration facilities to export financing. Innovation policy without industrial policy is incomplete. Industrial policy without sustained innovation becomes brittle. The report’s examination of fusion energy as a frontier domain reinforces the stakes. Structured “Races to First” in breakthrough technologies introduce transparency and accountability to ambition.¹³ Fusion may or may not achieve grid-connected scale within the next generation, but its resurgence signals renewed willingness to invest in bold experimentation. Energy history counsels humility. It also rewards persistence. Beneath the report’s data lies a deeper transformation. Digitalization, artificial intelligence, advanced materials science, and computational modeling are converging with turbines, batteries, and transmission systems. The boundary between energy and information technology is dissolving. The next gains in productivity may emerge as much from algorithms as from alloys. Innovation policy must recognize and accelerate this synthesis. The State of Energy Innovation 2026 is therefore more than a statistical compendium. It is a strategic mirror. It reflects a world in which energy innovation has become an arena of economic competition and collaboration. It challenges policy makers to think in decades while acting in quarters. It reminds regulators that prudence and progress must coexist. It signals to private capital that disciplined alignment with structural transition pathways will define competitive advantage. History offers perspective. Coal once appeared inexhaustible, underwriting empires that mistook abundance for permanence. Oil reshaped the twentieth century, delivering prosperity while entangling nations in dependencies they did not fully anticipate. Electricity became so ubiquitous that its fragility faded from memory—until crisis revealed the thinness of resilience beneath reliability. Each energy era rewarded institutions capable of sustained innovation. Not those that merely possessed resources, but those that cultivated systems: research systems, financial systems, regulatory systems, educational systems. They understood that energy is not only extracted from the ground. It is engineered through governance. The defining resource of the twenty-first century will not be a seam of coal or a reservoir of oil. It will be institutional capacity for innovation. The ability to align public capital with private ingenuity. The discipline to embed resilience in digitalized grids. The foresight to integrate artificial intelligence with heavy industry. The patience to think in decades while acting in quarters. The IEA has illuminated the terrain. The race is not theoretical. It is underway. And as in every previous energy transformation, the outcome will belong to those who build institutions strong enough to carry innovation across generations. ## References International Energy Agency. [The State of Energy Innovation 2026](https://www.iea.org/reports/the-state-of-energy-innovation-2026?ref=aixenergy.io). Paris: IEA, 2026. Notes 1. International Energy Agency (IEA), The State of Energy Innovation 2026 (Paris: IEA, 2026), Executive Summary, 7. 2. IEA, State of Energy Innovation 2026, 7. 3. IEA, State of Energy Innovation 2026, 7. 4. IEA, State of Energy Innovation 2026, 10. 5. IEA, State of Energy Innovation 2026, 8. 6. IEA, State of Energy Innovation 2026, 8. 7. IEA, State of Energy Innovation 2026, 7–8. 8. IEA, State of Energy Innovation 2026, 8. 9. IEA, State of Energy Innovation 2026, 10. 10. IEA, State of Energy Innovation 2026, Table of Contents, Chapter 6. 11. IEA, State of Energy Innovation 2026, 9. 12. IEA, State of Energy Innovation 2026, 9. 13. IEA, State of Energy Innovation 2026, 10. ### Extending AI Governance into Grid Governance: A Review of the OECD AI Report from an Infrastructure Systems Perspective URL: https://www.aixenergy.io/extending-ai-governance-into-grid-governance-a-review-of-the-oecd-ai-report-from-an-infrastructure-systems-perspective/ Last updated: 2026-07-10T17:30:59.000Z The November 2025 [OECD](https://www.oecd.org/en.html?ref=aixenergy.io) AI report, [Assessing Potential Future Artificial Intelligence Risks, Benefits and Policy Imperatives](https://www.oecd.org/content/dam/oecd/en/publications/reports/2024/11/assessing-potential-future-artificial-intelligence-risks-benefits-and-policy-imperatives%5F8a491447/3f4e3dfb-en.pdf?ref=aixenergy.io), is a serious and structured intervention in the global AI governance debate. It defines risk families. It articulates benefit domains. It proposes policy levers—disclosure regimes, lifecycle oversight, concentration safeguards, and compute-based regulatory triggers. The report is comprehensive in scope and disciplined in architecture. As a framework for governing advanced AI systems, it is analytically coherent and forward-looking. Author's Summary 0:00 /77.728027 1× But it is designed for a platform world. Energy appears in the report in two ways. First, as an application domain—AI enabling real-time grid management and digital twins for optimization (p. 16). Second, as a critical infrastructure sector vulnerable to AI-enabled failure or cyber harm (pp. 22–23). These references acknowledge the intersection of AI and power systems. What they do not do is extend the governance logic into the operational mechanics of electricity: frequency stability, reserve margins, ramp rates, contingency performance, transmission constraints, or cost causation. That distinction matters. The OECD framework is built around model capability and systemic exposure within digital ecosystems. Its primary concerns are information asymmetry, misuse risk, concentration of compute power, and lifecycle monitoring. Risk is framed as a function of scale, capability, and control. Electric grids operate under a different set of constraints: - Physics governs outcomes. - Disturbances propagate in seconds. - Reliability standards are enforceable. - Cost impacts flow directly to ratepayers. When AI is embedded into grid operations—forecasting engines, dispatch optimization tools, control systems—it becomes part of a reliability regime. When AI scales as hyperscale data center demand, it becomes concentrated electrical load measured in megawatts. At that point, governance cannot remain at the level of abstract capability. It must move to the level of physical consequence. This review has two objectives. First, to assess the OECD report on its own terms—examining how it frames energy as a use case, a critical system, and a concentration dynamic. Second, to extend that framework into infrastructure governance. That means translating compute thresholds into megawatt thresholds. Converting disclosure into operational telemetry. Reframing lifecycle oversight as reliability assurance. Mapping concentration of compute into concentration of electrical load and system leverage. The central thesis is straightforward: AI governance must evolve from model-centric oversight to consequence-centric infrastructure management. The OECD report provides a foundation. It does not yet close the loop between digital capability and physical system exposure. As AI becomes embedded in both control logic and concentrated demand, that loop must be closed. When compute becomes megawatts, governance must follow the physics. ## Framing: Platform Governance vs. Infrastructure Governance The OECD report is structured around risk families, benefit domains, and policy actions. Its orientation is primarily platform-centric. It focuses on model capability, information asymmetry, systemic misuse, concentration of power, and lifecycle oversight. Electric grids operate under different constraints: - Physics is non-negotiable. - Disturbances propagate at sub-second timescales. - Reliability standards are enforceable. - Cost allocation affects ratepayers. When AI enters grid operations and becomes a large electrical load, governance must operate at the level of physical consequence, not abstract capability. ## Energy as a Use Case The report references AI applications in "real-time energy grid management" and digital twins for forecasting and optimization (p. 16). This framing recognizes that energy systems are becoming AI-mediated decision environments. However, the analysis remains at the optimization layer. It does not address: - Frequency response obligations - Voltage stability constraints - Reserve margin requirements - Ramping dynamics - Dispatch coordination under volatility Optimization assumes constraints. The report does not describe those constraints. AI does not merely optimize grid behavior, it changes decision velocity and load characteristics. Governance must account for that shift. ## Critical Systems Risk The OECD identifies integration of AI into complex critical systems as a source of cascading failure risk (pp. 22–23). It also names energy infrastructure as a plausible target of AI-enabled cyber activity (p. 23). This section is directly applicable to modern power systems. Three layers are converging: 1. AI embedded in grid operations and forecasting. 2. AI embedded in market trading and dispatch analytics. 3. AI operating as hyperscale electrical load. Failure in any layer can propagate across the others. The report correctly signals scaling risk. It does not translate that risk into reliability metrics or operational thresholds. For infrastructure systems, the relevant question is not whether an AI system is advanced. The relevant question is whether it interacts with high-consequence physical systems. ## Power Concentration and Compute Thresholds The report identifies "power concentration" as a structural risk linked to compute access and cost (p. 18). It further references compute thresholds as potential regulatory triggers (p. 26). This framing is important but incomplete from a grid perspective. Compute concentration is also load concentration. Advanced AI facilities cluster where firm, low-cost electricity is available and deliverable. The OECD discusses compute thresholds in terms of model capability oversight. It does not connect compute thresholds to physical-system thresholds such as: - Megawatt scale - Ramp rate variability - Transmission deliverability limits - Operating reserve dependency Without that translation, compute governance remains abstract. Infrastructure governance requires mapping compute intensity to electrical impact. ## Disclosure Frameworks Policy Action 3 emphasizes disclosure, including standardized documentation such as model cards and datasheets (pp. 14–15). The purpose is to reduce information asymmetry and improve accountability. This is the strongest bridge between OECD governance and grid governance. If AI systems require model cards, then high-impact AI loads and AI-operated grid tools require operational disclosure standards. A "Power Card" would describe system-relevant behavior rather than proprietary architecture. It would include: - Maximum and typical load envelope - Ramp characteristics (upward and downward) - Curtailment capability and response time - Backup generation behavior - Telemetry availability and latency - Failover modes and restoration sequences - Operational coordination protocols Such disclosure does not expose intellectual property. It communicates behavior that affects shared infrastructure. The OECD disclosure principle supports this extension. ## Lifecycle Risk Management — Operational Assurance The report emphasizes lifecycle monitoring, incident response, and corrective deployment actions (pp. 28–29). This aligns with utility-grade reliability governance. In electric systems, assurance is continuous. Systems are logged, audited, and subject to post-event analysis. AI-mediated operational decisions in critical infrastructure should meet similar standards: - Decision logging - Event-triggered reporting - Audit traceability - Rollback capability - Structured post-incident review The OECD framework legitimizes lifecycle oversight. Infrastructure systems require that oversight to operate at sub-hourly resolution. ## Data Centers as Critical National Infrastructure The report notes that the United Kingdom has classified data centers as critical national infrastructure (p. 27). This classification shift is consequential. It implies: - Enhanced cyber protection standards - Elevated reporting obligations - Potential resilience and operational requirements If data centers are critical infrastructure, then their electrical behavior becomes a reliability matter, not only a commercial one. ## Environmental Burden — Acknowledged but Unspecified The OECD notes that policymakers are considering the environmental burden of AI systems (p. 27). The discussion remains high level. It does not address: - Nodal energy intensity - Water intensity under thermal constraints - Marginal emissions under peak dispatch - Sub-hourly volatility impacts Annual energy totals are insufficient. Infrastructure impact occurs at operational timescales. ## Structural Gaps Identified Four structural gaps remain: ### Absence of Load Dynamics Analysis There is no sustained treatment of load volatility, ramping, or dispatch interaction. ### No Reliability Metric Translation The report does not connect AI governance to reserve margins, contingency response, or disturbance management. ### Compute Thresholds Not Linked to MW Thresholds Governance triggers are not mapped to physical-system triggers. ### No Engagement with Cost Allocation There is no analysis of tariff design, cost causation, or system cost recovery. ## Extension Framework The OECD report provides governance scaffolding. The next step is translation into infrastructure standards with these five extensions: 1. A Grid-Relevant Disclosure Standard (Power Card). 2. Sub-hourly Load Volatility Metrics tied to reserve cost. 3. Incident Reporting Standards for AI-mediated operational tools. 4. Compute-to-Megawatt mapping thresholds for regulatory triggers. 5. Environmental intensity reporting at operational resolution. These extensions do not contradict OECD recommendations. They operationalize them for electricity systems. ## Conclusion The OECD AI report provides a rigorous and well-structured governance architecture for advanced artificial intelligence systems. It identifies systemic risk categories, concentration dynamics, disclosure principles, and lifecycle oversight mechanisms. As a platform-governance document, it is disciplined, forward-looking, and analytically sound. But AI is no longer confined to platforms. It is embedded in control systems. It is embedded in markets. And it is embedded in megawatt-scale physical demand. The report stops at the capability layer. It does not extend its framework into the operational language of power systems—frequency stability, ramp rates, reserve margins, contingency performance, transmission constraints, or cost causation. It does not translate compute thresholds into physical-system thresholds. It does not convert lifecycle oversight into reliability assurance. It does not map concentration of compute into concentration of electrical load. That translation is essential. As AI becomes both decision logic and concentrated load, governance must move from abstract model risk to measurable infrastructure exposure. It must answer not only whether systems are powerful or aligned, but whether they compress operating margins, alter dispatch dynamics, increase reserve requirements, or shift cost burdens to ratepayers. The OECD framework should therefore be treated as a foundation—not a finished structure. Its taxonomy can support infrastructure governance. Its disclosure logic can evolve into operational transparency standards. Its lifecycle principles can anchor grid-grade audit and rollback protocols. Its concentration analysis can be extended into energy-system leverage and reliability exposure. The next phase of AI governance is not about models alone. It is about systems of consequence. When compute becomes megawatts, governance must follow the physics. Download the full OECD report here [Assessing Potential Future Artificial Intelligence Risks, Benefits and Policy Imperatives](https://www.oecd.org/content/dam/oecd/en/publications/reports/2024/11/assessing-potential-future-artificial-intelligence-risks-benefits-and-policy-imperatives%5F8a491447/3f4e3dfb-en.pdf?ref=aixenergy.io). ### The Grid’s New Blind Spot: Why Demand Volatility May Be the Next Frontier in Rate Design URL: https://www.aixenergy.io/the-grids-new-blind-spot-why-demand-volatility-may-be-the-next-frontier-in-rate-design/ Last updated: 2026-06-30T02:28:43.000Z AI-driven digital loads are reshaping not just how much electricity the grid uses, but how it behaves. Traditional tariffs price magnitude and timing, not volatility. Aligning rates with demand behavior—without mandates—may be the next quiet evolution in cost-of-service regulation. _This post is for paying subscribers only._ ### The IEA’s Electricity 2026 Meets AI: What the World’s Top Power Outlook Says—And What It Doesn’t URL: https://www.aixenergy.io/electricity2026/ Last updated: 2026-07-15T02:40:45.000Z IEA’s Electricity 2026 isn’t an AI report—but AI runs through it. Data centers drive demand, grids become the bottleneck, queues grow, and reliability risks rise. AI isn’t just adding load—it’s forcing new rules for how the grid is governed. _This post is for paying subscribers only._ ### When AI Agents Talk to Each Other: What the Electric Power Grid Reveals About the Next Phase of Intelligence URL: https://www.aixenergy.io/when-ai-agents-talk-to-each-other-what-the-electric-power-grid-reveals-about-the-next-phase-of-intelligence/ Last updated: 2026-06-30T02:28:25.000Z As AI shifts from single models to interacting agents, intelligence becomes a coordination problem. The electric grid—bound by physics, reliability, and accountability—reveals why agent-to-agent reasoning must be structured, legible, and constrained long before deployment. _This post is for paying subscribers only._ ### Artificial Intelligence Has Entered the Grid: What U.S. Utilities are Doing with AI—and Why Scale Remains Constrained URL: https://www.aixenergy.io/artificial-intelligence-has-entered-the-grid-quietly-what-u-s-utilities-are-actually-doing-with-ai-why-most-deployments-remain-constrained-and-what-that-reveals-about-the-future-of-crit/ Last updated: 2026-06-30T02:28:18.000Z U.S. utilities are adopting AI cautiously—piloting tools for forecasting, reliability, and large-load management while avoiding autonomous control. The constraint isn’t technology, but governance, data, regulation, and risk tolerance. _This post is for paying subscribers only._ ### From Paradox to Progress — A Review of WEF’s Net-Positive AI Energy Framework URL: https://www.aixenergy.io/from-paradox-to-progress-a-review-of-wefs-net-positive-ai-energy-framework/ Last updated: 2026-07-10T17:35:00.000Z The [WEF-Accenture report](https://www.weforum.org/publications/from-paradox-to-progress-a-net-positive-ai-energy-framework/?ref=aixenergy.io) is correct in its opening premise: artificial intelligence has crossed a threshold at which its energy implications are no longer secondary effects. AI-driven compute growth, hyperscale data centers, and the upstream supply chains for advanced chips are now influencing electricity demand, infrastructure investment, and grid planning at national and regional scales.² The report’s projection that global data-center electricity demand could exceed 1,200 TWh by 2035—nearly triple 2024 levels—captures the scale of the challenge.³ Author's Summary 0:00 /125.240317 1× What makes the report especially timely is its alignment with current events. In the United States, the Energy Information Administration projects record electricity consumption in 2026 and 2027, explicitly citing data centers and AI workloads as key drivers.⁴ At the same time, grid operators and regulators are grappling with transformer shortages, interconnection backlogs, and novel load configurations, including co-located generation and behind-the-meter arrangements.⁵ These pressures are now shaping tariff design, planning assumptions, and political debates over affordability. Internationally, similar dynamics are unfolding. Ireland’s experience—where data centers have become a material share of national electricity demand—has pushed regulators to tighten connection rules and require stronger system contributions from large loads.⁶ These cases underscore a central insight of the WEF framework: AI growth is inevitable, but unmanaged AI growth can undermine energy security, public acceptance, and climate goals. ## **Strengths of the WEF–Accenture Approach** 💡 WEF’s net-positive AI energy framework moves AI from aspiration to accountability, defining measurable lifecycle outcomes. Its focus on efficiency by design, impact-driven deployment, and demand shaping reflects mature energy economics and practical decision-making grounded in real system constraints. The report’s definition of “net-positive AI energy”—a condition in which the energy and resource savings enabled by AI exceed its full lifecycle consumption—is a major contribution.⁷ It moves the debate from abstract claims about AI’s potential to a measurable standard that can, in principle, be audited and compared across sectors and jurisdictions. This framing aligns well with the academic literature calling for lifecycle-based disclosure of AI energy and emissions impacts. Research in machine learning and natural language processing has demonstrated that model choice, training regime, inference patterns, and data-center characteristics can dramatically alter energy use and carbon intensity.⁸ By anchoring the discussion in lifecycle accounting, WEF situates AI within the same analytical tradition long used in energy policy and environmental assessment. The three action drivers—*Design for efficiency*, *Deploy for impact*, and *Shape demand wisely*—are thoughtfully chosen and mutually reinforcing. Each corresponds to a well-documented failure mode in current AI deployment trajectories. *Design for efficiency* recognizes that AI’s energy footprint is not fixed. Advances in hardware efficiency, cooling systems, model architectures, and data-center design can yield large gains.⁹ WEF’s emphasis on embedding efficiency at the design stage is consistent with evidence showing that early technical choices dominate long-run energy outcomes. *Deploy for impact* shifts attention from raw capability to contextual value. AI applications that reduce losses, improve forecasting, or optimize industrial and grid operations can deliver genuine system benefits, especially under constrained conditions. This focus aligns with energy-economics research showing that the value of efficiency depends on where and when it occurs. *Shape demand wisely* is particularly important. By explicitly addressing rebound effects, unconscious consumption, and “dark data,” the report avoids a common pitfall of efficiency narratives.¹⁰ The historical literature—from Jevons onward—demonstrates that efficiency gains alone do not guarantee reduced resource use.¹¹ WEF’s inclusion of demand shaping signals a mature understanding of energy-system dynamics. The report’s strategic enablers—consumer education and workforce upskilling, ecosystem collaboration, and transparent measurement and accountability—are pragmatic and well grounded. Energy transitions are not achieved through technology alone; they depend on human capacity, institutional trust, and shared standards. Accenture’s influence is evident here in a positive sense. The report is oriented toward decision-makers who must act under real constraints, balancing cost, reliability, and sustainability. The use of over 130 real-world examples reinforces the framework’s credibility and applicability. ## Underdeveloped Areas 💡 AI load growth is outpacing grid build cycles, creating near-term stress even when deployments are net-positive over time. Clearer rules on timing, system boundaries, affordability, and accountability are needed so efficiency gains don’t undermine reliability, equity, or public trust. One underdeveloped tension in the framework concerns timing. AI-driven load growth is arriving faster than generation, transmission, and distribution investments can be planned and built. Even with aggressive clean-energy deployment, infrastructure lead times create periods in which marginal demand may rely on existing fossil assets or exacerbate congestion.¹² The report acknowledges this constraint, but future work could go further by examining how net-positive criteria should be applied during transitional periods. An AI deployment that is net-positive over its lifecycle may still impose short-term system costs that require explicit mitigation or compensation. WEF is right to emphasize transparent measurement, but the complexity of system boundaries deserves additional attention. Net-positive claims depend on assumptions about geographic scope, temporal resolution, and upstream impacts. Does the accounting include embodied energy in semiconductors? How are regional grid emissions factors treated? How are water and land impacts incorporated alongside electricity use? The academic literature has repeatedly highlighted these challenges. Patterson et al. show that location and processor choice can swing emissions outcomes dramatically, while Masanet et al. demonstrate how global data-center energy estimates can vary based on methodological choices.¹³ Strengthening standardized disclosure requirements would enhance the framework’s robustness. Another area for expansion is distributional impact. The report correctly notes the risk that AI growth could raise electricity prices or concentrate benefits in energy-rich regions.¹⁴ These are not merely side effects; they shape political feasibility. Recent reporting on large technology firms’ commitments to fund grid upgrades reflects an emerging recognition that social license matters.¹⁵ Future iterations of the framework could more explicitly address how costs and benefits are allocated among customer classes, regions, and communities, and how net-positive outcomes are reconciled with affordability objectives. Governance, as discussed in [*The Cognitive Grid*](https://www.amazon.com/Cognitive-Grid-Artificial-Intelligence-Infrastructure/dp/B0GDP8P1GD?ref=aixenergy.io), is best understood not as an abstract principle but as the practical question of how decisions are authorized, executed, and reviewed in public-interest systems. AI systems that inform prioritization, planning, or real-time operations can influence outcomes even when they remain “advisory.” [The Cognitive Grid BookThe Cognitive Grid warns judgment is shifting into automation, creating a latency gap where machine-speed action outruns oversight. It proposes “constitutional” execution-time governance—separating proposal from authorization via constraints, permissions, proofs, audits, isolation.![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/icon/Icon1-149.png)AIxEnergyBrandon Owens![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/thumbnail/Cover5_upscayl_4x_upscayl-standard-4x-1.png)](https://www.aixenergy.io/cognitivegrid/) The WEF framework gestures toward accountability, but additional specificity—around decision rights, escalation pathways, and traceability—would strengthen its applicability in highly regulated environments such as electric grids. This is less a critique than an invitation: WEF is well-positioned to convene the actors needed to translate high-level principles into operational governance templates. The framework reflects a clear understanding of how enterprises actually make decisions: through capital-allocation processes, performance metrics, and operational incentives. Without this grounding, calls for net-positive AI energy would risk remaining aspirational. At the same time, enterprise pragmatism must be complemented by public-sector authority. Utilities and regulators operate under statutory obligations to ensure reliability and fairness, not merely efficiency. The report would benefit from drawing this distinction more explicitly—not to diminish enterprise action, but to clarify where voluntary leadership ends and binding rules begin. ## A Foward-Looking Research Agenda 💡 Next steps for the WEF–Accenture framework include applying net-positive metrics at the grid level, accounting for infrastructure timing, standardizing AI energy disclosures, addressing affordability and equity, and clarifying governance for AI-influenced infrastructure decisions. Building on the strong foundation laid by WEF and Accenture, several areas merit deeper exploration: 1\. Operationalizing net-positive metrics at the grid and jurisdictional level, not only within enterprises. 2\. Integrating infrastructure timing into net-positive assessments, distinguishing short-term system stress from long-term benefits. 3\. Developing standardized disclosure templates for AI energy use, emissions, and resource impacts. 4\. Addressing distributional effects explicitly, including affordability and regional equity. 5\. Clarifying governance pathways for AI systems that influence consequential infrastructure decisions. These steps represent natural extensions of its logic. ## Conclusion 💡 **From Paradox to Progress* reframes AI and energy in practical, institutional terms. By creating a shared language around systems and constraints, it advances serious action. Further refinement on timing, measurement, equity, and governance would sharpen—not dilute—its impact. *From Paradox to Progress* is a constructive and timely contribution. It reframes the AI–energy relationship in terms that energy institutions understand: systems, constraints, trade-offs, and accountability. It avoids simplistic optimism, acknowledges real risks, and proposes a coherent structure for action. The framework’s greatest value may be that it creates a shared language through which enterprises, utilities, regulators, and civil society can engage. Strengthening its treatment of infrastructure timing, measurement boundaries, economic distribution, and institutional governance would not weaken that language—it would make it more precise. Progress, in energy systems, is rarely linear. But frameworks like this one matter because they shape how problems are named and which solutions are considered legitimate. In that respect, WEF and Accenture have provided a foundation that deserves both recognition and continued refinement. ## Sign up for AIxEnergy AIxEnergy analyzes how artificial intelligence is changing electricity demand, grid operations, and energy infrastructure, with a focus on system risk, policy, and institutional decision-making. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## Notes 1. Brandon N. Owens, *The Cognitive Grid: Artificial Intelligence and the Governance of Delegated Power in Critical Infrastructure* (New York: Independently published, 2025). 2. World Economic Forum, *From Paradox to Progress: A Net-Positive AI Energy Framework* (Geneva: World Economic Forum, December 2025), executive summary. 3. Ibid., 6. 4. Reuters, “US power use to beat record highs in 2026 and 2027, EIA says,” January 13, 2026. 5. World Economic Forum, *From Paradox to Progress*, 6. 6. European Commission and Irish regulatory reporting on data-center electricity demand and connection policy. 7. World Economic Forum, *From Paradox to Progress*, 3–4. 8. Emma Strubell, Ananya Ganesh, and Andrew McCallum, “Energy and Policy Considerations for Deep Learning in NLP,” *Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics* (2019). 9. World Economic Forum, *From Paradox to Progress*, 13–14. 10. Ibid., executive summary. 11. William Stanley Jevons, *The Coal Question* (London: Macmillan, 1865); Harry D. Saunders, “A View from the Macro Side: Rebound, Backfire, and Khazzoom–Brookes,” *Energy Policy* 28, no. 6–7 (2000): 439–449. 12. International Energy Agency, *Energy and AI* (Paris: IEA, 2025). 13. David Patterson et al., “Carbon Emissions and Large Neural Network Training,” arXiv:2104.10350 (2021); Eric Masanet et al., “Recalibrating global data center energy-use estimates,” *Science* 367, no. 6481 (2020): 984–986. 14. World Economic Forum, *From Paradox to Progress*, 6. 15. Financial Times, “Microsoft vows to ‘pay its way’ as it seeks to defuse data centre backlash,” January 2026. ## Bibliography European Commission. “In focus: Data centres – an energy-hungry challenge.” November 17, 2025. Financial Times. “Microsoft vows to ‘pay its way’ as it seeks to defuse data centre backlash.” January 2026. International Energy Agency. *Energy and AI*. Paris: IEA, 2025. Jevons, William Stanley. *The Coal Question*. London: Macmillan, 1865. Masanet, Eric, et al. “Recalibrating global data center energy-use estimates.” *Science* 367, no. 6481 (2020): 984–986. Owens, Brandon N. *The Cognitive Grid: Artificial Intelligence and the Governance of Delegated Power in Critical Infrastructure*. New York: Independently published, 2025. Patterson, David, et al. “Carbon Emissions and Large Neural Network Training.” arXiv:2104.10350 (2021). Reuters. “US power use to beat record highs in 2026 and 2027, EIA says.” January 13, 2026. Saunders, Harry D. “A View from the Macro Side: Rebound, Backfire, and Khazzoom–Brookes.” *Energy Policy* 28, no. 6–7 (2000): 439–449. Strubell, Emma, Ananya Ganesh, and Andrew McCallum. “Energy and Policy Considerations for Deep Learning in NLP.” In *Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics*. 2019. World Economic Forum. *From Paradox to Progress: A Net-Positive AI Energy Framework*. Geneva: World Economic Forum, December 2025. ### The Year the Grid Met AI: 2025 and the Collision Between Exponential Compute and Linear Infrastructure URL: https://www.aixenergy.io/the-year-the-grid-met-ai-2025-and-the-collision-between-exponential-compute-and-linear-infrastructure/ Last updated: 2026-07-13T01:39:46.000Z Artificial intelligence did not arrive at the grid as a marginal new end‑use. It arrived as a structural shock—fast, concentrated, capital‑intensive, and deeply entangled with decisions made far outside traditional utility planning. In 2025, utilities across the United States confronted a form of load growth that behaved unlike anything in their historical data. This was not demand creeping in at the edges. It was infrastructure arriving all at once. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2026/01/10Lessons_upscayl_2x_upscayl-standard-4x.png) What unfolded over the year was not a series of tactical adjustments, but a set of hard lessons about how electricity systems, institutions, and regulatory frameworks respond when computation becomes a primary driver of demand. Taken together, these lessons clarify why generative AI load cannot be treated as a forecast update. It is a reordering force. 1. The first lesson was immediate and uncomfortable: **forecasting models broke**. Traditional demand forecasting is built on gradual adoption curves, stable elasticities, and sectoral proxies that assume continuity. Gen‑AI violated every assumption. Hyperscale data centers moved from site selection to interconnection requests in compressed timeframes. Load appeared in steps, not slopes. Utilities found themselves reacting to public commitments already made by technology firms. The models were not statistically wrong; they were structurally misaligned with a new class of demand that behaves more like infrastructure than consumption. 2. Closely related was a second realization: **planning cycles were simply too slow**. Integrated Resource Plans, transmission studies, and distribution upgrade programs operate on multi‑year cadences optimized for deliberation and stability. AI development operates on quarters. Throughout 2025, utilities repeatedly discovered that by the time a load scenario entered a formal docket, the underlying project had already scaled, relocated, or multiplied. The mismatch was not procedural—it was temporal. Institutions built for incremental change were being overtaken by a sector optimized for speed. 3. Not all lessons pointed toward rigidity. One of the more surprising discoveries was that **load flexibility was real—and significant**. AI workloads, particularly training and non‑latency‑critical inference, proved more shiftable than many traditional industrial loads. With the right incentives and coordination, compute could be paused, relocated, or time‑shifted to relieve grid stress. This reframed AI not only as a source of demand, but as a potential system resource. The constraint was not technical feasibility; it was governance—who controls the shift, under what conditions, and with what accountability. 4. At the same time, **AI was no longer merely something utilities planned around. It became something they used**. In 2025, AI embedded itself into grid operations—forecasting, congestion management, outage prediction, asset monitoring. A recursive dynamic emerged: the same technology driving new load was also being deployed to manage the system absorbing that load. Utilities began to experience AI as both pressure and tool, raising new questions about dependence, transparency, and operational trust. 5. **As AI loads concentrated geographically, reliability risks intensified**. Large data centers stressed substations, transmission corridors, and local balancing resources. In several regions, single interconnection requests rivaled the peak demand of entire cities. Reliability planning, historically diffuse and probabilistic, now had to contend with discrete, high‑impact nodes whose failure modes were tightly coupled to digital uptime requirements. The grid was no longer serving many small customers; it was supporting a few extremely large ones—with very low tolerance for interruption. 6. This reality reinforced another lesson: **infrastructure must lead**. Waiting for firm load to materialize before investing proved untenable. By the time AI demand arrived, it was already too late to build transmission, reinforce substations, or site generation. Utilities learned—often painfully—that anticipatory investment, while politically and regulatorily challenging, was the only viable strategy in regions courting AI development. The cost of being early was lower than the cost of being late. 7. **These pressures exposed shortcomings in rate design**. Existing tariffs were not built for customers whose load profiles could double overnight and whose capital decisions were shaped by marginal power prices. Utilities and regulators began confronting the need for new, purpose‑built tariff structures for hyperscale AI customers—structures that balance cost recovery, fairness, and economic development. The old industrial rate classes were too blunt for a customer that was simultaneously system‑critical and system‑flexible. 8. **Environmental and equity tensions sharpened as well**. AI growth raised emissions concerns in regions where incremental supply came from fossil generation. It raised fairness questions where infrastructure investments driven by data centers were socialized across broader rate bases. Utilities were forced to confront a fundamental uncertainty: would AI load accelerate decarbonization by pulling forward clean, firm investment—or slow it by crowding out limited resources? The answer varied by region. The tension was universal. 9. **The nature of customer engagement changed**. AI firms did not behave like traditional large customers. They arrived with their own models, timelines, and strategic priorities—and often expected utilities to act as partners in siting, infrastructure planning, and long‑term power strategy. Managing AI load required executive‑level engagement, cross‑sector coordination, and a willingness to operate outside standard playbooks. 10. Finally, **cyber and resilience considerations moved from background concern to central imperative**. As operational AI proliferated and digital load became more critical, the grid’s attack surface expanded. Reliability was no longer only about weather and equipment failure; it was about software integrity, data governance, and coordinated response across cyber‑physical systems. Resilience planning had to assume intelligent adversaries alongside extreme events. By the end of 2025, one conclusion was unavoidable. Generative AI did not simply add megawatts. It altered the economics, governance, and tempo of the electric system itself. Utilities that absorbed these lessons began to reframe their role—not as passive providers of power, but as active stewards of an infrastructure now inseparable from computation. ## Sign up for AIxEnergy AIxEnergy analyzes how artificial intelligence is changing electricity demand, grid operations, and energy infrastructure, with a focus on system risk, policy, and institutional decision-making. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### The Utility-AI Leadership Edge Part II: The Utility CIO’s Mandate in the Age of AI URL: https://www.aixenergy.io/the-utility-cios-mandate-in-the-age-of-ai-managing-risk-while-unlocking-grid-scale-opportunity/ Last updated: 2026-07-29T19:40:59.000Z *This is the first part in a four part article series exploring how utility C-suite leaders can harness the power of AI to deliver resilience, reliability, and growth.* For decades, the role of the utility CIO was defined by stability. Keep the lights on digitally. Protect mission-critical systems. Deliver reliability, security, and cost discipline across an increasingly complex technology stack. Success was measured by uptime, compliance, and the quiet absence of disruption. Today, that mandate has expanded—and sharpened. Audio Summary 0:00 /51.962063 1× Artificial intelligence is no longer an emerging technology confined to innovation labs or pilot programs at the margins of the enterprise. It is becoming embedded directly into grid operations, system planning, asset management, customer engagement, and storm response. AI is shaping how utilities forecast load, predict failures, dispatch crews, manage distributed resources, and respond to volatility driven by electrification and climate risk. For utility CIOs, this represents one of the most powerful opportunities in a generation—and one of the most consequential sources of enterprise risk. The question is no longer whether AI will shape the utility of the future, but how it will be governed, scaled, and integrated responsibly across systems that underpin public safety, economic stability, and critical infrastructure. The shift is not incremental. It is structural. ## **AI Is Moving Faster Than Traditional Utility Governance** One of the defining tensions for CIOs today is speed versus control. AI pilots are proliferating across utilities, often initiated by operations, engineering, or innovation teams responding to real and urgent challenges. Common use cases include outage prediction, vegetation risk management, asset failure detection, load and DER forecasting, customer insight generation, and storm restoration optimization. Many of these pilots deliver early value and demonstrate clear potential. The intent is sound. The urgency is real. Without strong CIO leadership, however, AI adoption can quickly fragment into disconnected tools, inconsistent data practices, opaque models, and unmanaged vendor relationships. What begins as innovation can quietly evolve into enterprise risk. From the CIO vantage point, the risk landscape is increasingly clear: - Operational risk: AI models are influencing operational decisions without sufficient explainability, validation, or defined human-in-the-loop controls. - Cybersecurity exposure: AI systems ingest real-time operational and customer data and integrate with core platforms, expanding attack surfaces and introducing new threat vectors. - Data integrity risk: Incomplete asset records, inconsistent telemetry, and biased training data can produce outputs that appear confident but are materially incorrect. - Regulatory and legal risk: As AI-influenced decisions affect reliability, affordability, and customer outcomes, those decisions must be auditable, defensible, and transparent. - Architectural debt: AI layered onto legacy environments without clear integration patterns increases fragility, complexity, and long-term cost. In utilities, these risks are not abstract. They intersect directly with public safety, reliability performance, and regulatory trust. CIOs are uniquely positioned to address them—not by slowing innovation, but by industrializing it. ## **AI as an Enterprise Force Multiplier** While the risks are real, the opportunity is substantial. Utilities possess some of the richest operational data sets in the economy: grid telemetry, asset histories, outage records, weather data, customer behavior, field activity, and market signals. Historically, much of this data has been siloed, underutilized, or too slow to inform real-time decisions. AI is the connective capability that allows this data to be translated into intelligence at scale. When deployed correctly, AI does not replace engineering judgment or operational expertise. It augments them. It enables utilities to move from static, deterministic planning toward probabilistic, adaptive decision-making across time horizons ranging from seconds to decades. This is increasingly evident in situations where utilities face large block load requests—often driven by data centers, electrified industrial processes, or transportation infrastructure—that exceed traditional feeder or substation assumptions. Initial responses frequently center on multi-year infrastructure reinforcement: new substations, reconductoring, or transmission upgrades. More advanced approaches reframe the challenge not as a pure capacity problem, but as a coordination and flexibility problem. By combining distributed energy resource management platforms with advanced forecasting and optimization, utilities can dynamically orchestrate distributed generation, flexible load, and operational constraints in near real time. Rather than planning exclusively for worst-case coincidence, the system continuously evaluates actual conditions and probabilistic risk. The result is the ability to support significant incremental load while maintaining reliability, avoiding unnecessary capital expenditure, and buying time for longer-term infrastructure upgrades. These outcomes are not purely technological. They depend on enterprise data integration across planning, operations, and customer systems; governance around how recommendations are generated and trusted; and tight operational alignment to ensure insights can be acted on safely in the field. The shift is from static planning to adaptive system management—enabled by intelligence rather than steel alone. ## **Breaking Silos Through Intelligence** As electrification accelerates and distributed resources proliferate, the limitations of siloed utility systems become increasingly acute. Enterprise data unification enables intelligence to cut across traditional boundaries between planning, operations, asset management, and customer platforms. When telemetry, asset condition, weather forecasts, and customer behavior are analyzed together, decision quality improves materially. Decision intelligence strengthens: - Capital planning, through scenario modeling that reflects uncertainty rather than single-point forecasts. - Storm preparedness, by probabilistically assessing damage risk, crew needs, and restoration sequencing. - Workforce deployment, by aligning skills, location, and urgency dynamically. Operational efficiency improves as analytics, diagnostics, and reporting are automated—reducing manual effort while increasing speed, consistency, and accuracy. Engineers and operators spend less time assembling data and more time applying judgment. At the same time, advanced analytics materially strengthen cybersecurity posture. Behavioral monitoring and anomaly detection improve visibility in environments where milliseconds matter and traditional rule-based tools fall short. Perhaps most importantly, these capabilities extend the impact of scarce engineering and analytics talent. As utilities face retirement waves and intensified competition for digital skills, intelligence allows smaller teams to manage greater complexity without sacrificing rigor. The CIO’s responsibility is to ensure these capabilities are built once, governed properly, and scaled across the enterprise—rather than recreated in disconnected pockets. ## **From Technology Gatekeeper to Strategic Integrator** The most effective utility CIOs are evolving from technology gatekeepers to strategic integrators. This requires a deliberate shift in posture. Rather than acting as a brake on innovation, CIOs increasingly enable the organization to move from asking whether advanced analytics should be used to understanding how they can be deployed safely and sustainably. Modern intelligence systems favor modular, API-driven architectures that allow innovation without destabilizing core platforms. This pushes CIOs to think in terms of platforms rather than point solutions—capabilities that can be reused, governed, and evolved over time. Ownership models are changing as well. Many advanced capabilities are delivered through partners rather than built entirely in-house. CIOs must orchestrate ecosystems, balancing speed, security, interoperability, and long-term viability. Vendor management now includes scrutiny of model behavior, data usage, update cycles, and intellectual property risk. Success metrics must evolve accordingly. Traditional IT KPIs are insufficient. Reliability, resilience, customer outcomes, emissions reduction, and cost avoidance increasingly become shared outcomes across IT, operations, finance, and risk. Close partnership with COOs, CFOs, and enterprise risk leaders is essential. Initiatives that succeed are jointly owned, operationally relevant, financially grounded, and technologically sound. ## **What Effective Governance Looks Like** Governance does not mean bureaucracy. It means clarity. Leading CIOs are establishing frameworks that are practical, proportional, and aligned with operational realities. Common elements include: 1. Clear prioritization tied directly to business value and system outcomes 2. Enterprise data standards, particularly for asset, telemetry, and event data 3. Model transparency and validation for decisions affecting safety and reliability 4. Cybersecurity and vendor risk controls tailored to advanced analytics environments 5. Change management and training that build operator trust and understanding In regulated environments, governance is not a constraint—it is the enabler that allows intelligence to move from pilot to production with confidence. It provides the foundation for explainability, auditability, and regulatory trust. Well-defined governance accelerates adoption by establishing clear rules of engagement. ## **The CIO’s Moment** Utilities are entering a decade defined by electrification, climate volatility, and unprecedented demand growth. Load trajectories that once followed predictable patterns are being reshaped by transportation, buildings, industry, and digital infrastructure. Advanced intelligence will be essential to navigating this complexity—but only if deployed with discipline, foresight, and leadership. This is the CIO’s moment to shape how intelligence enters the grid. Those who lead will help their organizations move beyond reactive decision-making toward adaptive, predictive, and resilient operations. They will reduce enterprise risk while accelerating innovation. And they will redefine the CIO role—from keeper of systems to architect of the intelligent utility. In an industry where trust is paramount, CIO leadership will determine whether advanced intelligence becomes a liability—or one of the most powerful strategic assets utilities have ever deployed. ## What’s Ahead in Parts III & IV The next two articles will explore the role that other C-Suite executives such as, Chief Public Affairs/Regulatory Officer and Chief Financial Officer, play in the utility-AI transformation. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### Managing Data Center Uncertainty Part V — From Uncertainty to Action: A Five-Mechanism Integrated Framework URL: https://www.aixenergy.io/managing-data-center-uncertainty-part-v-from-uncertainty-to-action-a-five-mechanism-integrated-framework/ Last updated: 2026-07-29T19:42:02.000Z *This article is part of the AIxEnergy Series: Managing Data Center Energy Uncertainty, drawn from the author's full research paper. The complete version is available directly from the author by request at michael.leifman1@gmail.com.* In [Part I](https://www.aixenergy.io/part-i-the-uncertainty-problem/) of this article series, we discussed how AI is driving unprecedented uncertainty in electricity demand, with U.S. data-center use swinging between 325 and 580 TWh by 2028\. The issue isn’t forecasting—it’s governance. Without transparency, flexibility, and better incentives, utilities overbuild, costs rise, and emissions lock in. In [Part II](https://www.aixenergy.io/managing-data-center-uncertainty-part-ii-phantom-data-centers-how-strategic-opacity-drives-overbuild/) of this article series, we reviewed how phantom data centers distort U.S. energy planning. Developers overfile interconnection requests, utilities profit from overbuilding, and regulators approve speculative capacity. Misaligned incentives create costly overbuild and fossil lock-in—requiring governance reform, transparency, and accountability. [Part III](https://www.aixenergy.io/managing-data-center-uncertainty-part-iii-the-utilization-paradox-scarcity-and-waste-inside-ai-infrastructure/) of this article series revealed that what looks like scarcity is, in many cases, inefficiency masked by opacity—a failure of synchronization between computation and electricity, compounded by a lack of data on workload types. The takeaway was clear: the cleanest power plant is the unused GPU. Unlocking that efficiency requires not more technology, but more transparency and governance that distinguish between workload types and create appropriate economic incentives for each. [Part IV](https://www.aixenergy.io/managing-data-center-uncertainty-part-iv-flexibility-is-the-new-capacity-unlocking-100-gw-without-new-generation/) revealed that the U.S. grid has nearly 100 GW of hidden capacity that could be unlocked through very small, well-timed data center curtailments—capacity that is technically feasible today but inaccessible under current economic conditions. With the right incentives, data centers can deliver this flexibility at a massive scale, enabling rapid AI-driven load growth without building new electricity generation capacity and transforming a looming grid constraint into a structural advantage. Article Summary 0:00 /36.571429 1× In this series concluding article, we contend that AI data center risk stems from poor governance, not bad forecasts, and we outline a coordinated regulatory framework to prevent overbuild, cut costs, and unlock grid flexibility before billions are stranded. This series has already established that U.S. data center electricity demand could reach 580 TWh by 2028—or plateau at 325 TWh. This 255 TWh uncertainty band, larger than Florida's total consumption, threatens tens of billions in stranded assets and decades of emissions lock-in. Strategic opacity by developers creates phantom data centers that distort planning. GPU utilization averaging just 60-70% conceals enormous flexibility potential. And geometric demand charges could unlock nearly 100 GW of grid capacity without new generation. But pricing flexibility alone won't prevent overbuild or accelerate decarbonization. Effective governance requires four additional mechanisms working in concert with demand charges: mandatory granular disclosure that reveals workload composition and flexibility potential; performance-based regulation that rewards forecast accuracy in time horizons where improvement is achievable; technology incentives that overcome capital cost barriers for efficiency and storage; and speed-to-market incentives that exploit developers' most valuable asset—time—while using costly commitments to signal genuine intent. Together, these five mechanisms address every dimension of the challenge. They reduce near-term uncertainty through better data, realign utility incentives away from capital spending bias, unlock technically feasible flexibility through compelling price signals, drive efficiency and decarbonization, and control costs by preventing unnecessary capacity additions. This is not theoretical—it builds on proven regulatory approaches, validated technical potential, and commercial demonstrations. The political obstacles are real. Utilities will resist changes to capital expenditure-based regulation. Developers will claim competitive harm from disclosure requirements. Some regulators will favor economic development over ratepayer protection. But the alternative—business-as-usual planning that treats speculative demand as certain—virtually guarantees the overbuild catastrophe these mechanisms prevent. ## The Affordability Crisis Demands Action Now Before detailing the mechanisms, the stakes must be clear. This is not merely an efficiency or climate problem—it is rapidly becoming a political crisis that threatens the coalition supporting grid modernization and decarbonization. Utilities have requested $29 billion in rate hikes for 2025, with AI data centers cited as a contributing factor. Resources for the Future notes that 25% of low-income households already face energy burdens exceeding 6% of income—the threshold for "highly burdensome." Electricity prices are rising faster than overall inflation. Residential customers already experiencing rate increases will face further pressure if tens of billions in overbuild costs are socialized across rate bases. The political coalition supporting decarbonization and grid modernization will fracture if ratepayers conclude they are paying for phantom data centers. Resources for affordable housing, education, and healthcare will be diverted to cover stranded generation assets that never should have been built. Most electricity affordability policy focuses on bill assistance, rate design reform, and subsidies—accepting rising costs as inevitable and focusing on who pays. Cost avoidance through accurate planning is an equally important but underutilized affordability strategy. If performance-based regulation prevents 4 GW of unnecessary gas generation, those $4-5 billion costs never enter the rate base. If geometric demand charges unlock Duke's 100 GW of flexibility with existing infrastructure, ratepayers avoid $10-20 billion in new generation and transmission investments. These are permanent cost reductions, not subsidies requiring ongoing appropriations. The mechanisms proposed address affordability directly: disclosure enables right-sized infrastructure; forecast accuracy PBR prevents stranded assets from entering rate base; geometric demand charges reduce need for expensive peaker plants; technology incentives lower system-wide capacity costs; speed-to-market for decarbonization eliminates fuel price volatility and reduces long-term generation costs. The affordability solution is not just bill assistance—it is cost avoidance through accurate planning. ## Mechanism 1: Granular Demand Disclosure Requirements The first mechanism addresses the measurement uncertainty that makes accurate planning impossible. Without transparent operational data, utilities cannot distinguish between speculative and genuine projects, between flexible and firm loads, or between facilities that can provide grid services and those that cannot. The Proposal: PUCs and ISOs require data center developers to provide: - Minute-by-minute power consumption data for 12+ months from three most recent comparable facilities - Multi-year demand projections (monthly average and peak) for proposed facilities - Technology specifications: chip generations, cooling systems, efficiency metrics - Workload composition: percentage breakdown of training vs. inference operations, with detail on workload types (batch processing vs. real-time services) and temporal flexibility characteristics - Demand response and load-shifting capabilities - Behind-the-meter generation and storage plans Why This Matters: No reliable public data exists on training/inference workload splits, yet this fundamentally determines which flexibility mechanisms are appropriate. Temporal load-shifting works for training; electrical storage is required for inference. Without this information, utilities cannot design targeted demand response programs and regulators cannot assess whether proposed facilities can realistically provide grid services. The absence of workload composition data is itself a critical planning challenge that mandatory disclosure directly addresses. Implementation: Phase in over 2-3 years, starting with AI data centers 50 MW and above. Facilities qualifying as "AI data centers" must demonstrate power densities ≥35 kW/rack average. Traditional enterprise data centers (<35 kW/rack average) remain under existing disclosure frameworks regardless of total MW. Protect commercially sensitive information through confidential filings to state PUCs, similar to existing protocols for utility resource planning data. Penalties for non-compliance: lower queue priority, higher interconnection charges, or denial of expedited approvals. The Rationale: This mechanism reduces measurement and parametric uncertainty by creating apples-to-apples comparisons across proposals. It signals developer seriousness—speculative projects avoid binding commitments. It creates accountability for projections by establishing benchmarks against which actual performance can be measured. And it enables informed decision-making by regulators who currently plan based on guesswork. ## Mechanism 2: Performance-Based Regulation for Forecast Accuracy and Uncertainty Management The second mechanism realigns utility incentives from capital expenditure to effective uncertainty management and forecast accuracy—but only where achievable. Different time horizons have fundamentally different uncertainty characteristics, and policy should target the uncertainties that can realistically be reduced. Critical Distinction: This proposes targeted, data center-specific performance incentives, not comprehensive performance-based regulation. The scope is deliberately narrow: PUCs would establish performance metrics and incentives/penalties specifically for utility forecasting and planning related to data center loads, implemented through a PUC order or limited rulemaking rather than comprehensive rate case restructuring. The Proposal: Under this data center-specific PBR, utilities would: - Earn bonuses for near-term (1-3 year) data center demand forecasts within ±10% of actual, where parametric and measurement uncertainty can be managed through better data collection and developer accountability - Earn bonuses for proactive uncertainty reduction efforts in all time horizons: requiring developer deposits, conducting facility audits, implementing pilot flexibility programs - Face penalties for data center-related overbuilding leading to <70% capacity utilization over rolling 5-year periods - Receive rewards for reducing peak demand through data center demand response programs - Be held accountable for stranded assets from excess data center capacity in near-term planning windows where uncertainty was reducible Distinguishing Time Horizons: Near-term forecasts (1-3 years) should achieve reasonable accuracy given known project pipelines and current technology. Parametric and measurement uncertainty dominate but can be substantially reduced through better data collection (Mechanism 1), developer deposits, and facility audits. Utilities should be held accountable for accuracy in this window because the tools for improvement exist. Long-term forecasts (5-10 years) face genuine Knightian uncertainty about AI use cases and technological evolution. Utilities cannot be held fully accountable for these unknowables, but can be incentivized to build flexibility and optionality into long-term planning—modular generation, transmission corridors designed for future expansion, contracts with flexibility provisions. The goal is not perfect long-term forecasting (impossible) but rather planning approaches that minimize regret regardless of which scenario unfolds. Implementation Timeline: Unlike comprehensive PBR reform requiring 3-5 years of multi-stakeholder processes, data center-specific PBR can be implemented through PUC investigatory dockets (6-12 months), targeted performance orders (12-18 months total), and pilot programs (24-36 months). Total realistic timeline: 2-3 years from initiation to implementation. Why This Is Feasible: Narrow scope focusing only on data center forecasting, not entire utility operations. High-stakes urgency creates political will. Data availability makes verification straightforward. Natural pilot jurisdictions exist—states with high data center concentration (Virginia, Ohio, Texas) face immediate need. Less disruptive than comprehensive reform because utilities maintain existing rate structures for other customers. The Rationale: This aligns utility incentives with ratepayer interests in time horizons where accuracy is achievable. It encourages utilities to pressure developers for accurate data in near-term planning. It reduces capital expenditure bias for data center-serving infrastructure specifically. It distinguishes between manageable near-term uncertainty and inherent long-term uncertainty. And it rewards uncertainty reduction efforts even when perfect forecasting remains impossible. Precedent Exists: The UK's RIIO framework rewards distribution network operators for efficient investment and innovation. Hawaii's PUC implemented PBR for customer satisfaction, renewable integration, and data sharing. Academic research documents that traditional rate-of-return regulation creates capital bias—this mechanism addresses it surgically for data center planning without requiring wholesale regulatory restructuring. ## Mechanism 3: Already Covered in Part IV Part IV detailed geometric demand charges and time-of-use surcharges—the pricing mechanism that makes flexibility economically compelling. The 150 MW example showing 0.9-year payback and $680+ million NPV demonstrates how appropriate price signals transform technical potential into business reality. This mechanism works in concert with the others: disclosure (Mechanism 1) reveals which facilities can provide flexibility, PBR (Mechanism 2) gives utilities incentives to pursue it, and the mechanisms below overcome capital cost barriers and accelerate deployment. ## Mechanism 4: Technology Incentives Coupled with Storage Mandates The fourth mechanism addresses the capital cost barriers that prevent voluntary adoption of flexibility and efficiency technologies, even when geometric demand charges make the economics attractive. The Proposal: Provide tax credits, accelerated depreciation, or expedited permitting for data centers that: - Install on-site battery storage ≥20% of peak load capacity - Commit to discharging storage during top 50-100 system peak hours - Deploy advanced cooling demonstrating >20% efficiency improvement - Achieve PUE <1.3 within two years Enhanced Cooling R&D Incentives: Given that cooling system constraints represent a major barrier to operational flexibility, regulators should establish dedicated R&D incentive programs for next-generation cooling technologies. These could include: - Grants or tax credits for facilities piloting novel cooling approaches (immersion cooling, advanced liquid cooling, phase-change systems) - Accelerated depreciation for cooling infrastructure investments that enable faster load ramping without thermal stress - Performance bonuses for facilities achieving both superior PUE (<1.2) and demonstrated ability to modulate load without thermal degradation - Public-private partnerships to develop and validate cooling systems specifically designed for flexible operations The Economic Logic: A 150 MW data center at PUE 1.2 uses 180 MW total. Same compute at PUE 1.6 uses 240 MW—60 MW saved per facility, enough to avoid building a $60-70 million peaker plant. Multiply across dozens of facilities and avoided generation costs reach billions in savings for all ratepayers. Battery storage provides multiple value streams beyond peak shaving. Research demonstrates data center energy storage can provide valuable grid services including frequency regulation, voltage support, and reactive power management while maintaining operational reliability. These services generate additional revenue that improves investment returns beyond what geometric demand charges alone provide. The Rationale: This mechanism directly addresses the peak demand problem that drives generation overbuild. It's technology-agnostic within categories—developers choose how to achieve thresholds based on their operational needs. It creates positive externalities because storage can provide grid services beyond data center operations. And cooling innovations enable both efficiency gains and operational flexibility, with improved PUE delivering permanent energy savings while better thermal management enables the load ramping that flexibility requires. ## Mechanism 5: Speed-to-Market Incentives for Verifiable Decarbonization Commitments The fifth mechanism exploits developers' most valuable asset—time—while using costly commitments to signal genuine intent and prevent phantom projects from distorting planning. The Context: First-mover advantages in AI are worth billions. A 6-12 month delay can be competitively catastrophic when training the next frontier model or launching a new service. Speed-to-market is more valuable to serious developers than any subsidy regulators could offer. The Proposal: Regulators offer expedited approvals, higher interconnection queue priority, or faster permitting timelines to developers who commit to one or more of the following: Option A: Clean Baseload Commitment - Place in escrow funds for CCS facility or new "clean firm" generation (advanced nuclear, enhanced geothermal) equivalent to 100% of facility load, operational within 5 years Option B: 100% Clean Energy Matching - Execute PPAs for renewable energy equivalent to 100% of annual consumption (must be additional capacity, not existing generation) Option C: Virtual Power Plant Financing - Finance VPP capacity equal to 15-20% of peak load, capable of dispatching during system peak hours Option D: Load Flexibility + Advanced Efficiency - Accept binding load caps during top 100 peak hours + deploy advanced cooling (PUE <1.2) + implement AI-driven load management Option E: Best-in-Class Efficiency Design - Commit to PUE <1.2 from day one with advanced cooling, AI load management, independent verification, and financial penalties for non-compliance Option F: Siting Flexibility for Grid Optimization - Commit to co-location strategies that leverage existing interconnection capacity (such as Colectric's "Power Couples" approach at retired power plant sites) and add substantial renewable generation and storage at the site Preventing Perverse Incentives: A legitimate concern is that expedited approval could encourage more speculative requests—developers submitting multiple phantom projects to capture queue positions. The escrow and binding commitment requirements are specifically designed to prevent this. Unlike interconnection requests that cost only modest application fees, these mechanisms require substantial upfront financial commitments. Option A places millions in escrow—speculators won't tie up capital they need elsewhere. Option B requires executing binding PPAs that create long-term financial obligations developers only sign for projects they intend to build. Option C requires creditworthy commitments that financial institutions vet through due diligence. Options D & E create downside risk through binding performance standards with financial penalties that phantom projects avoid. All options include provisions where failure to meet milestones results in forfeiture of queue position, financial penalties, or return of expedited approval benefits. The Key Insight: Revealed preference through costly action. Cheap talk (interconnection requests, zoning applications) is easy to multiply. Expensive commitments (escrow, PPAs, binding penalties) signal genuine intent. Only developers with serious projects will make these investments, naturally filtering speculative requests without regulatory assessment of "seriousness." The Rationale: This mechanism exploits speed-to-market value while encouraging strategic siting that reduces transmission costs and accelerates deployment. Facilities that can locate flexibly gain competitive advantage through faster interconnection (4-6 years vs. 8+ years for traditional approaches), while the grid benefits from avoided transmission infrastructure and efficient use of existing interconnection capacity at retired power plant sites. Co-location at sites with surplus interconnection capacity represents win-win optimization: developers get speed, ratepayers avoid transmission costs, and the system efficiently utilizes existing infrastructure. The binding commitments ensure only serious projects capture these benefits while advancing decarbonization goals. Recent Validation: Enhanced geothermal projects like Fervo Energy's Utah Innovative Power Pathway demonstrate commercial viability of clean firm power for data centers. Google has committed to procuring 115 MW of next-generation geothermal energy from Fervo. Colectric's analysis demonstrates significant potential for co-location approaches, identifying technical potential to serve 470 GW of data center load at existing U.S. power plant sites with 85% clean energy. ## How the Five Mechanisms Work Together The power of this framework lies in integration. Each mechanism addresses a different dimension of the challenge, and together they reinforce rather than duplicate. Mechanism 1 (Disclosure) creates the information foundation. Without knowing workload composition and flexibility characteristics, the other mechanisms cannot be designed effectively or targeted appropriately. Mechanism 2 (Performance-Based Regulation) realigns utility incentives to reward accuracy in near-term horizons and flexibility in long-term planning, addressing the capital expenditure bias that drives overbuild. Mechanism 3 (Geometric Demand Charges) makes flexibility economically rational once utilities and developers know which facilities can provide it and what types of flexibility are appropriate for their workloads. Mechanism 4 (Technology Incentives) overcomes the capital cost barriers that might prevent investments even when payback periods are attractive, accelerating the deployment of efficiency and flexibility technologies. Mechanism 5 (Speed-to-Market) filters phantom projects through costly commitments while accelerating decarbonization and encouraging optimal siting that avoids transmission costs. These mechanisms serve multiple constituencies simultaneously. Ratepayers avoid tens of billions in stranded assets. Data center developers gain speed-to-market advantages and regulatory certainty. Utilities' incentives align with accurate planning rather than capital expansion. Climate advocates see accelerated decarbonization. Grid operators gain flexibility resources reducing peaker plant needs. The political coalition potential is substantial if advocates articulate how preventing overbuild serves all these interests. ## Implementation: Three Parallel Tracks The solution is not waiting for perfect evidence or universal buy-in before any jurisdiction moves. Implementation should proceed on three parallel tracks: Track 1: Early Adopters (12-24 months) States facing acute pressure—Virginia, Ohio, Georgia—can implement Mechanisms 1, 3, 4, and 5 through PUC orders within twelve to twenty-four months. These early adopters provide real-world validation and demonstrate whether the mechanisms deliver promised benefits. They prevent billions in imminent overbuild while generating evidence for broader adoption. Track 2: Structured Pilots (12-36 months) States with significant but less immediate growth—Texas, North Carolina, Arizona—can run structured pilots testing refined approaches. These generate rigorous evidence about effectiveness, identify implementation challenges, and allow mechanisms to be adapted to different market structures and regulatory contexts. Pilot jurisdictions benefit from learning from early adopters while contributing to evidence base. Track 3: Evidence-Based Adoption (24-48 months) Jurisdictions with minimal current activity can monitor results from early adopters and pilots, then adopt refined mechanisms incorporating lessons learned. Late adopters benefit from proven approaches and avoid costly trial-and-error, while still preventing overbuild before it occurs in their regions. This parallel approach maximizes learning while minimizing delay. Early adopters prevent billions in imminent overbuild. Pilots generate evidence and refine mechanisms. Late adopters benefit from proven approaches. Action starts now through multiple tracks rather than waiting for comprehensive evidence before any jurisdiction moves. Timeline expectations must be realistic: Even targeted reforms face political challenges. Mechanisms 1 (disclosure), 3 (geometric charges), 4 (technology incentives), and 5 (speed-to-market) can be implemented relatively quickly (12-24 months) through PUC orders. Mechanism 2 (targeted PBR) requires more stakeholder engagement and pilot testing (24-36 months). The urgency stems from capacity decisions being made now for 2028-2035 demand. A 2-3 year regulatory process still prevents most imminent overbuild if initiated promptly. The window for action is closing, but it has not closed. ## The Political Obstacles Are Real But Not Insurmountable Implementation faces genuine resistance. Utilities will resist changes to capital expenditure-based regulation because their business model depends on it. Developers will claim competitive harm from disclosure requirements, arguing transparency helps competitors. Some regulators will favor economic development over ratepayer protection, viewing data centers as desirable industrial load. But political conditions in specific jurisdictions create opportunities. States with concentrated data center development and vocal ratepayer advocates—Virginia, Ohio, Georgia—face immediate pressure as electricity bills rise. Utility commissions that have already initiated data center investigations—Texas ERCOT, Oregon—have established momentum. Climate-progressive states—California, Washington, New York—can frame these mechanisms as decarbonization strategies aligned with their climate goals. The alternative makes political resistance unsustainable. Business-as-usual planning that treats speculative demand projections as certainties virtually guarantees tens of billions in stranded assets, decades of emissions lock-in, and political backlash undermining grid modernization. When ratepayers see electricity bills rising to cover generation that serves phantom data centers, the political coalition supporting energy transition will fracture. Preventing that fracture requires acting before the crisis becomes undeniable. The data center energy transition is fundamentally a governance and affordability challenge. We have the analytical tools to understand the uncertainty, the technical capabilities to provide flexibility, and the policy instruments to incentivize efficient outcomes. What we need is regulatory will to deploy them before tens of billions in potentially unnecessary commitments lock in for three decades. ## Conclusion: The Choice Between Design and Drift This series has traced a clear arc. Part I established that the defining challenge is structural uncertainty, not forecasting inadequacy. Part II exposed how strategic opacity and perverse incentives create phantom data centers that distort planning. Part III revealed the utilization paradox—60-70% GPU utilization concealing enormous flexibility potential. Part IV quantified how geometric demand charges can unlock nearly 100 GW without new generation. And this final installment has provided the complete policy toolkit: five integrated mechanisms that address every dimension of the challenge. The mechanisms are not theoretical. They build on proven regulatory approaches, validated technical potential, and commercial demonstrations. Disclosure requirements draw on standard practices in utility resource planning. Performance-based regulation follows models from the UK and Hawaii. Geometric demand charges apply pricing principles long used in industrial rate design. Technology incentives mirror successful programs for renewable deployment. Speed-to-market mechanisms use revealed preference logic common in project finance. The AI data center energy challenge is typically framed as a crisis: exploding demand, inadequate supply, climate setbacks, affordability pressures. This framing is not wrong, but it is incomplete. The deeper truth is that the challenge stems from design choices that can be redesigned. Strategic opacity by developers, perverse utility incentives under rate-of-return regulation, and regulatory information asymmetries are institutional arrangements that policy can address. Current rate structures that fail to incentivize technically feasible flexibility can be reformed. Affordability concerns that focus exclusively on bill assistance can be supplemented with cost avoidance strategies that prevent overbuilding. The opportunity is to transform uncertainty from crisis to catalyst for smarter, cleaner, more affordable energy systems. The mechanisms exist. The technical potential is validated. The economic logic is sound. The political conditions in some jurisdictions are favorable. The question is whether regulators will deploy these mechanisms before the window closes—or whether we'll drift into a future of stranded assets, emissions lock-in, and ratepayer backlash that could have been prevented. The choice is between design and drift. Between governance that shapes outcomes and inertia that accepts them. Between policy mechanisms that prevent overbuild and political mechanisms that allocate blame after the fact. The tools are ready. The evidence is compelling. The time is now. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## **References** Resources for the Future (2025). "Electricity Affordability 101." PowerLines (2025). "Utilities Have Requested $29 Billion in Rate Hikes for 2025, Surpassing 2024." January 2025. Lawrence Berkeley National Laboratory (2024). *United States Data Center Energy Usage Report.* LBNL-2001637. Norris, T.H., Patiño-Echeverri, D., & Dworkin, M. (2025). *Rethinking Load Growth: Assessing the Potential for Integration of Large Flexible Loads in US Power Systems.* Nicholas Institute for Environmental Policy Solutions, Duke University. Ofgem (2014). *RIIO-ED1: Final determinations for the slowtrack electricity distribution companies.* Office of Gas and Electricity Markets, UK. November 2014. Hawaii Public Utilities Commission (2020). *Decision and Order No. 37787 Establishing Performance-Based Regulation Framework.* Docket No. 2018-0088\. December 23, 2020. Ghatikar, G., Ganti, V., Matson, N., & Piette, M.A. (2012). *Demand Response Opportunities and Enabling Technologies for Data Centers: Findings From Field Studies.* Lawrence Berkeley National Laboratory. LBNL-5763E. Fervo Energy (2025). *UIPA: The Enhanced Geothermal Data Center Corridor.* July 2025. Google (2024). Google commits to Fervo Energy's next-generation geothermal energy project in Nevada. Corporate press release. Colectric (2025). "Unlocking Clean Power for Data Centers Where No One Else Can." Company presentation, October 2025. Michael Leifman (2025). *Managing Data Center Energy Uncertainty: A Framework to Prevent Overbuild, Control Costs, and Unlock Grid Flexibility.* Full research paper available from author at michael.leifman1@gmail.com. ### The Role of Artificial Intelligence in World Energy Employment 2025 URL: https://www.aixenergy.io/the-role-of-artificial-intelligence-in-world-energy-employment-2025/ Last updated: 2026-07-10T17:35:29.000Z The International Energy Agency’s (IEA) [World Energy Employment 2025](https://iea.blob.core.windows.net/assets/e5b9e908-26fe-4228-8978-e4ee36c555ed/WorldEnergyEmployment2025.pdf?ref=aixenergy.io) offers a comprehensive assessment of global labor trends across the energy sector. While the report’s core narrative focuses on job growth, skilled-worker shortages, and demographic pressures, it also provides a substantive—though understated—examination of how artificial intelligence (AI) is beginning to influence energy-sector employment. This review synthesizes and evaluates the IEA’s treatment of AI, highlighting its implications for workforce development, competitiveness, and the execution of the global clean energy transition. Article Summary 0:00 /58.017959 1× ## AI as an Emerging Contributor to Energy-Sector Productivity The report identifies artificial intelligence as “a powerful productivity tool in energy,” noting that early deployment is concentrated in administrative functions, permitting real-time monitoring, and safety applications (p. 7).¹ Virtual reality (VR)–based training is highlighted as a promising use case, particularly in offshore environments, where AI-enabled simulations reduce the need for costly and hazardous field training.² These findings align with broader industry trends: AI adoption in energy currently serves to optimize workflows and augment decision-making rather than substitute for human labor. Automation is concentrated in documentation-heavy tasks, data analysis, and predictive maintenance—domains where AI’s comparative advantage is clearest. However, the report is equally explicit in acknowledging AI’s limits. Current tools do not significantly reduce the need for field-based technical labor, which remains dominated by manual, situational, and safety-critical tasks that AI “is not currently well suited to replace” (p. 7).³ This distinction is significant. It suggests that AI is reshaping the structure of work without mitigating the acute labor shortages that characterize technical roles across the power, construction, and grid sectors. In effect, AI elevates productivity at the administrative and cognitive layers of the workforce while increasing pressure on the physical and technical layers—an important systemic insight. ## AI Talent Scarcity as a Strategic Vulnerability The IEA highlights a marked gap in AI-relevant talent between the energy sector and other industries. Between 2018 and 2024, the concentration of AI-skilled workers in utilities, oil and gas, and mining was on average 40% lower than in education, technology, finance, and media (p. 43).⁴ This talent deficit represents a critical strategic vulnerability for the sector. Several factors contribute to this disparity: **Salary Competition**—Entry-level wages in the technology sector are approximately 30% higher than in energy (p. 43).⁵ **Undefined AI Strategy**—Many energy firms lack clear roadmaps for AI deployment, making it difficult to justify specialized hiring (p. 44).⁶ **Limited Reskilling Infrastructure**—Workers report barriers such as training cost, lost wages, and limited program awareness (p. 8).⁷ These structural challenges constrain the sector’s ability to recruit and retain AI professionals at a time when digital capabilities are becoming essential for grid optimization, distributed energy resource (DER) management, asset forecasting, and advanced analytics. The report’s comparative framing underscores an important trend: while the global energy transition is accelerating, energy-sector employers are not yet competitive destinations for AI practitioners. This misalignment risks slowing modernization efforts and increasing dependence on external vendors. ## AI’s Impact on Job Roles and Occupational Structure The report describes AI as a catalyst for the evolution of job roles rather than a destroyer of occupations. It notes that AI “shifts the nature of work… requiring individuals and organisations to rethink job roles” (p. 43).⁸ Administrative functions—procurement, permitting review, compliance, and documentation—are the most exposed to automation. However, the IEA finds no evidence that AI is materially reducing employment demand in the most labor-constrained roles, including electricians, lineworkers, heavy-equipment specialists, turbine technicians, and grid engineers. These occupations, which form the backbone of the energy transition, rely on physical dexterity, in-field problem solving, and judgment under uncertainty—areas where current AI systems lack capability. The report’s occupation-level analysis reinforces this conclusion. Although AI enhances training, safety, and operational quality, “current use cases do not significantly reduce demand for applied technical workers in construction, operations, and maintenance” (p. 7).⁹ This distinction is crucial for workforce planning: AI is reconfiguring roles but not relieving the underlying shortage of skilled labor. ## Geographic and Institutional Disparities in AI Adoption The report emphasizes the uneven distribution of AI competencies across regions. Emerging market and developing economies (EMDEs) face constraints due to limited digital infrastructure, lower digital literacy, and reduced access to advanced training programs (p. 43).¹⁰ As a result, these economies face a dual challenge: they must scale labor-intensive infrastructure buildouts while simultaneously adapting to a digitalizing global energy system. Advanced economies, by contrast, typically possess greater access to digital tools, training infrastructure, and higher-skilled labor pools. However, they face demographic pressures—particularly aging workforces in nuclear and grid occupations—where AI may be leveraged for knowledge capture and augmented decision support but cannot fully compensate for retirement-driven attrition (p. 6).¹¹ The report’s cross-economy comparison reveals an important strategic implication: AI adoption may accelerate divergence between digitally mature and digitally constrained energy systems, with downstream effects on cost, deployment speed, and labor intensity. ## Education, Training, and Reskilling: The Human Prerequisites for AI Integration AI readiness requires complementary human capital investments. The report outlines several persistent obstacles: **Insufficient collaboration between firms and educational institutions**—fewer than 25% of companies participate in curriculum development for AI-related or technical programs (p. 8).¹² **Inadequate vocational pipelines**—applied technical graduates increased only 9% from 2015–2022 despite a 16% increase in demand across the economy (p. 7).¹³ **High retirement rates**—especially in nuclear and grid subsectors, threaten institutional knowledge continuity (p. 6).¹⁴ **Limited training uptake**—hindered by cost, lost wages, and low awareness (p. 8).¹⁵ Where AI is deployed as a training and knowledge-transfer tool, it shows promise—particularly through VR simulations, digital twins, and AI-assisted safety training. But the report concludes that workforce advancement depends on broad-based improvements in education partnerships, reskilling incentives, and institutional capacity-building. AI, in other words, is not a substitute for workforce development; it is dependent on it. ## Overall Assessment of the IEA’s Treatment of AI IEA’s discussion of AI is measured and empirically grounded. It avoids both overstatement and technological determinism. However, the report’s treatment remains conservative in two respects. First, the analysis underestimates AI’s strategic significance. Although the report identifies AI as a contributor to productivity and training enhancements, it does not fully integrate AI into its longer-term scenario planning, nor does it quantify AI’s potential role in addressing workforce bottlenecks in planning, forecasting, and system coordination. Second, the report understates the competitive disadvantage posed by the AI talent gap. A 40% deficit in AI-skilled workers is not merely a labor statistic; it is a threat to modernization, grid reliability, and the pace of energy system transformation. Nonetheless, the document provides a rigorous foundation for understanding how AI intersects with energy labor markets. It makes clear that AI is already changing the shape of work—even if it is not yet changing the volume of it. ## Conclusion World Energy Employment 2025 presents a nuanced portrait of a global workforce navigating simultaneous pressures: rapid infrastructure expansion, demographic strain, and emerging digitalization. The report’s findings on AI, though subtle, carry major strategic implications: 1. AI elevates productivity and administrative efficiency but does not solve skilled labor shortages. 2. AI talent scarcity is a defining competitiveness challenge for energy companies and national labor markets. 3. Job displacement effects are concentrated in administrative roles, while technical trades remain shielded. 4. Without investment in training and education ecosystems, AI adoption may widen regional and occupational disparities. For policymakers, energy firms, investors, and workforce planners, the message is clear: AI is an essential enabler of the energy transition, but it cannot—and will not—replace the need for a robust, well-trained human workforce. Its role is complementary, not substitutive. In this sense, the report underscores a central reality: the future of energy work will be defined not by the contest between humans and machines, but by the quality of integration between them. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## Notes 1. World Energy Employment 2025, 5. 2. Ibid., 7. 3. Ibid., 7. 4. Ibid., 7. 5. Ibid., 43. 6. Ibid., 44. 7. Ibid., 8. 8. Ibid., 43. 9. Ibid., 7. 10. Ibid., 43. 11. Ibid., 6. 12. Ibid., 8. 13. Ibid., 7. 14. Ibid., 6. 15. Ibid., 8. ### Managing Data Center Uncertainty Part IV — Flexibility Is the New Capacity: Unlocking 100 GW Without New Generation URL: https://www.aixenergy.io/managing-data-center-uncertainty-part-iv-flexibility-is-the-new-capacity-unlocking-100-gw-without-new-generation/ Last updated: 2026-07-29T19:42:06.000Z *This article is part of the AIxEnergy Series: Managing Data Center Energy Uncertainty, drawn from the author’s full research paper. The complete version is available directly from the author by request at* *michael.leifman1@gmail.com* *.* Article Summary 0:00 /50.207347 1× In [Part I](https://www.aixenergy.io/part-i-the-uncertainty-problem/) of this article series, we discussed how AI is driving unprecedented uncertainty in electricity demand, with U.S. data-center use swinging between 325 and 580 TWh by 2028\. The issue isn’t forecasting—it’s governance. Without transparency, flexibility, and better incentives, utilities overbuild, costs rise, and emissions lock in. In [Part II](https://www.aixenergy.io/managing-data-center-uncertainty-part-ii-phantom-data-centers-how-strategic-opacity-drives-overbuild/) of this article series, we reviewed how phantom data centers distort U.S. energy planning. Developers overfile interconnection requests, utilities profit from overbuilding, and regulators approve speculative capacity. Misaligned incentives create costly overbuild and fossil lock-in—requiring governance reform, transparency, and accountability. [Part III](https://www.aixenergy.io/managing-data-center-uncertainty-part-iii-the-utilization-paradox-scarcity-and-waste-inside-ai-infrastructure/) of this article series revealed that what looks like scarcity is, in many cases, inefficiency masked by opacity—a failure of synchronization between computation and electricity, compounded by a lack of data on workload types. The takeaway was clear: the cleanest power plant is the unused GPU. Unlocking that efficiency requires not more technology, but more transparency and governance that distinguish between workload types and create appropriate economic incentives for each. In this installment, we show that the U.S. grid has nearly 100 GW of hidden capacity that could be unlocked through very small, well-timed data center curtailments—capacity that is technically feasible today but inaccessible under current economic conditions. The conclusion is that with the right incentives, data centers can deliver this flexibility at a massive scale, enabling rapid AI-driven load growth without building new electricity generation capacity and transforming a looming grid constraint into a structural advantage. The 100 GW estimate cited here comes from recent Duke University research that indicated that U.S. balancing authorities could integrate nearly 100 GW of new load—equivalent to 10% of national peak demand—if data centers reduced consumption by just 0.5-1% during the grid's top 177 hours annually. The average curtailment would last about 2.1 hours per event, with half the load continuing to run even during these brief reductions. This isn't theoretical: it builds on technically feasible flexibility in AI training workloads that already experience multi-day runtimes due to data delivery bottlenecks. Yet this enormous potential remains untapped. Current rate structures make flexibility economically irrational. A data center paying 6-8¢/kWh faces electricity costs representing 15-25% of operating expenses—substantial in absolute terms, but insufficient to drive behavioral change when capital costs dominate decision-making and traditional demand charges run just $10-20/kW-month. The solution isn't gentle nudges. It's making flexibility obviously profitable through geometric demand charges that create cliff effects—where crossing a threshold costs millions annually—combined with time-of-use surcharges during the grid's most stressed hours. A 150 MW facility investing $98 million in batteries, advanced cooling, and load management could see a 0.9-year payback period and $680+ million in net present value over 15 years. At that return, flexibility stops being a favor to the grid and becomes core business strategy. This article details the pricing mechanisms that can unlock Duke's 100 GW opportunity, explains why different facility types will pursue different strategies based on their workload characteristics, and shows how pass-through payment structures can cascade price signals through the entire compute value chain. ## **The 100 GW Opportunity: What Duke Found** Duke University's Nicholas Institute analysis introduced "curtailment-enabled headroom"—a mechanism through which additional load can be absorbed using existing capacity with only modest, short-duration usage reductions. The findings quantify an enormous untapped resource: nearly 100 GW of new large loads could be integrated across the 22 largest U.S. balancing authorities—covering 95% of U.S. load—if data centers can reduce power usage by as little as 0.5-1% during peak demand periods. The numbers are striking. Duke Energy Progress alone could gain 1.3 GW of capacity for data centers if facilities could withdraw 0.5% of power during peak system stress. At the 0.5% curtailment rate—representing 177 hours per year—load curtailment would average about 2.1 hours per event. Critically, on average for 88% of curtailment time, half the new load would continue running. These are partial reductions, not complete shutdowns. The study assumes constant load additions at 100% utilization for modeling purposes. In the context of current demand growth, AI data centers represent the primary application of this flexibility potential, though the analysis applies to "large flexible loads" generally. But Duke's analysis comes with important caveats. The researchers didn't consider transmission constraints, which could limit available headroom in specific regions. They based calculations on peak demand levels without factoring in reserve margin capacity, which could increase potential headroom. And most critically, the study quantifies theoretical grid capacity without examining whether the operational and economic barriers to flexibility can be overcome. ## **The Technical Barriers Are Real But Solvable** While Duke's analysis proves the grid has spare capacity, practical obstacles exist. Data centers already operate at 60-70% GPU utilization due to I/O bottlenecks—but as Part III of this series established, this reveals which workloads are flexible rather than constraining flexibility potential. State persistence and coordination create challenges. Many workloads cannot simply pause. Database transactions must complete, active sessions require continuity, and model training involves thousands of synchronized GPUs. User-facing services—the most economically valuable workloads—are time-sensitive and difficult to interrupt. Multi-tenant complexity looms large for colocation data centers serving hundreds of customers. Determining whose workloads get curtailed while maintaining service level agreements—typically 99.99% uptime, meaning just 52.6 minutes of acceptable downtime per year—creates enormous operational complexity. Cooling and thermal management systems respond slowly and are optimized for steady-state operation. Rapid load changes create thermal transients, hot spots, or force inefficient part-load operations that can stress equipment. Economic incentives present the deepest barrier. Data centers represent $10-15 million capital investment per megawatt. The business model assumes 60-80%+ utilization. Deliberately underutilizing capacity means accepting lower returns unless compensation mechanisms are very attractive—and current rate structures provide minimal rewards. Duke's curtailment-enabled headroom analysis applies most readily to facilities with inherently flexible workloads—particularly those running I/O-constrained AI training jobs that are not latency-sensitive. For such facilities, the 0.5-1% curtailment during peak periods (averaging 2.1 hours per event at the 0.5% rate) could be achieved through temporal load-shifting: deferring batch training jobs to off-peak hours with minimal operational disruption. The computational output completes slightly later, but the business impact is negligible for workloads already experiencing multi-day training times due to I/O bottlenecks. In contrast, facilities serving real-time user requests face different constraints. For these compute-optimized facilities serving latency-sensitive workloads—inference, web services, database operations—achieving the same grid flexibility requires electrical demand management via battery storage that maintains computational continuity while reducing peak electrical draw, rather than direct computational curtailment that would violate service requirements. A fundamental data gap undermines precise quantification. No reliable public data exists on what proportion of projected AI data center growth represents training versus inference workloads. Without this information, we cannot determine what portion of Duke's 100 GW opportunity is achievable through temporal load-shifting versus requiring alternative mechanisms like battery storage. Anthropic's forecast of roughly equal training and inference capacity by 2028 for frontier AI suggests perhaps 40-60 GW of Duke's potential could come from training workload flexibility, with the remainder requiring electrical rather than computational flexibility—but this remains speculative absent better disclosure. Accounting for the technical and economic barriers documented above, actual achievable flexibility will be lower than the 100 GW theoretical potential, but without workload composition data and further analysis, we cannot determine whether achievable flexibility falls in the range of 75-90 GW or as low as 10-25 GW. This uncertainty itself demonstrates why mandatory disclosure is essential for effective policy design. ## **The Economic Logic: Why Current Pricing Fails** The technical potential for flexibility documented by Duke and MIT raises an obvious question: if flexibility is feasible and valuable, why haven't AI data centers embraced it through existing demand response programs? The answer lies in opportunity costs. Data centers represent $10-15 million capital investment per megawatt. The business model assumes 60-80%+ utilization. Deliberately underutilizing capacity means accepting lower returns unless demand response compensation offsets foregone revenue. Contractual service level agreements create legal constraints. Cloud providers guarantee 99.9-99.99% availability—just 52.6 minutes acceptable downtime per year. These are legally binding with financial penalties for violations. Demand pricing complexity matters. The most economically valuable workloads are user-facing and time-sensitive. Building a business around interruptible compute means serving less profitable market segments. Stranded capacity risk is real. Committing to demand response requires maintaining spare capacity year-round with ongoing costs but no revenue during normal operations. Multi-tenancy coordination is especially challenging. Colocation facilities face daunting obstacles deciding whose workloads get curtailed while maintaining diverse SLAs. Hyperscalers controlling their full stack have clear advantages. Most fundamentally, inadequate price signals render flexibility irrational. Data centers in 2024-2025 pay 6-8¢/kWh for electricity, with electricity costs representing 15-25% of operating expenses. Yet even substantial absolute costs do not drive flexibility investments when rate structures lack appropriate price signals. Current electricity costs and traditional demand charges of $10-20/kW-month are too modest to change behavior where capital costs and revenue generation dominate decision-making. ## **The Solution: Geometric Demand Charges That Create Cliff Effects** The proposed mechanism implements a two-part pricing structure that makes flexibility investments economically compelling. **Part 1: Geometric Monthly Demand Charges** based on the facility's highest 30-minute average demand during each month, with rates increasing geometrically: - 0-100 MW: $15/kW-month - 100-125 MW: $26.25/kW-month (1.75× multiplier) - 125-150 MW: $45.94/kW-month (1.75× multiplier) - 150+ MW: $80.39/kW-month (1.75× multiplier) **Part 2: Time-of-Use Energy Surcharges** of $0.50/kWh (versus baseline $0.09/kWh) during the grid's top 100 stress hours annually, identified through day-ahead ISO price signals. These mechanisms work together. Demand charges incentivize year-round peak reduction through efficiency and right-sizing. TOU surcharges reward load shifting and battery discharge specifically when the grid needs help most. Both make the same flexibility investments—battery storage, advanced cooling, load management—economically rational. Consider a 150 MW data center with typical utilization patterns. Under current rate structures, the facility pays roughly $86 million annually in electricity costs: $64.4 million in energy charges at $0.09/kWh for 715,600 MWh consumed yearly (assuming 55% average utilization), plus approximately $21.6 million in standard demand charges at $12/kW-month. Under the proposed structure without flexibility measures, costs rise dramatically. Monthly peaks at 150 MW trigger the highest geometric tier at $80.39/kW-month, generating $144.7 million in annual demand charges alone. Operating at full 150 MW capacity during the 100 grid-stressed hours adds $6.15 million in TOU surcharges. Total annual electricity costs reach $215.3 million—a 150% increase over current rates. However, investing $98 million in flexibility infrastructure transforms the economics. The investment breakdown: a 30 MW, 4-hour battery system ($48M at approximately $1,600/kW installed cost), advanced cooling retrofits for 40% of racks ($40M for immersion or advanced liquid cooling systems), and AI-driven load management platforms ($10M for enterprise-scale DCIM systems with real-time optimization). By keeping monthly peaks at 125 MW through battery discharge and load management, demand charges drop to $39.4 million annually. Reducing consumption 20% during the 100 grid-stressed hours through battery discharge and temporal load shifting cuts TOU surcharges to $4.92 million. Total electricity costs: $108.7 million, saving $106.6 million annually compared to the no-flexibility scenario. Against the $98 million capital investment, this yields a 0.9-year payback period and a 15-year net present value exceeding $680 million at a 7% discount rate. Even at current baseline rates, the avoided costs justify the investment. The geometric structure simply makes flexibility investments obviously rational rather than marginally attractive. ## **The Power of Cliff Effects** The geometric progression creates powerful threshold effects. A facility peaking at 124 MW pays $26.25/kW-month ($39.3M annually), while one crossing to 126 MW pays $45.94/kW-month ($69.4M annually)—a $30.1 million annual penalty for 2 MW. These cliff effects make investing in flexibility to stay below thresholds economically compelling. The mechanism is technology-neutral: facilities choose whether to achieve reductions through temporal load-shifting, battery storage, improved cooling efficiency, or workload optimization based on their operational characteristics. One critical detail: the geometric demand charge structure charges ALL capacity at the highest tier reached during the month, not marginal pricing. If a facility peaks at 126 MW even once during the month, the entire 126 MW is charged at the $45.94/kW-month rate. This creates even stronger incentives for peak management than marginal pricing would provide. Of course, different facility types will respond differently based on their workload characteristics. Facilities running I/O-constrained training workloads may emphasize temporal shifting, deferring batch jobs to off-peak hours. The computational output completes slightly later, but the business impact is negligible for workloads already experiencing multi-day training times. Compute-optimized facilities serving latency-sensitive inference workloads will invest more heavily in battery storage to maintain computational continuity while reducing electrical peaks. Direct computational curtailment would violate service requirements. Multi-tenant colocation facilities may develop curtailable compute markets with pass-through payment structures, where demand charge savings flow to compute customers who accept occasional curtailment. This cascades price signals through the value chain. Hyperscalers controlling their full stack can implement flexibility more readily than facilities serving diverse external customers, but all face consistent price signals. ## **Pass-Through Payments: Cascading Price Signals** Pass-through payment structures deserve emphasis. Just as utilities offer demand response payments to customers who curtail electricity usage, data centers can compensate compute customers for accepting curtailment. When a data center reduces peak demand and avoids geometric penalties, those savings flow through as payments to customers whose workloads were deferred. This cascades price signals through the value chain: utilities compensate data centers through reduced charges, data centers compensate compute customers for flexibility, and the entire system efficiently allocates curtailment capability. Regulators should encourage transparent pass-through mechanisms, potentially requiring standardized curtailment product offerings that distinguish "guaranteed compute" from "curtailable compute with advance notice" at differentiated prices reflecting grid value. Implementation must address operational constraints. Thermal management systems respond slowly. On-site battery storage allows instant response to price signals while gradually ramping compute loads, avoiding thermal stress. Service level agreements matter. The proposal's 100-hour annual duration with day-ahead notice allows operators to maintain contractual commitments through temporal shifting, geographic load-balancing, or battery discharge rather than violating SLAs. Multi-tenant coordination remains challenging, but technical assistance grants or automated curtailment platforms can address this barrier. The key is making the economic case overwhelming enough that facilities invest in solving operational challenges. ## **Academic Validation** Academic research validates the approach. Studies demonstrate data centers can reduce coincident peak by 15-30% through workload shifting with minimal disruption, confirming the technical feasibility of required reductions. Battery storage research shows data center energy storage provides valuable grid services including frequency regulation and voltage support while maintaining reliability. Duke's analysis quantifying 100 GW of integration potential with 0.5-1% curtailment provides evidence that required flexibility magnitudes are achievable, though primarily for training rather than inference workloads. The mechanism works alongside other policy tools. Technology incentives help overcome capital cost barriers for storage and advanced cooling. Disclosure requirements ensure utilities understand which facilities can provide flexibility. Performance-based regulation gives utilities incentives to pursue demand-side solutions. Speed-to-market rewards facilities combining flexibility with decarbonization. Together, these transform flexibility from theoretical possibility to economically rational business strategy. MIT's comprehensive analysis of generative AI's climate impact identifies multiple efficiency opportunities that compound flexibility benefits. Algorithmic improvements show efficiency gains from new model architectures that solve complex problems faster are doubling every eight or nine months. This means the same computational result requires progressively less energy over time. Hardware optimization demonstrates that "turning down" GPUs so they consume about three-tenths the energy has minimal impacts on AI model performance. This isn't theoretical—it's operational practice at facilities prioritizing efficiency. Training efficiency reveals about half the electricity used for training an AI model is spent to get the last 2-3 percentage points in accuracy. Early stopping can save substantial energy with minimal quality impact for many applications. Temporal load shifting proves splitting computing operations so some are performed when more electricity is from renewable sources can significantly reduce a data center's carbon footprint. This aligns naturally with the geometric demand charge structure. Long-duration storage could be a "game-changer" because data centers could use stored renewable energy during high-demand periods. This converts intermittent renewables into dispatchable capacity while providing grid flexibility. These efficiency improvements enable flexibility rather than competing with it. A facility that solves I/O bottlenecks achieves higher baseline utilization AND can provide more flexibility through electrical storage because the computational output per watt increases. ## **What's Next** Part IV has shown that the 100 GW flexibility opportunity Duke identified is real and achievable—if economic incentives align with technical potential. Geometric demand charges and time-of-use surcharges create that alignment, making sub-one-year payback periods possible for comprehensive flexibility investments. Part V, the final installment of this series, will complete the policy framework. Beyond pricing flexibility, effective governance requires four additional mechanisms: mandatory granular disclosure that reveals workload composition and flexibility potential; performance-based regulation that rewards forecast accuracy over capital expansion; technology incentives that overcome upfront cost barriers; and speed-to-market incentives that accelerate decarbonization while preventing phantom projects. Together, these five integrated mechanisms transform data center energy uncertainty from crisis to opportunity—preventing overbuild, controlling costs, and unlocking the flexibility that makes rapid AI growth compatible with grid reliability and climate commitments. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## **References** Norris, T.H., Patiño-Echeverri, D., & Dworkin, M. (2025). *Rethinking Load Growth: Assessing the Potential for Integration of Large Flexible Loads in US Power Systems.* Nicholas Institute for Environmental Policy Solutions, Duke University. MIT News (2025). "Responding to the climate impact of generative AI." September 30, 2025. Lawrence Berkeley National Laboratory (2024). *United States Data Center Energy Usage Report.* LBNL-2001637. Turner & Townsend (2024). "Data Center Cost Index 2024." International Construction Market Survey. Amazon Web Services (2024). "Amazon Compute Service Level Agreement." AWS Legal Documentation. Aikema, D., Simmonds, R., & Zareipour, H. (2012). "Data centre demand response: Avoiding the coincident peak via workload shifting and local generation." *Performance Evaluation*, 70(10), 770-791. Ghatikar, G., Ganti, V., Matson, N., & Piette, M.A. (2012). *Demand Response Opportunities and Enabling Technologies for Data Centers: Findings From Field Studies.* Lawrence Berkeley National Laboratory. LBNL-5763E. Michael Leifman (2025). *Managing Data Center Energy Uncertainty: A Framework to Prevent Overbuild, Control Costs, and Unlock Grid Flexibility.* Full research paper available from the author. ### What’s Holding Utilities Back from Scaling the Grid for AI? URL: https://www.aixenergy.io/whats-holding-utilities-back-from-scaling-the-grid-for-ai/ Last updated: 2026-07-29T19:42:55.000Z The energy transition was never meant to unfold at the speed of software. Yet that is precisely the tempo AI has imposed on the grid. Over the past decade, utilities have operated inside a familiar cadence—plan, permit, procure, build—each stage measured in years. AI has no patience for these intervals. A hyperscale data center can surface on a development roadmap in January and demand gigawatts of power by the following summer. EV fast-charging networks scale in quarters, not decades. Industrial automation arrives as a full-stack digital retrofit—instantaneous in operational impact, relentless in its energy appetite. Article Summary 0:00 /46.419592 1× The contrast is stark: AI-driven demand materializes on the order of months, while grid infrastructure evolves on the order of years. It is a temporal mismatch, but more importantly, it is a strategic one. If left unresolved, the grid becomes the bottleneck to the AI economy—a constraint that shapes where compute clusters form, where innovation concentrates, and ultimately, where economic value migrates. Utilities sense the inflection point, even if the legacy structures around them were never designed for it. And so the central question emerges: What is truly holding utilities back from scaling the grid for AI? The answer is not singular. It is systemic—woven through regulation, engineering, supply chains, workforce capabilities, and financial architecture. But beneath those structural barriers lies a deeper shift: the grid must evolve from a system built solely to deliver supply into a system designed to orchestrate demand. ## **The Temporal Dilemma: When Data Outpaces Steel** The challenge begins with timing. AI loads behave unlike any other demand category utilities have ever faced. Industrial parks once took a decade to reach peak load. Residential growth followed demographic trends. Even earlier waves of digital infrastructure—co-location facilities, telecom nodes, server farms—scaled gradually. But hyperscale AI centers do not scale gradually. They appear fully formed, with procurement cycles measured in fiscal quarters and power needs that rival medium-sized municipalities. Utilities, by contrast, must navigate a planning environment defined by multi-year transmission studies, environmental reviews, permitting processes, and construction lead times. In many regions, a transmission project that begins today will not energize until the early 2030s. This asynchrony forces a strategic reckoning: utilities built for predictability must now operate in a world defined by immediacy. And immediacy is unforgiving. When demand surges ahead of infrastructure, reliability risk grows. Economic opportunity migrates. Stakeholders—from regulators to OEMs to data center developers—begin searching for alternative jurisdictions that can match the tempo of innovation. Utilities are not resistant to change. They are constrained by the architecture within which they operate. And five constraints, in particular, define the boundary conditions of their response. ### *1\. Regulation and Permitting: Governance at a Different Speed* Regulation is not a barrier by design; it is a safeguard. It protects communities, ensures transparency, and preserves environmental integrity. But the regulatory framework governing transmission was not built for a world in which gigawatts of new load can materialize with almost no warning. Every mile of transmission becomes a negotiation across dozens of stakeholders. Environmental impact assessments unfold in sequential reviews. Right-of-way discussions cascade across local jurisdictions. Interconnection studies, already burdened by renewable queues, become chokepoints that delay projects measured in months of engineering work but years of administrative throughput. The result is a predictable pattern: utilities can see the load coming, but cannot break ground before the load arrives. Some regions are beginning to experiment with fast-track zones—pre-assessed corridors designed to compress approval timelines. Others are developing more robust, data-driven regulatory cases that quantify reliability risk, economic opportunity, and the cost of inaction. Still others are reshaping their General Rate Case and Risk Assessment filings to anticipate the shifting geography of AI demand. These efforts point toward a new regulatory rhythm—one that maintains rigor but recognizes that delay itself has become a systemic risk. ### *2\. Physical Capacity: When Conductors Become the Constraint* If regulation governs the pace, physics governs the limits. Much of the U.S. transmission system is already operating close to thermal and voltage thresholds. The grid was not overbuilt; it was right-sized for an era of steady, incremental load growth. AI breaks that equilibrium. Paradoxically, the challenge is not always generation. In many regions, ample power exists. The constraint is the ability to move that power—across constrained corridors, through aging substations, into rapidly growing load pockets that lack the infrastructure to serve them. Utilities have begun turning to real-time intelligence to stretch existing assets. Dynamic Line Ratings—a technology mandated in spirit by FERC Order 881—offer a revealing insight: capacity is not a fixed number but a dynamic condition influenced by weather, conductor temperature, and loading patterns. With real-time monitoring, many lines can carry 10–30% more power safely, without new steel or new permits. Advanced conductors, topology optimization, and probabilistic planning further expand what existing networks can handle. Yet the principle remains the same: physical infrastructure is slow, digital augmentation is fast—utilities must leverage both. ### *3\. Supply Chains: Scarcity as a Structural Reality* Even when engineering designs are complete and permits approved, utilities face a challenge that rarely makes headlines but determines the fate of nearly every project: the supply chain. Large power transformers now carry lead times of 18–36 months. Protection and control systems routinely take more than a year. Switchgear, relays, and voltage control systems—once predictable procurement items—are now global bottlenecks. The irony is striking: the grid is held back not only by regulatory process or physical constraints, but by the availability of steel, copper, and silicon. Some utilities have responded by standardizing equipment designs to reduce procurement fragmentation. Others have negotiated long-term manufacturing capacity reservations, much like hyperscalers reserving cloud capacity. A growing number are pursuing co-investment models in which data center developers help finance procurement commitments in exchange for accelerated timelines. The lesson is clear: grid readiness is not only an engineering challenge—it is a supply chain strategy problem. ### *4\. Workforce and Skills: A Grid Built for Determinism Entering an Era of Probabilistic Operations* The modern grid is becoming a digital organism. Its signals come not only from SCADA but from millions of AMI meters, phasor measurement units, DERMS platforms, and IoT devices. Managing this organism requires fluency in machine learning, probabilistic modeling, cybersecurity, digital twins, and real-time automation. Utilities possess deep expertise in deterministic operations—rotating machines, fault currents, protection coordination, contingency planning. But the AI-era grid behaves differently. It demands an operational model that blends engineering intuition with algorithmic intelligence. The workforce challenge is not a skills gap alone; it is a paradigm gap. Utilities must expand their technical repertoire while preserving the reliability ethos that defines their culture. Training programs in AI, advanced analytics, and OT/IT integration are beginning to take shape. Partnerships with engineering firms and hyperscalers are becoming strategic rather than transactional. Joint innovation centers are emerging as accelerators where grid operators learn directly from technology OEMs and vice versa. This is not a transition from mechanical to digital—it is a convergence. And convergence demands a workforce fluent in both. ### *5\. Funding and Regulatory Approval: When Digital Value Meets Analog Frameworks* The final constraint is conceptual rather than technical. Regulators understand the value of physical assets—substations, transformers, wires. Their benefits can be quantified, depreciated, and recovered within familiar frameworks. Digital infrastructure is different. Its benefits are systemic, distributed, and often probabilistic. An AI analytics platform can improve reliability, but it does so indirectly—by predicting failures, shaping load, preventing overload conditions. A DERMS platform can mitigate wildfire risk, but the value is expressed as avoided events, not constructed assets. And so utilities encounter a paradox: the tools needed to manage AI-era load growth are digital, but the regulatory mechanisms that approve investments in those tools remain analog. Some utilities have begun framing digital investments explicitly in terms that regulators recognize—SAIDI/SAIFI improvements, wildfire mitigation outcomes, and reliability risk reduction. Others are leveraging federal funding to reduce ratepayer exposure. Still others are experimenting with shared investment models in which hyperscalers co-finance digital modernization as part of interconnection agreements. The shift is gradual but inevitable. As the grid becomes a digitally orchestrated system, the regulatory models governing it must evolve. ## **A Deeper Transformation: From Building Capacity to Orchestrating Flexibility** Solving these five constraints accelerates infrastructure deployment—but acceleration alone cannot close the temporal gap between AI load growth and grid expansion. The deeper shift is conceptual. For more than a century, utilities have operated under a single assumption: demand is firm and must be served with firm supply. AI upends that assumption. Modern data centers are not rigid consumers of electricity. They possess on-site backup generation, battery storage, advanced load management systems, and the ability to shift non-time-sensitive compute tasks. They are flexible—even if today’s regulatory structures do not treat them as such. A study from Duke University crystallizes the opportunity. If data centers curtailed load for just 22 hours per year—0.25% of the time—the grid could accommodate an additional 76 GW of demand without adding a single power plant or transmission line. The implication is profound: the fastest path to scaling the grid for AI is not only building more capacity, but unlocking flexibility in the demand that uses it. This is not theoretical. Flexible interconnection agreements are beginning to emerge, offering non-firm service that connects faster with lower cost. Dynamic tariff structures are aligning economic incentives with periods of grid stress. And utilities are increasingly exploring how behind-the-meter assets—backup generators, battery systems, thermal storage—can participate directly in grid operations. In this future, demand becomes an orchestrated asset. Data centers shift load in response to grid conditions. AI workloads dynamically migrate across regions based on real-time capacity signals. The grid stops treating load as an immutable requirement and instead treats it as a resource—a tool that can be shaped, modulated, and dispatched. This is the paradigm shift: from infrastructure as destiny to flexibility as strategy. ## **The Path Forward: Utilities at the Center of the AI Economy** The utilities that lead in the AI era will not follow a single strategy. They will follow a dual one. They will accelerate traditional infrastructure—permitting reform, dynamic line ratings, standardized procurement, workforce evolution, and modernized funding. But they will pair that acceleration with something far more transformative: a commitment to load orchestration as a core operational capability. The grid becomes not merely a delivery mechanism, but a platform. A system where physical capacity and digital intelligence reinforce one another. A system where flexibility carries as much strategic value as steel. A system where utilities shape demand rather than chase it. AI does not wait for infrastructure. It expands along the fastest available pathway. The regions that modernize their grid operations will become gravitational centers for AI investment, economic growth, and technological leadership. Those who do not will find themselves constrained not by engineering limitations but by the assumptions they failed to abandon. The future of the grid will not be defined solely by the assets utilities build. It will be defined by the intelligence they deploy, the flexibility they unlock, and the pace at which they adapt to a world where demand moves faster than steel can be set. Utilities now stand at a crossroads—one defined not by risk, but by opportunity. The AI economy is forming its infrastructure map today. The question is simple: who will build the grid it needs, and who will watch opportunity flow to those who do? ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### The Great Shift in AI Infrastructure: From Hyperscale Training to Edge-Based Real-Time Action URL: https://www.aixenergy.io/the-great-shift-in-ai-infrastructure-from-hyperscale-training-to-edge-based-real-time-action/ Last updated: 2026-06-30T02:27:44.000Z AI’s future isn’t in desert mega-campuses—it’s emerging in old warehouses wired with power across America. As data-center demand strains grids and latency needs rise, these forgotten industrial buildings are becoming the critical, low-latency nodes of the AI era. _This post is for paying subscribers only._ ### The Genesis Mission and the Architecture of National Scientific Capacity URL: https://www.aixenergy.io/the-genesis-mission-and-the-architecture-of-national-scientific-capacity/ Last updated: 2026-07-10T17:35:40.000Z On November 24, 2025, the White House issued an executive order establishing the Genesis Mission, a coordinated national effort to accelerate scientific discovery through artificial intelligence. While the document invokes historical analogies and national aspirations, its operational content is technical, detailed, and administrative. It directs the Department of Energy (DOE) to create a unified computational and data environment—the American Science and Security Platform—capable of supporting large-scale AI research across scientific domains. Author's Summary 0:00 /69.668571 1× From an AIxEnergy perspective, this order represents a structural development in how the United States organizes its scientific assets. It positions AI not as a standalone innovation but as an integrated capability dependent on computing capacity, secure datasets, advanced modeling, and the physical infrastructure of the grid. Although the order reflects the political objectives of the issuing administration, its mechanisms and timelines outline a practical federal effort aimed at upgrading national scientific capability. This analysis provides a neutral, clear-eyed interpretation of what the Genesis Mission does, what it requires, and how it intersects with the energy and computing landscape. ### **The Purpose of the Genesis Mission: AI as an Accelerator of Discovery** The executive order frames the Genesis Mission around a straightforward premise: scientific discovery can be accelerated by modern AI techniques, particularly those capable of training on large, federally curated datasets. The document identifies multiple goals: - to increase the speed of scientific experimentation and hypothesis testing; - to integrate federal datasets into unified, AI-ready formats; - to support national laboratories and research institutions with shared AI capabilities; - to strengthen U.S. competitiveness in fields such as biotechnology, semiconductors, quantum science, and advanced manufacturing; and - to improve the effectiveness of federal research and development. The core idea is that AI can analyze large volumes of scientific data, generate models across domains, automate components of scientific workflows, and help identify new directions for research. This aligns with trends already underway in materials science, chemistry, fusion, and climatology, where machine learning is increasingly used to accelerate discovery cycles. ### **The Structure of the Mission: DOE Leadership and NSTC Coordination** The order places primary responsibility for implementation with the **Secretary of Energy**. DOE is tasked with integrating its laboratories, data infrastructure, and high-performance computing resources into a single secure platform. The Secretary may appoint a senior official to oversee operations, emphasizing the scale and complexity of the initiative. Policy coordination is assigned to the Assistant to the President for Science and Technology, who acts through the National Science and Technology Council (NSTC). This reflects the cross-agency nature of the Mission: federal datasets reside across multiple departments, and foundational scientific challenges extend beyond DOE. While the document is aspirational in tone, the operational responsibilities it assigns are pragmatic: asset cataloging, data standardization, cross-agency guidance, and secure platform construction. ### **The American Science and Security Platform: Infrastructure Requirements** The most substantial component of the order is Section 3, which creates the American Science and Security Platform. This platform is envisioned as an integrated system combining compute, data, modeling tools, and experimental resources. Its required capabilities include: 1. High-performance computing resources from DOE laboratories and secure cloud environments. 2. AI modeling and analysis frameworks, including AI agents capable of exploring scientific design spaces and operating within research workflows. 3. Simulation and predictive modeling tools supporting a range of scientific disciplines. 4. Domain-specific foundation models trained on curated scientific data. 5. Secure access to datasets, including proprietary and classified information, governed by federal data standards. 6. Experimental and production tools, enabling autonomous or AI-assisted research. These elements together define a distributed scientific computing ecosystem. For AIxEnergy, the implications are clear: such infrastructure requires significant electrical capacity, resilient data centers, stable networks, and secure facilities. While the order does not explicitly address the grid, the underlying infrastructure demands will inevitably interact with regional and national energy systems. ### **Timelines and Milestones: A Phased Approach** The order sets multiple deadlines: - 90 days: Identify federal computing and networking resources, including cloud partnerships. - 120 days: Identify initial datasets, model assets, and risk-based cybersecurity measures. - 240 days: Review capabilities at national laboratories for AI-directed experimentation. - 270 days: Demonstrate an initial operating capability aimed at one national challenge. These timelines indicate an intent to move rapidly, but actual implementation will depend on appropriations, infrastructure readiness, and interagency cooperation. The milestones serve primarily as organizing mechanisms rather than strict indicators of scientific output. ### **Identifying National Challenges: A Domain-Based Approach** Section 4 directs DOE to produce a list of at least twenty scientific and technological challenges that could benefit from the Genesis Mission. These domains reflect national priorities identified in the 2025 National Science and Technology Memorandum, including: - advanced manufacturing; - biotechnology; - critical minerals and materials; - nuclear fission and fusion; - quantum information science; and - semiconductors and microelectronics. This list frames the Mission around applied scientific challenges with economic and strategic implications. AIxEnergy notes that many of these areas are deeply interconnected with energy systems and compute requirements. For example, semiconductor manufacturing requires stable high-quality power; fusion research involves intensive simulation workloads; and critical-mineral discovery depends on geospatial modeling and predictive analytics. ### **Interagency and External Collaboration: Practical Governance Structures** Section 5 outlines the mechanisms for interagency coordination and collaboration with external partners. Key provisions include: - aligning federal AI programs to avoid duplication; - integrating available agency datasets into the Platform; - facilitating funding opportunities and prize competitions; - enabling federal research facilities to host students and early-career researchers; and - developing standardized frameworks for partnerships. These measures reflect a recognition that scientific AI development requires broad collaboration among government, industry, and academia. The order also emphasizes security, including classification controls, cybersecurity requirements, export controls, and vetting of users. From a neutral perspective, these guardrails are consistent with existing federal research governance. They acknowledge both the potential benefits and the risks of integrating sensitive data and powerful computational models. ### **Implications for Compute, Energy, and Infrastructure** Although the executive order focuses on scientific discovery, its implementation will require substantial infrastructure support. AIxEnergy highlights several areas where energy systems intersect with the Mission: - High-performance computing and training clusters require large and stable electricity supplies. - Secure cloud-based AI environments may rely on commercial data centers with specific interconnection and resilience needs. - Autonomous laboratories and production facilities depend on high-reliability power. - Domain-specific foundation models may require multiple training cycles, creating ongoing compute demand. Because the infrastructure supporting AI is energy-intensive, utility partnerships and grid modernization will play a supporting role even if they are not explicitly named in the order. This dynamic aligns with broader national trends, where AI is increasingly influencing load forecasts and regional planning processes. ### **Evaluating the Mission: Opportunities and Constraints** A clear-eyed assessment recognizes both the potential benefits and the practical limitations of the Genesis Mission. Opportunities: - It may accelerate scientific discovery across multiple disciplines. - It improves coordination across federal datasets and compute assets. - It strengthens national laboratory capabilities. - It offers a platform for public–private collaboration. Constraints: - Implementation depends on appropriations and infrastructure readiness. - Grid and data-center capacity may limit compute expansion. - Cybersecurity risks increase as resources are federated. - The initiative requires sustained leadership across administrative transitions. These factors make the Genesis Mission a long-term endeavor rather than a short-term transformation. ### **What Success Requires—and the Challenges Ahead** For the Genesis Mission to achieve its stated goals, several conditions will be essential. First, sustained federal funding is required to support the expansion of compute resources, modernization of data infrastructure, and secure operation of autonomous research environments. Second, long-term coordination across agencies must persist beyond administrative cycles, ensuring continuity in standards, interoperability, and governance. Third, the nation will need significant upgrades in energy and data‑center infrastructure, since the computational intensity of scientific AI will place new pressures on grid reliability, transmission capacity, and siting processes. Finally, the Mission will hinge on secure integration of sensitive data—balancing scientific openness with national security requirements. The challenges are equally substantial. Implementing a platform of this scale may encounter delays in procurement, interconnection, or regulatory approvals. Cybersecurity threats will increase as datasets and compute systems are federated across facilities. Workforce capacity—in AI engineering, cybersecurity, and advanced scientific modeling—may limit speed of deployment. And scientific progress will depend on the ability to coordinate diverse stakeholders with different missions, incentives, and risk tolerances. These obstacles do not diminish the Mission’s potential, but they underscore that success hinges on aligning technical, institutional, and energy‑system capabilities over a multi‑year horizon. ### **Conclusion:** A National Scientific Platform with Energy and Compute Implications The Genesis Mission, as outlined in the executive order, seeks to create a national infrastructure for AI-enabled scientific discovery. It integrates DOE leadership, federal datasets, computing resources, and research facilities into a coherent framework. While the document carries the tone of a presidential directive, its operational components are technical and administrative in nature, focused on building secure, scalable scientific infrastructure. From an AIxEnergy perspective, the most significant implication is structural: AI-driven science requires energy, compute, and secure data at scale. The Genesis Mission does not explicitly address the role of the grid, but the underlying infrastructure depends on reliable electricity and advanced compute environments. As the Platform develops, energy systems and scientific systems will become increasingly interdependent. In this way, the Genesis Mission marks an evolution in how the United States organizes its scientific enterprise—not as a rhetorical gesture, but as a coordinated attempt to adapt federal research infrastructure to the era of artificial intelligence. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### The Utility-AI Leadership Edge Part I: How Utility COOs Can Harness AI to Deliver Resilience, Reliability, and Growth URL: https://www.aixenergy.io/the-utility-ai-leadership-edge-part-i-how-utility-coos-can-harness-ai-to-deliver-resilience-reliability-and-growth/ Last updated: 2026-07-29T19:41:05.000Z *This is the first part in a four part article series exploring how utility C-suite leaders can harness the power of AI to deliver resilience, reliability, and growth.* Imagine you are a North American utility COO in 2025\. You are responsible for delivering power to millions of residential and commercial customers—yet every lever of your operating environment is under unprecedented strain. Electrification is accelerating faster than forecast. Data centers are consuming capacity at a scale unseen in decades. Extreme weather is turning into a year-round condition. And the infrastructure you rely on was built for a different era entirely. Article Summary 0:00 /69.172245 1× Meanwhile, customer expectations continue to climb. The tolerance for outages—momentary or sustained—is shrinking. Reliability is no longer measured only in SAIDI and SAIFI; it is measured in economic continuity, community trust, and brand credibility. The mandate is unmistakable: deliver power that is more reliable, more resilient, more dynamic, and more affordable—while managing risk, cost, and a workforce under pressure. Yet even the best-run operating models were not designed for today’s level of volatility or velocity. What the most forward-facing COOs now understand is that the defining challenge is not operational excellence alone. It is operational intelligence—the ability to see, predict, and adapt the system in real time. And the most powerful lever for that shift is artificial intelligence. This article—Part I of the series—examines how AI is already transforming the operational core of utilities and giving COOs the ability to move from reactive management to predictive, adaptive, and optimized systems. ## Power Delivery Under Pressure: Moving from System Awareness to System Prediction Every COO knows the truth: the grid is now the bottleneck for nearly every strategic initiative. EV adoption, DER interconnection, wildfire mitigation, load growth from hyperscale data centers—none of these stall because of generation. They stall because of delivery. Five years ago, operators could see the system in fragments. Today, AI enables continuous situational awareness and an unprecedented degree of system foresight. How Leading COOs Are Using AI in Power Delivery - Predictive load-flow modeling that integrates DER behavior, customer usage patterns, weather forecasts, and real-time performance—surfacing overload conditions before the system strains, and recommending alternative switching strategies. - AI-assisted substation monitoring that identifies anomalous transformer or breaker behavior weeks or months before visible degradation. - Dynamic hosting-capacity analytics that unstick DER interconnections by identifying safe, flexible capacity without needing broad, costly upgrades. - AI-enabled switching and recloser optimization that improves fault isolation and accelerates sectionalization as load increases. ### What These Tools Deliver? More than anything, they deliver time—the one variable operators can never manufacture. Time to plan instead of react. Time to redesign circuits with accuracy. Time to allocate crews where they matter most. AI is not replacing operational judgment. It is making that judgment exponentially stronger, more precise, and more actionable. ## Resilience & Storm Risk: From Post-Event Response to Pre-Event Preparedness Extreme weather is no longer episodic; it is embedded. A single storm can cost a utility hundreds of millions of dollars in restoration costs, service interruptions, and reputational damage. Over the past decade, COOs have become experts at storm response. The shift now is becoming masters of storm prediction and prevention. AI is proving to be the inflection point enabling this pivot. Where AI Is Strengthening Storm and Resilience Operations - Pre-storm outage prediction models that integrate wind forecasts, flooding projections, vegetation conditions, asset age, soil saturation, and micro-climate data—pinpointing the feeders and assets most likely to fail. - Vegetation-risk analytics that go beyond location and identify *hazard zones*, directing trimming budgets toward spans with the highest outage probability. - AI-driven crew routing and staging that minimize travel time and accelerate restoration under rapidly evolving field conditions. - Automated drone and satellite damage assessment, processed within minutes, and delivered directly into OMS and work-management platforms. ### The Impact No AI system will ever eliminate weather risk. But AI can significantly reduce the duration, cost, and scale of storm impacts. It allows COOs to move from triage to preparedness—acting hours or days earlier with clarity and confidence. This shift is not theoretical. It is already underway in the utilities reshaping industry best practice. ## Managing Explosive Load Growth: From One-Way Planning to Adaptive, AI-Native Planning The load growth curve unfolding across the U.S. is rewriting the role of the COO. After decades of relatively flat demand, utilities are entering a new era where 2–3% annual growth becomes 5–7%—with pockets of 20–40% growth in regions experiencing hyperscale data center expansion, manufacturing reshoring, and deep electrification. Traditional planning cycles—deterministic studies, annual refreshes, reinforcement-heavy portfolios—cannot keep pace with growth of this magnitude. AI enables a fundamentally different operating environment. How AI Brings Planning and Operations Together - Adaptive planning models that update continuously as new data enters the system—customer pipelines, interconnection queues, DER growth, transformer loading, substation constraints, and feeder performance. - AI-enabled non-wires alternative identification, spotlighting where virtual power plants, targeted demand flexibility, or feeder-level optimization can defer or avoid costly capital upgrades. - Scenario-based investment portfolios that help COOs and CFOs jointly evaluate capital decisions under multiple weather, load, and economic futures. - Grid digital twins that allow teams to test capital plans, maintenance strategies, load growth, and operational constraints *before* investing dollars or deploying crews. ### What This Means for COOs This is not merely technology transformation. It is business transformation. AI empowers COOs to grow system capacity faster, with higher accuracy, and at significantly lower cost than traditional reinforcement alone. It turns the grid from a static asset into a living system—one that adapts to demand instead of being overwhelmed by it. Where COOs Need to Lead: The Human and Organizational Shift AI adoption is not a software exercise. It is a leadership exercise. The most successful COOs are distinguishing themselves in four key ways: 1. They set an operational AI agenda—anchored in reliability, resilience, safety, and affordability—not a generic “innovation” agenda. 2. They build AI-literate field and control-room teams, using training, simulation, and change management to ensure frontline operators trust and understand the tools. 3. They invest in high-quality operational data, especially around assets, vegetation, and real-time conditions, recognizing that AI only performs as well as the data behind it. 4. They co-own the strategy with CIOs, CFOs, and customer organizations, ensuring AI solutions strengthen reliability, manage cost trajectories, and support long-term growth. Operators trust authenticity and clarity. The COO’s voice carries authority because they own the systems, the crews, and the customer experience. When COOs champion AI, the organization listens. ## The Moment for AI-Driven Operations Has Arrived Utilities face extraordinary pressure—and extraordinary opportunity. AI is not a silver bullet, but it is the most powerful operational tool the industry has had in decades. The COOs who embrace this shift will deliver grids that are safer, more resilient, more predictable, and more prepared for the economic and social growth ahead. They will be the leaders who move their organizations from managing the grid to mastering it. In a decade defined by electrification, volatility, and exponential demand, mastery is no longer a goal—it is the mandate for operational excellence. In the end, the portrait that emerges is of a utility operating landscape being remade at a pace that outstrips the tools of the past. COOs are discovering that the constraints they face—runaway load growth, weather volatility, aging infrastructure, and demanding customers—are not solvable with incremental process improvements or legacy planning cycles. They require a shift from system awareness to system prediction, from static design to adaptive operation, from episodic response to continuous readiness. Artificial intelligence is the catalyst for that shift, giving operators time, foresight, and optionality in a world where every decision carries higher stakes. The implication is unmistakable: utilities that embrace AI-native operations will accelerate capacity, reduce risk, and strengthen reliability at a cost and speed unachievable through traditional means. Those that resist will find themselves increasingly outpaced by the demands of electrification and the expectations of the communities they serve. ## What’s Ahead in Parts II–IV This Part I article established the operational foundation: how AI is reshaping grid intelligence, real-time operations, storm preparedness, and load-growth planning for COOs. The next three articles will explore the role that other C-Suite executives, Chief Customer Officer, Chief Public Affairs/Regulatory Officer and Chief Financial Officer, play in the utility-AI transformation. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### Managing Data Center Uncertainty Part III — The Utilization Paradox: Scarcity and Waste Inside AI Infrastructure URL: https://www.aixenergy.io/managing-data-center-uncertainty-part-iii-the-utilization-paradox-scarcity-and-waste-inside-ai-infrastructure/ Last updated: 2026-07-29T19:42:11.000Z *This article is part of the AIxEnergy Series: Managing Data Center Energy Uncertainty, drawn from the author’s full research paper. The complete version is available directly from the author by request at* *michael.leifman1@gmail.com* *.* In [Part I](https://www.aixenergy.io/part-i-the-uncertainty-problem/) of this article series, we discussed how AI is driving unprecedented uncertainty in electricity demand, with U.S. data-center use swinging between 325 and 580 TWh by 2028\. The issue isn’t forecasting—it’s governance. Without transparency, flexibility, and better incentives, utilities overbuild, costs rise, and emissions lock in. In [Part II](https://www.aixenergy.io/managing-data-center-uncertainty-part-ii-phantom-data-centers-how-strategic-opacity-drives-overbuild/) we of this article series, reviewed how phantom data centers distort U.S. energy planning. Developers overfile interconnection requests, utilities profit from overbuilding, and regulators approve speculative capacity. Misaligned incentives create costly overbuild and fossil lock-in—requiring governance reform, transparency, and accountability. Article Summary 0:00 /97.697959 1× This article explores the gap between apparent scarcity and actual inefficiency, arguing that the energy crisis of AI is not one of shortage but of synchronization. Drawing on research from Lawrence Berkeley National Laboratory, Duke University, and detailed technical studies from Google and Microsoft, this article reveals that GPU utilization across AI data centers averages just 60-70 percent. The culprit is a fundamental infrastructure mismatch: unoptimized training deployments can experience 30-50% GPU idle time waiting for data, with data preprocessing consuming up to 65% of epoch time in the worst cases. The solution is not to build more power but to use power more intelligently. By distinguishing between training and inference workloads and recognizing which facilities possess genuine operational flexibility, the grid can unlock significant capacity. Duke University research demonstrates nearly 100 GW of potential load integration using existing infrastructure if data centers reduce consumption by just 0.5-1% during peak periods, while inference workloads require different mechanisms focused on electrical demand management via storage. ## The Architecture of Inefficiency 💡 AI data centers consume enormous power yet routinely run at only 60–70 percent GPU utilization because storage and preprocessing cannot feed data fast enough, creating idle cycles that utilities cannot see and forcing the grid to plan for load that is far less efficient than it appears. AI data centers are vast ecosystems of silicon and cooling fluid. Inside, compute clusters built to train large language models draw more power per rack than an entire suburban neighborhood. Yet these same clusters often sit partially idle. Unlike legacy industrial facilities, AI systems operate in pulses—bursts of intense computation followed by intervals where GPUs wait for data. During active training, GPUs operate at high capacity. But these periods are punctuated by delays as storage systems struggle to deliver training data fast enough. Research reveals that unoptimized AI training deployments can experience 30-50% GPU idle time waiting for data, with data preprocessing consuming up to 65% of epoch time in the worst cases, achieving only 60-70% GPU utilization—not because computational demand is lacking, but because data delivery infrastructure cannot keep pace with GPU processing speed. Major cloud providers acknowledge this reality. Aggregate GPU utilization across hyperscale facilities rarely exceeds this 60-70% range. The remaining capacity goes unused while GPUs wait for preprocessing, network transfers, or storage systems to deliver the next batch of training data. Utilities, however, cannot see this inefficiency. From their perspective, the data center appears to be a continuous load. The grid must plan for peak capacity as though it were constant. ## The I/O Bottleneck: Where Billions in Compute Power Stalls 💡 Data-starvation research shows that AI training is often bottlenecked by slow storage and preprocessing—not computation—leaving costly GPUs idle for 30–65 percent of training time and pushing real-world utilization far below theoretical capacity. The technical reality beneath these utilization figures reveals why even expensive AI infrastructure operates far below its theoretical capacity. Studies from Google and Microsoft document that model training time is often dominated not by computation but by data delivery—a phenomenon researchers call “data starvation.” A 2021 analysis of deep neural network training at scale found that workloads spend up to 65% of each training epoch simply on data preprocessing—loading training examples, applying transformations, and shuffling batches—before GPUs can perform any calculations. This preprocessing burden creates systematic bottlenecks that cascade through the entire training pipeline. The causes are structural. Storage throughput limitations mean GPUs process data orders of magnitude faster than storage systems can deliver it. Large language model training requires sustained read performance exceeding 10 GB/s, while computer vision training demands even higher throughput. Traditional storage infrastructure often cannot meet these requirements, creating systematic delays where expensive GPUs—costing $30,000-40,000 per H100 unit—sit idle waiting for data to arrive. Geographic separation compounds the problem. Cloud training has become standard practice, but distributed architectures mean slow I/O when computation and storage span different regions. A Google study documented that training workloads spend on average 30% of total training time simply on input data pipeline operations—time where GPUs contribute nothing to model improvement but continue consuming power. CPU preprocessing bottlenecks create additional delays. During training, CPUs handle data augmentation, normalization, and batch preparation. When CPUs reach maximum utilization preparing the next batch, GPUs sit partially idle waiting for work. The random access patterns required for training deep neural networks using Stochastic Gradient Descent create particularly punishing demands on storage infrastructure. Systems must read terabytes of data just to use gigabytes, while expensive GPUs idle. The economic implications are substantial. Poorly optimized data pipelines can reduce GPU utilization to 40-60%, directly impacting both training speed and infrastructure return on investment. Organizations with optimized data loading achieve 90%+ GPU utilization during training—but this requires significant engineering investment in custom pipelines, distributed caching strategies, and infrastructure co-design that most organizations defer. ## The I/O Bottleneck as Flexibility Indicator 💡 The I/O bottleneck exposes which AI workloads are truly flexible, showing that most training jobs can shift in time while real-time inference cannot, and without transparency about this distinction utilities misclassify all data-center load as firm—driving costly overbuild that smarter, workload-aware planning could avoid. This technical reality has profound implications for grid planning. The I/O bottleneck reveals which data center workloads possess inherent temporal flexibility. A training job that requires 48 hours due to data delivery constraints is fundamentally different from latency-sensitive inference serving real-time user requests. The distinction is not merely technical but economic: if a training workload were genuinely time-critical, the organization would invest the significant engineering resources required to optimize data pipelines and achieve 90%+GPU utilization. The fact that many organizations accept 60-70% utilization indicates these workloads can tolerate substantial delays. Deferring a training job’s start by a few hours to avoid grid peaks adds modest calendar time to multi-day runs already experiencing delays from I/O constraints. Organizations that accept multi-day training times are demonstrating revealed preference: these workloads are not latency-sensitive. The I/O bottleneck doesn’t prevent flexibility; it identifies which workloads can provide it most readily. Duke University research demonstrates the scale of this opportunity. Their analysis found that nearly 100 GW of new large loads could be integrated across the 22 largest U.S. balancing authorities with minimal impact if data centers can reduce power usage by 0.5-1% during peak demand periods. For the 0.5% curtailment rate, averaging about 2.1 hours per event over 177 hours per year, the existing grid could handle much of expected demand growth with no additional generation. Importantly, on average for 88% of curtailment time, half the new load would continue running—meaning partial rather than complete shutdowns. However, this flexibility opportunity applies primarily to training workloads. For inference operations serving real- time user requests, achieving grid flexibility requires electrical demand management via battery storage that maintains computational continuity while reducing peak electrical draw, rather than direct computational curtailment that would violate service requirements. The problem is institutional. Workloads remain opaque to regulators, and utilities treat all megawatts as firm. Without reliable data on training versus inference workload composition, utilities cannot design targeted demand response programs and regulators cannot assess whether proposed facilities can provide grid services. The Cost of Misclassification Utilities treat AI data centers as steady, predictable, and inflexible. But AI is dynamic, algorithmic, and constantly evolving. By classifying data centers as firm industrial load, regulators force utilities to build for a world that no longer exists. This misclassification has profound consequences. Utilities commit to constructing generation and transmission assets assuming 24/7 operation at design capacity, even though real-world utilization rarely justifies it. In Virginia, utility forecasts incorporated speculative AI facilities at near-continuous utilization. Half lacked financing or operational schedules. Yet those assumptions justified billions in new gas-fired generation. The same story repeats across Georgia, Texas, and Oregon. A more rational classification would tie load designations to computational behavior—training versus inference, flexible versus firm. This would align planning with technical realities and create incentives for data centers to disclose operational flexibility. Transparency is the foundation of efficiency. ## From Waste to Value 💡 In the AI era, grid reliability hinges on how efficiently power is used, not how much is built, demanding transparency on GPU utilization, performance-based incentives that reward reduced waste, and flexibility markets tailored to the distinct behaviors of training and inference workloads. For a century, the energy sector equated reliability with capacity. In the AI age, reliability will depend on utilization—on how effectively each watt converts into useful computation. To make this transition real, three governance mechanisms stand out: First, utilization disclosure. Require all AI facilities above 50 MW to report GPU utilization, workload composition (training versus inference percentages), and flexibility potential during interconnection approval. Regulators can then distinguish between continuous and shiftable load. Second, performance-based regulation. Extend PBR frameworks to reward utilities for verified reductions in system waste—deferring capacity expansions through coordinated management of data center loads where technically feasible. This converts efficiency into a profit center rather than a cost. Third, differentiated flexibility markets. Create demand response mechanisms that reward operators for aligning load timing with grid conditions where technically feasible (primarily for training workloads), while recognizing that inference workloads require different approaches such as electrical storage. ## The Governance Imperative 💡 The grid’s real challenge is not excess demand but opaque, inefficient AI workloads, and true resilience will come from transparency and governance that reward efficiency, distinguish workload types, and unlock the vast flexibility already hidden inside today’s underused GPUs. The utilization paradox exposes a simple truth: the grid’s greatest risk is not overuse, but misuse combined with opacity. Energy institutions struggle to govern a world lacking the data needed to distinguish between workload types. Regulators equate prudence with overbuild, and utilities profit from expansion. Yet the path to resilience lies in visibility, accuracy, and coordination. To govern the AI age, policymakers must reward efficiency per watt rather than megawatts built. They must treat transparency as infrastructure and recognize that different workload types require different flexibility mechanisms. Training workloads offer temporal flexibility potential; inference workloads require electrical flexibility via storage. Both contribute to grid resilience through fundamentally different mechanisms. The takeaway is clear: the cleanest power plant is the unused GPU. Unlocking that efficiency requires not more technology, but more transparency and governance that distinguish between workload types and create appropriate economic incentives for each. ## What’s Next Part III has revealed that what looks like scarcity is, in many cases, inefficiency masked by opacity—a failure of synchronization between computation and electricity, compounded by lack of data on workload types. Part IV will detail the specific policy mechanisms that can transform this inefficiency into value. By showing how targeted interventions can unlock grid headroom while respecting different workload characteristics, it will argue that the next frontier of reliability is not generation—it is intelligence paired with transparency. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## References ## Lawrence Berkeley National Laboratory. *2024 United States Data Center Energy Usage Report.* Berkeley, CA: LBNL, 2024\. LBNL-2001637. Norris, T. H., D. Patiño-Echeverri, and M. Dworkin. *Rethinking Load Growth: Assessing the Potential for Integration of Large Flexible Loads in U.S. Power Systems.* Durham, NC: Duke University Nicholas Institute, 2025. Mohan, Jayashree, et al. “Analyzing and Mitigating Data Stalls in DNN Training.” *Proceedings of the VLDB Endowment* 14 (2021). Accessed \[date\]. [https://www.cs.utexas.edu/\~vijay/papers/vldb21-datastalls.pdf](https://www.cs.utexas.edu/~vijay/papers/vldb21-datastalls.pdf?ref=aixenergy.io). Spjut, Josef, and Ted Purcell. “The New Bottlenecks of ML Training: A Storage Perspective.” *SIGARCH Blog*, July 2021\. Accessed \[date\]. [https://www.sigarch.org/the-new-bottlenecks-of-ml-training-a-storage-perspective/](https://www.sigarch.org/the-new-bottlenecks-of-ml-training-a-storage-perspective/?ref=aixenergy.io). Leifman, Michael. *Managing Data Center Energy Uncertainty: A Framework to Prevent Overbuild, Control Costs, and Unlock Grid Flexibility.* Washington, DC: AIxEnergy, 2025\. Available from the author at [michael.leifman1@gmail.com](mailto:michael.leifman1@gmail.com). ### The Age of Electricity Begins: AI Is Rewriting the Global Grid, Says IEA URL: https://www.aixenergy.io/the-age-of-electricity-begins-ai-is-rewriting-the-global-grid-says-iea/ Last updated: 2026-07-10T17:35:48.000Z The International Energy Agency’s *World Energy Outlook 2025* reads like a dispatch from a grid that is being quietly rewritten by code. The report was conceived as a global energy survey, but buried in its tables and scenario charts is something more specific: the first full attempt by a major institution to quantify what the rise of artificial intelligence and data centres will do to electricity systems, fuels, and investment over the next decade and beyond. 0:00 /1:41 1× For AIxEnergy, this is home turf. What follows is a fully WEO-aligned re-telling of that story: how much power AI is likely to consume, where the demand will land, what fuels will be pulled into service, and how AI itself pushes back against the very energy pressures it creates. Every number and factual claim below is drawn directly from *World Energy Outlook 2025* unless otherwise noted, and citations are provided inline. ## **From oil shocks to inference shocks** The WEO was born in the aftermath of the 1970s oil crises, when barrels and tankers defined energy security. In 2025, the report opens on a different landscape: energy is still geopolitical, but the points of vulnerability now stretch from oil and gas fields to critical minerals, transmission grids, and the neural racks of hyperscale data centres. The IEA is explicit that we are entering an “Age of Electricity,” in which electricity demand grows much faster than total energy use. In both its Current Policies Scenario (CPS) and Stated Policies Scenario (STEPS), global electricity demand rises by about 40 percent between 2024 and 2035, to roughly 37,800 terawatt-hours, while overall energy demand grows more slowly. Within that expansion, AI and data centres are not the largest driver, but they are among the fastest and most concentrated. They are also the least negotiable: unlike home air conditioners or industrial motors, a modern AI cluster cannot simply be “turned down” in the middle of a training run. ## **The $580 billion pivot: when data overtakes oil** The hinge of the AI story in WEO 2025 is not a single gigawatt figure but a pair of investment numbers. In 2025, the IEA estimates that global investment in data centres reaches around 580 billion United States dollars, surpassing the roughly 540 billion dollars projected for global oil supply investment in the same year. This crossover is more than symbolism. For half a century, upstream oil spending was one of the main barometers of energy security and macroeconomic risk. The IEA’s comparison signals a structural shift: capital is increasingly flowing into digital infrastructure that sits electrically “downstream” of generation, yet has the power to reshape generation itself. In the AIxEnergy frame, this is the moment when electricity stops being just the fuel for computation and becomes the constraint that defines how fast intelligence can scale. ## **How much power AI actually uses** The WEO’s quantitative core on AI sits in section 1.5, “How much difference will AI make to the future of energy?” There, the IEA does two important things at once: it resists sensationalism while acknowledging the scale of the shift. On the demand side, the agency projects that electricity consumption by AI-optimised servers increases fivefold by 2030 as data centre capacity surges. This surge in AI-tuned hardware is enough to double total data-centre electricity consumption by 2030 relative to 2024. Yet even after this doubling, data centres as a whole account for less than 10 percent of global electricity demand growth between 2024 and 2030 in the STEPS. Other sources—industrial loads, electric vehicles, air conditioning, and other end-uses—still dominate the incremental growth of electricity demand. Looking out to 2035, the Executive Summary crystallises the story: electricity consumed by data centres roughly triples by that date, but still represents under 10 percent of total global electricity demand growth across the period. In other words, AI is important not because it is the main driver of global load, but because it is a fast-growing, highly concentrated sliver that arrives precisely where grids are already taut. ## **Geography: three regions, one bottlenecked grid** The WEO is unusually clear about the spatial pattern of this demand. Data centres today are geographically concentrated, with the United States, China, and Europe accounting for about 82 percent of global capacity. Over the next few years, more than 85 percent of new capacity additions are expected in these same three regions. In China and the European Union, data centres account for an estimated 6 to 10 percent of electricity demand growth to 2030\. In the United States—the world’s largest data-centre market—they are far more dominant, accounting for around half of electricity demand growth over the same period. A geospatial analysis in the report adds further texture. More than half of data centres under construction or announced are being developed in or near cities of at least one million people, where grids already serve large loads. About 55 percent of the new centres in the pipeline are larger than 200 megawatts; once fully online, each of those facilities will consume as much electricity annually as roughly 200,000 households. Nearly two-thirds of future capacity is sited in existing data-centre clusters rather than new greenfield locations. The result is a kind of digital urbanism that rides atop twentieth-century grid topology. Load arrives not as a diffuse swell but in discrete, multi-hundred-megawatt chunks at the edges of metropolitan substations whose transformers and rights-of-way were never designed with AI in mind. ## **Queues, delays, and the new face of “speed to power”** By the time we reach the middle of chapter 1, the WEO turns from demand to friction. Grid congestion and connection queues for new data centres are already lengthening in many regions. In the United States, connection times are typically in the range of one to three years, but in northern Virginia—the densest data-centre cluster on Earth—they can stretch to seven years. Europe is not immune. In the United Kingdom and parts of continental Europe, average queue times have been reported at seven to ten years. In Dublin, a major European data-centre hub, new connection requests have been paused until 2028\. Supply chains for key equipment—transformers, cables, gas turbines, and the critical minerals embedded in them—are already under pressure. Taking these bottlenecks together, the IEA estimates that about 20 percent of projected data-centre additions to 2030 could be at risk of delay. For AIxEnergy, this is the core strategic constraint. The limiting factor is no longer the willingness of capital to build compute, but the ability of physical infrastructure to deliver electrons to it. “Speed to power,” in the WEO’s language, becomes the determining variable for how fast AI capacity can scale, especially in the United States and Europe. ## **Where the extra electricity comes from** On the supply side, the report asks a straightforward question: if data centres are going to double or triple their electricity use, which technologies actually provide that incremental power? In the STEPS, renewables remain the leading source of additional electricity for data centres. Between now and 2035, they provide around 45 percent of the growth in electricity consumed by these facilities, equivalent to nearly 400 terawatt-hours of extra renewable output dedicated to data-centre load. In the more conservative CPS, renewables still contribute the largest share of incremental supply, though their share falls modestly to about 40 percent. Natural gas also plays a critical role, particularly in the United States and the Middle East. Globally, gas-fired generation serving data-centre demand increases by about 220 terawatt-hours to 2035 in the STEPS and by roughly 285 terawatt-hours in the CPS. This growth comes despite a surge in orders for new gas turbines over the past two years, which has strained supply chains and lengthened delivery times and costs. Nuclear power emerges as a third, quieter pillar. With renewed policy support and rising interest from technology companies, nuclear plants—both existing reactors and new designs such as small modular reactors—are increasingly being positioned as dedicated power sources for data centres. The WEO notes that in the STEPS, nuclear provides around 190 terawatt-hours of additional electricity for data centres by 2035, and records the first power-purchase agreement between a data-centre operator and a demonstration-scale small modular reactor. In the AIxEnergy language, the pattern is clear: renewables set the direction of travel, gas provides much of the dispatchable flexibility, and nuclear offers long-duration, zero-carbon baseload for hyperscale clusters that cannot tolerate volatility. ## **AI as a demand shock—and an efficiency engine** The WEO is careful not to treat AI solely as a source of new load. Section 1.5.3, “AI for energy,” turns the lens around and asks what AI does inside the energy system itself. Here, the IEA’s modelling suggests that widespread uptake of AI-enabled solutions across the energy sector—optimised logistics in oil and gas, predictive maintenance in power plants and grids, smarter control of industrial processes and mobility—could boost efficiency by roughly 3 to 10 percent across the transport and industry sectors globally by 2035\. The aggregate energy savings from this optimisation are on the order of 13.5 exajoules, slightly larger than the total energy demand of Indonesia today. This is the paradox at the heart of AIxEnergy. The same compute that drives new power demand can, if appropriately governed, reduce the amount of primary energy required to deliver a given level of economic output. In macro terms, AI is both a demand accelerator and a potential brake—an inference engine that, in principle, can help the grid think its way into a more efficient configuration. ## **Risk hierarchy: from critical minerals to climate overshoot** The AI story in WEO 2025 also sits within a broader risk hierarchy. The report spends significant attention on critical minerals, noting that a single country is the dominant refiner for nineteen out of twenty strategic energy-related minerals, with an average market share of about seventy percent. Many of those minerals feed directly into the equipment that will serve AI loads—grid transformers, power electronics, batteries, and the chips that sit on AI accelerator boards. At the same time, the IEA is blunt about climate trajectories. In the STEPS, energy-related carbon dioxide emissions fall below 30 gigatonnes by mid-century, but still leave the world on a path toward roughly 2.5 degrees Celsius of warming by 2100\. In the CPS, emissions remain higher and are broadly consistent with almost 3 degrees of warming. In the Net Zero Emissions by 2050 (NZE) Scenario, the 1.5-degree threshold is overshot for several decades before being re-approached late in the century via deep decarbonisation and large-scale carbon-dioxide removal. AI, in the WEO framing, does not rescue us from these trajectories by itself. It intensifies pressure on the grid in the near term and offers tools for mitigation and efficiency, but the decisive variables remain policies, investment patterns, and the speed at which low-emissions technologies are built out and integrated. ## **What this means for AIxEnergy** When the IEA writes that “explosive growth in electricity demand for data centres and AI is concentrated in advanced economies and China,” it is summarising a story that is both reassuring and alarming. Reassuring, because even in the central scenarios, AI does not blow up the global power system; its share of incremental electricity demand remains under 10 percent. Alarming, because that relatively small share is focused in specific corridors—Northern Virginia, Dublin, the Rhine-Ruhr, coastal China—precisely where grids are already strained, land is contested, and queues are stretching from years toward decades. For AIxEnergy, the implications are clear: First, the centre of gravity of AI-driven electricity demand is the United States, China, and the European Union, with the United States in particular seeing data centres account for about half of projected electricity demand growth to 2030 in the STEPS. Any serious “speed to power” strategy must therefore start with these regions’ permitting, interconnection, and grid-planning regimes. Second, the composition of the marginal kilowatt-hour for data centres will be set by the interplay of renewables buildout, gas-turbine availability, and the politics of nuclear restarts and SMR deployment. Renewables are on track to supply roughly 40 to 45 percent of additional data-centre demand by 2035; gas and nuclear fill most of the rest. The carbon intensity of AI will follow. Third, AI itself is a lever. A 3 to 10 percent efficiency gain across global transport and industry, corresponding to about 13.5 exajoules of avoided energy demand, is not a rounding error; it is an entire Indonesia’s worth of energy carved out of the system. Whether that potential is realised depends on governance, incentives, and the degree to which AI tools are aimed at optimisation rather than pure scale. Finally, the WEO reminds us that while AI is the most novel of the new energy loads, it is not the only one. The same scenarios that double or triple data-centre demand also see rising electricity use for cooling, vehicles, and advanced manufacturing. The AI question, in other words, is not “Will the grid survive?” but “Who will get to plug in, on what terms, and at whose expense?” Those are the questions AIxEnergy was built to interrogate. The *World Energy Outlook 2025* does not close the debate; it gives us the most rigorous, globally consistent baseline yet for understanding the energy economics of AI. From here, the task is to translate those scenario curves into investment decisions, regulatory frameworks, and governance architectures that keep the intelligence boom aligned with a stable, decarbonising grid. **International Energy Agency.** [***World Energy Outlook 2025***](https://iea.blob.core.windows.net/assets/9228d782-4207-4648-9c75-e6bb908cdb90/WorldEnergyOutlook2025.pdf?ref=aixenergy.io)**. Paris: OECD/IEA, 2025.** ### The AI Factory? Artificial Intelligence, Manufacturing, and the Energy Realities of Reindustrialization URL: https://www.aixenergy.io/the-cognitive-factory-ai-manufacturing-and-the-energy-realities-of-reindustrialization/ Last updated: 2026-06-30T02:27:21.000Z AI is reshaping U.S. manufacturing into a data-driven, energy-intensive ecosystem. Reindustrialization hinges on reliable, low-carbon power and grid modernization—linking America’s industrial competitiveness to its energy infrastructure. _This post is for paying subscribers only._ ### Powering the Intelligence Age: How Hyperscalers Are Reviving U.S. Nuclear Plants to Fuel AI URL: https://www.aixenergy.io/powering-the-intelligence-age-how-hyperscalers-are-reviving-u-s-nuclear-plants-to-fuel-ai/ Last updated: 2026-06-30T02:27:13.000Z Tech giants are turning to nuclear power to fuel AI. Google, Amazon, and Microsoft are backing reactor restarts and next-gen designs—reviving U.S. plants in Iowa and Pennsylvania to secure 24/7 carbon-free energy and anchor America’s digital infrastructure for the AI era. _This post is for paying subscribers only._ ### Managing Data Center Uncertainty Part II — Phantom Data Centers: How Strategic Opacity Drives Overbuild URL: https://www.aixenergy.io/managing-data-center-uncertainty-part-ii-phantom-data-centers-how-strategic-opacity-drives-overbuild/ Last updated: 2026-07-29T19:42:15.000Z ### Preface *This article is part of an AIxEnergy article series: Managing Data Center Energy Uncertainty, drawn from the author’s full research paper, “Managing Data Center Energy Uncertainty: A Framework to Prevent Overbuild, Control Costs, and Unlock Grid Flexibility” (Leifman, 2025). The complete version, which provides more detailed analysis, policy design frameworks, and full citations, is available directly from the author by request* *at michael.leifman1@gmail.com* *and may be shared at his discretion..* Article Summary 0:00 /64.026122 1× U.S. utilities are planning and permitting gigawatts of new generation capacity based on speculative signals from data center developers. Yet industry insiders acknowledge that developers file five to ten times more interconnection requests than projects they intend to build. These “phantom data centers” distort demand projections, inflate capital spending, and entrench fossil generation. The problem is structural: developers profit from optionality, utilities profit from capital investment, and regulators—rewarded for caution—approve capacity to avoid blame for shortages. The result is a feedback loop of overbuild and excess cost. This part of the Managing Data Center Uncertainty article series examines the political economy of opacity that drives this dynamic. It argues that overbuilding is not a product of greed or error but of rational behavior within a flawed system. Developers, utilities, and regulators each act logically under misaligned incentives that collectively yield irrational outcomes. The solution lies in governance reform—through mandatory disclosure of interconnection data, queue accountability, and performance-based regulation that rewards accuracy, not expansion. ## **The Illusion of AI Demand** Utilities across the United States are forecasting unprecedented growth in load, much of it attributed to artificial intelligence and data centers. Yet a significant share of that projected demand may never materialize. Data center developers routinely submit multiple interconnection requests—often five to ten times more than they plan to construct—to secure positions in congested queues and retain future options. Under current market rules, there is little cost to doing so. The Federal Energy Regulatory Commission (FERC) reports that total queued load for U.S. data centers now exceeds 120 gigawatts—more than double any plausible buildout by 2030\. In some regions, more than 80 percent of proposed projects will never reach construction. Still, utilities must plan as if every request is real. Rate-of-return regulation further distorts this process: utilities earn profits on capital expenditures, not on the accuracy of their forecasts. When in doubt, they build. Regulators compound the problem by approving generation and transmission projects aligned to the upper bounds of demand projections. No public official wants to be blamed for a shortfall. The result is systemic overbuild—tens of billions in new capacity justified by speculative demand. ## **Rational Behavior, Irrational Outcomes** Every actor in the current framework behaves rationally. Developers overfile because queue positions are valuable options in an opaque and competitive market. Utilities build because their profit models reward capital expansion. Regulators approve because they are structurally risk-averse. The irrationality lies not in any individual decision but in the incentive structure itself. Virginia’s “Data Center Alley” illustrates this cycle. Between 2022 and 2025, Dominion Energy revised its 15-year load forecast upward by more than 40 percent, driven largely by speculative data center proposals in Loudoun County. Internal estimates reported by *The Economist* (September 2025) indicate that fewer than half of those proposed facilities had secured financing or land. Yet these forecasts underpinned billions of dollars in new infrastructure investment, including gas peaking plants with decades-long lifespans. In Georgia, the Public Service Commission approved more than 6 GW of new natural gas generation in 2025 to serve projected AI-related load. Within six months, developers had delayed or withdrawn nearly half of the associated data center projects. These plants—built for phantom demand—now risk becoming stranded assets. This cycle is not driven by greed but by governance. Every participant acts prudently under a system that mistakes uncertainty for growth. ## **The Queue Problem** Interconnection queues, once designed as planning tools, have become instruments of speculation. Developers can secure queue positions with minimal financial commitment, often long before securing land or financing. Because queue data are not publicly verified, utilities and regulators lack the transparency needed to separate speculative projects from real ones. The opacity of interconnection data converts manageable forecasting uncertainty into systemic volatility. Without transparency into project status, workload type, or operational timeline, planners treat every interconnection request as an impending load. The consequence is a generation pipeline sized for a future that will not exist. FERC’s proposed reforms to interconnection procedures—such as milestone-based deposits and queue audits—are steps in the right direction, but they remain limited in scope. True accountability requires that speculative projects bear the financial risk of inaction. ## **Load Underwriting: A Governance Mechanism for Accuracy** A practical response to speculative interconnection is load underwriting—a regulatory mechanism requiring developers to backstop interconnection requests with proof of intent or partial financial commitment. This approach mirrors the project-finance discipline used in generation development, where offtake agreements and milestone deposits ensure credibility. Load underwriting can take multiple forms: 1. Milestone Deposits: Developers post refundable deposits tied to project progress; funds are forfeited if milestones are missed without cause. 2. Disclosure Requirements: Developers report site control, financing status, and estimated operational timelines. 3. Queue Audits: Regulators or independent system operators regularly review interconnection data to remove inactive or speculative projects. These measures would replace speculation with accountability and convert interconnection queues from opaque wish lists into transparent planning tools. ERCOT in Texas and Dominion Energy in Virginia have already begun testing such mechanisms, linking queue access to verified site control and equipment procurement. Early results suggest these reforms can significantly reduce phantom projects without slowing legitimate development. ## **The Cost of Overbuild** The financial and environmental costs of speculative overbuild are substantial. Each gigawatt of unnecessary capacity costs between $1 and $2 billion to construct. With 40 to 50 GW of phantom load in U.S. forecasts, the potential stranded investment exceeds $70 billion. The environmental implications are equally significant. Utilities frequently meet speculative demand with gas-fired generation because it can be deployed quickly. Once built, these plants operate for decades, locking in emissions and crowding out renewable investment. Overbuild also diverts capital from grid modernization and flexibility—undermining the very resilience the investments are meant to create. ## **Governance Reform: From Expansion to Accuracy** To manage speculative demand, governance must shift from rewarding expansion to rewarding accuracy. Three reforms are critical: 1. *Queue Accountability:* Require deposits and milestone verification for interconnection rights. Projects that fail to progress forfeit their positions. 2. *Transparency and Disclosure:* Mandate public reporting of project viability indicators—site control, financing, workload composition, and operational timelines. 3. *Performance-Based Regulation (PBR):* Align utility incentives with forecasting accuracy. Compensate utilities for cost savings achieved through deferred or avoided capital investment. Together, these reforms would build a market where truth, not speculation, drives infrastructure decisions. ## **Conclusion: From Phantom to Real** Phantom data centers are a symptom of structural misalignment, not individual misconduct. As AI accelerates energy demand, the temptation to overbuild will persist. Governance must therefore evolve to distinguish between potential and probable demand. The challenge is not to eliminate uncertainty but to ensure that uncertainty no longer rewards excess. Each megawatt planned should correspond to a verifiable commitment. Each dollar invested should map to a real, not imagined, future. The next installment, Part III — The Utilization Paradox: Scarcity and Waste Inside AI Infrastructure, will look inside the data center itself—where underutilized GPUs and flexible workloads reveal how the grid’s next 100 gigawatts of capacity could come not from new generation, but from smarter governance of what already exists. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## **References** 1. Brian Martucci, “A Fraction of Proposed Data Centers Will Get Built,” *Utility Dive*, May 15, 2025. 2. *The Economist*, “The Murky Economics of the Data Center Investment Boom,” September 30, 2025. 3. Lawrence Berkeley National Laboratory, *Data Center Energy Use Forecast 2024* (Berkeley, CA: LBNL, 2024). 4. International Energy Agency, *AI Is Set to Drive Surging Electricity Demand from Data Centres*, April 10, 2025. 5. Duke University Nicholas Institute, *Curtailment-Enabled Headroom: Integrating Large Loads with Existing Capacity* (Durham, NC: Duke University, 2024). ### The State Playbook for Equitable Data-Center Growth URL: https://www.aixenergy.io/the-state-playbook-for-equitable-data-center-growth/ Last updated: 2026-06-30T02:27:05.000Z States must ensure AI-driven data center growth strengthens—not strains—the grid. Fair pricing, consumer protection, and community reciprocity can align private innovation with public equity, turning energy demand into shared resilience. _This post is for paying subscribers only._ ### Managing Data Center Uncertainty Part I — The Uncertainty Problem URL: https://www.aixenergy.io/part-i-the-uncertainty-problem/ Last updated: 2026-07-29T19:42:19.000Z Artificial intelligence has shattered the linear logic of energy demand. By 2028, U.S. data-center electricity use could soar to 580 terawatt-hours—or stall at 325—a swing larger than Florida’s annual consumption. This volatility is not a failure of forecasting; it is a failure of governance. Institutions built for predictability are misfiring in an era defined by structural uncertainty. Developers obscure real load data, utilities are rewarded for overbuilding, and regulators—fearing shortage more than surplus—approve speculative capacity that locks in stranded assets, higher rates, and new emissions. AI data centers are not merely larger versions of what came before—they are a new species of infrastructure: power-dense, computationally volatile, and capital-intensive. Yet our policy frameworks still assume they behave like steady industrial loads. The result is a dangerous feedback loop: incomplete data drives uncertain forecasts; uncertainty fuels fear; fear drives overbuild; and overbuild deepens the uncertainty it was meant to solve. Drawing on new analyses from Lawrence Berkeley National Laboratory and Duke University, this first installment of the *AIxEnergy Framework Series* argues that the solution is not better prediction but smarter governance. The path forward begins with mandatory data disclosure, performance-based regulation that rewards accuracy over expansion, and flexible pricing that makes adaptability profitable. The question is no longer how to forecast demand—but how to build systems resilient enough to thrive when we can’t. 0:00 /1:49 1× ### **The New Energy Frontier** By 2028, U.S. data-center electricity consumption could reach 580 terawatt-hours—or it could plateau near 325\. That 255 TWh swing—larger than Florida’s entire annual usage—isn’t a rounding error. It is a governance gap. Our institutions still assume that energy demand grows linearly, incrementally, and predictably. Artificial intelligence has shattered that assumption. The AI data center is not an incremental technology. It is an entirely new species of infrastructure—power-dense, computationally volatile, and capital-hungry. Each new facility operates at a power density ten times that of a traditional data center, with clusters of GPUs consuming megawatts per rack. AI load is not a steady hum of cloud activity; it is a pulsing, asymmetric wave of compute intensity that shifts with algorithms, model architectures, and data availability. This volatility is reshaping the electric grid faster than regulation can adapt. In the past 18 months, utilities from Georgia Power to Dominion have announced double-digit revisions to load forecasts driven by AI data center demand. Yet the spread of uncertainty among those forecasts remains massive. Some expect AI demand to add 7 percent to system load; others predict 20 percent. The range isn’t analytical noise—it’s policy paralysis in numerical form. We are living through the most rapid structural change in electricity demand since the postwar industrial boom, but without the tools to govern it. The result is a dangerous feedback loop: incomplete data produces uncertain forecasts; uncertainty fuels fear of shortage; fear drives overbuild; and overbuild deepens the very uncertainty it was meant to cure. Our energy institutions were designed for a world that changed slowly, where infrastructure could be planned in decades and demand modeled in trends. That world no longer exists. AI is forcing utilities, regulators, and investors to plan in real time for technologies whose load profiles shift in months. Yet we continue to apply the same static governance logic—rewarding confidence, not accuracy; expansion, not adaptability. The paradox is that our forecasts are getting better while the world they describe is getting less predictable. Data quality improves, but the system’s behavior becomes more chaotic. The result is an explosion of risk—financial, operational, and political—locked inside an illusion of certainty. ### **The Shape of Uncertainty** Lawrence Berkeley National Laboratory’s 2024 analysis projects that U.S. data-center electricity use could fall anywhere between 325 and 580 TWh by 2028—an 80 percent swing in expected demand.¹ That range, the largest uncertainty band in any modern U.S. load forecast, reveals a structural flaw in how the energy sector handles unknowns. Some of the uncertainty stems from missing data, some from unpredictable variables like chip efficiency and cooling, and some from the unknowable—future AI applications that might multiply or moderate electricity use. Forecasting assumes uncertainty is random noise around a trend line. But AI introduces uncertainty that is systemic and strategic. Developers withhold operational data, utilities retain structural incentives to overbuild, and innovation keeps resetting the baseline. This is not a modeling problem. It is a governance problem. Utilities, regulators, and policymakers are being asked to plan for an industry whose power demand could double or plateau—and to do so without access to verifiable operational data. The opacity is extraordinary. Even as utilities scramble to expand generation portfolios, they lack basic visibility into how much power is being used, when, and for what type of workload. Is the load continuous inference, which runs 24/7, or intermittent training, which can flex in time? Those distinctions matter profoundly for grid design, but right now they are guesses. The implications are enormous. The 255 TWh uncertainty band represents 40–50 gigawatts of potential generation capacity. Building to the high end and landing near the low could strand tens of billions of dollars in assets while locking in decades of emissions.² The risks extend beyond cost—overbuilds crowd out renewables, inflate rates, and erode public support for the energy transition. And yet the opposite risk—underbuilding—could stall innovation and economic growth. The policy problem is not choosing between the two; it is designing a governance system capable of adjusting between them. ### **A Taxonomy for Turbulence** Uncertainty is not a single phenomenon. It operates across multiple layers, each demanding a different kind of response. 1. **Aleatoric (natural)** — Weather and load variability, statistically manageable and well understood. 2. **Measurement** — Missing or unreliable operational data from data centers and interconnection queues. 3. **Parametric** — Unknowns about chip efficiency, rack density, and cooling innovation trajectories. 4. **Epistemic** — Ambiguity about the system itself: what counts as a “data center” in a globally distributed compute network that shifts workloads across continents? 5. **Knightian (deep)** — The unknowable future: Will AI saturate, or will every device, vehicle, and system become a micro data center? Measurement and parametric uncertainty—short-term and reducible—can be addressed through mandatory disclosure, standardized reporting, and performance-based regulation (PBR) that rewards accuracy within a 1–3 year horizon. Epistemic and Knightian uncertainty—long-term and irreducible—require flexibility and optionality, not misplaced precision. Governance must distinguish between what can be measured and what must be managed. Treating all uncertainty as forecast error is like treating turbulence as bad piloting—it blames the wrong mechanism. We cannot smooth the flight path of AI’s energy curve, but we can build a better aircraft. ### **When Forecasting Becomes Fiction** Utilities are planning and permitting gigawatts of capacity based on speculative signals from developers. Industry insiders acknowledge that data-center developers file five to ten times more interconnection requests than projects they actually intend to build.³ These “phantom data centers” distort demand projections and drive unnecessary capacity commitments. Rate-of-return regulation turns this uncertainty into profit. Utilities are rewarded for capital investment, not for accuracy. When in doubt, they build. Regulators approve capacity for the high-end scenario because no one wants to be blamed for shortage. The result: systematic overbuild, socialized cost, and fossil lock-in.⁴ This is not malfeasance—it is inertia. Every actor behaves rationally within a flawed incentive structure. Developers inflate requests to secure optionality; utilities expand capital portfolios to earn regulated returns; regulators approve to avoid risk. The outcome is predictable: excess capacity, higher costs, and slower decarbonization. The only irrational behavior is the collective assumption that uncertainty can be ignored. In this ecosystem, uncertainty isn’t an error margin—it’s a business model. ### **A Governance Reframe** If uncertainty is the problem, the solution must begin with disclosure. We cannot manage what we cannot measure. Every interconnection approval should require transparent reporting of workload composition (training vs. inference), annual utilization rates, cooling type, and flexibility potential. These disclosures would allow regulators to distinguish between continuous and interruptible load—transforming forecasting from conjecture into constraint-based planning. Next, regulators must reward accuracy, not volume. Performance-Based Regulation (PBR) frameworks should compensate utilities for narrowing forecast error and penalize chronic overestimation. A utility that saves ratepayers billions by deferring unnecessary capacity should earn returns comparable to those gained from new capital investment. The incentive should shift from building to managing uncertainty. Third, we must price flexibility. Duke University’s Nicholas Institute found that a 0.5–1 percent reduction in peak data-center load could integrate nearly 100 GW of new demand with existing infrastructure.⁵ Geometric demand charges and narrow time-of-use (TOU) surcharges make such micro-curtailments financially rational. Once flexibility becomes an asset class, overbuild loses its logic. Finally, we must build optionality into infrastructure itself. Flexible interconnection contracts, modular substations, and reversible investments can make systems resilient to uncertainty. If AI demand slows, capacity can be repurposed. If it accelerates, it can be expanded without starting from scratch. Resilience is no longer about redundancy—it’s about reversibility. ### **The Clock Is Ticking** Global data-center electricity use is expected to more than double by 2030, driven primarily by AI computing loads.⁶ In the past year alone, Google, Amazon, and Microsoft announced 15 GW of new data-center capacity—projects that will test the limits of U.S. grid infrastructure.⁷ Meanwhile, local governments are beginning to push back. In Northern Virginia—the world’s largest data-center hub—officials are debating moratoriums on new interconnections until power and land constraints ease. In Oregon, environmental regulators have halted expansions over water-use concerns. In Georgia, new substations are facing community opposition over noise and visual impact.⁸ At the same time, utilities have requested $29 billion in rate increases for 2025, more than twice the average of the past decade.⁹ For regulators and ratepayers, the message is clear: uncertainty is expensive. What once looked like prudent planning now resembles speculative overreach. The political tolerance for overbuilding is shrinking. The next phase of the AI energy era will not be defined by shortage—it will be defined by accountability. ### **From Forecasting to Governance** We have entered the post-forecast era of energy planning. The question is no longer how to predict demand, but how to remain resilient when we can’t. That means treating transparency as infrastructure, rewarding prudence over projection, and designing markets that value adaptability as much as capacity. This transformation will not be easy. It requires regulators to trade control for clarity, utilities to exchange growth for accuracy, and developers to give up opacity for credibility. But the alternative—a grid built on guesswork—is unsustainable. Uncertainty is not the enemy. It is the environment. The grid’s real crisis isn’t forecasting—it’s the assumption that the future can still be known. ## **Roadmap for the Series** This first article defines the foundation: that the defining challenge of the AI-powered grid is not forecasting error but structural uncertainty. The rest of the *AIxEnergy Framework Series: Managing Data Center Energy Uncertainty* builds on this insight, translating diagnosis into design. **Part II — Phantom Data Centers: How Strategic Opacity Drives Overbuild** An investigation into how speculative interconnection practices and opaque disclosure allow developers to flood queues with fictitious projects—and how utilities’ capital incentives perpetuate the illusion of demand. We’ll examine policy tools that can separate signal from noise, including disclosure reform, probabilistic planning, and “load underwriting.” **Part III — The Utilization Paradox: Scarcity and Waste Inside AI Infrastructure** A technical deep dive into AI computing itself. We’ll explore how GPU underutilization (often just 60–70%) and I/O bottlenecks conceal massive temporal flexibility. We’ll show how distinguishing between training (flexible) and inference (rigid) workloads redefines grid operations—and turns waste into capacity. **Part IV — Flexibility Is the New Capacity: Unlocking 100 GW Without New Generation** This article will quantify how small, well-priced adjustments in load timing can unlock enormous system value. We’ll translate Duke University’s “curtailment-enabled headroom” findings into actionable pricing mechanisms—geometric demand charges, TOU surcharges, and pass-through incentives for data centers and their compute tenants. **Part V — From Uncertainty to Action: A Five-Mechanism Framework for AI-Era Governance** The series culminates in a full policy blueprint. Five mechanisms—granular disclosure, short-term PBR incentives, flexibility pricing, technology mandates tied to decarbonization, and speed-to-market rules for verified clean power—form a unified framework for adaptive governance. Together, these five articles will outline a new approach to managing data-center energy uncertainty—one that treats uncertainty not as a forecasting failure but as a design constraint for a faster, cleaner, and more accountable grid. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### **References** 1. Lawrence Berkeley National Laboratory, *Data Center Energy Use Forecast 2024* (Berkeley, CA: LBNL, 2024). 2. Michael Leifman, *Managing Data Center Energy Uncertainty: A Framework to Prevent Overbuild, Control Costs, and Unlock Grid Flexibility* (Washington, DC: AIxEnergy, 2025). 3. Brian Martucci, “A Fraction of Proposed Data Centers Will Get Built,” *Utility Dive*, May 15, 2025. 4. “The Murky Economics of the Data Center Investment Boom,” *The Economist*, September 30, 2025. 5. Duke University Nicholas Institute, *Curtailment-Enabled Headroom: Integrating Large Loads with Existing Capacity* (Durham, NC: Duke University, 2024). 6. International Energy Agency, “AI Is Set to Drive Surging Electricity Demand from Data Centres,” April 10, 2025. 7. Lisa Martens, “Tech Giants Announce 15 GW of New AI Data Center Capacity,” *Reuters*, April 22, 2025. 8. John Funk, “Local Governments Push Back Against AI Data Centers,” *Canary Media*, July 3, 2025. 9. U.S. Energy Information Administration, *Electric Power Monthly*, June 2025. *Next in the AIxEnergy Framework Series: Part II — Phantom Data Centers: How Strategic Opacity Drives Overbuild.* *Michael Leifman is a strategist with 20+ years advancing clean energy, technology, and policy. Former ERM Partner leading utility and state climate strategies; Founding Principal of Tenley Consulting; ex-GE innovation strategist; early DOE/EPA analyst. He holds M.S. degrees from Johns Hopkins and Carnegie Mellon, and a B.A. from the University of Chicago.* ### From Reactive Reliability to Dynamic Intelligence: How AI Is Building the Grid We Actually Need URL: https://www.aixenergy.io/from-reactive-reliability-to-dynamic-intelligence-how-ai-is-building-the-grid-we-actually-need/ Last updated: 2026-07-29T19:41:10.000Z Article Summary 0:00 /52.871837 1× AI is abuzz in the utility industry - and not only because of the surge in power demand from data centers and electrification across the U.S. and abroad. The deeper transformation is systemic: AI is moving the grid from reactive reliability to dynamic intelligence. The shift is no longer “nice-to-have”- it’s the only viable path to serve rising peaks, integrate variable and behind-the-meter resources, and shield customers from climate-driven volatility without overbuilding transmission, distribution, or peaking capacity. This next phase of grid modernization demands orchestration, beyond the current planned addition. The new control architecture- spanning Distributed Energy Resource Management Systems (DERMS), Virtual Power Plants (VPPs), AI-driven asset management, predictive outage restoration, and customer-centric load flexibility- is the connective tissue of a modern, responsive power system. But these technologies only reach full potential when stitched together by market access, data governance, and regulatory alignment under frameworks like FERC Order 2222. ## **The AI Imperative** U.S. peak demand is climbing, driven by data centers, electrification, and extreme weather. The U.S. Department of Energy’s *Pathways to Commercial Liftoff* calls for deploying roughly 80–160 GW of VPPs by 2030—- effectively a fast, modular alternative to some traditional capacity additions. The 2025 update underscores that peak demand could rise from \~800 GW in 2024 to \~900 GW by 2030, intensifying the need for flexible capacity. At the same time, FERC Order 2222 opens wholesale markets to aggregated DER- rooftop solar, batteries, EVs, and smart devices- turning previously passive load into dispatchable grid resources. This is a regulatory unlock for value stacking at scale and will yield market disrupters as well as opportunities for traditional business models like regulated utilities. Think of Advanced Distribution Management Systems (ADMS) as the utility’s real‑time air traffic control for the distribution system- switching, voltage/VAR optimization, outage restoration. DERMS specializes in orchestrating fast, granular, often customer‑owned resources at the grid edge. In practice, the two should interoperate tightly: ADMS ensures system safety and topology‑aware constraints; DERMS dispatches fleets of behind‑the‑meter assets to meet grid and market objectives. Utilities are already using advanced Outage Management Systems (OMS)/ADMS with predictive analytics for faster restoration and better crew guidance. Leading OEM offerings (such as that of GE Vernova) highlight fault location, damage assessment, and decision support capabilities that shorten SAIDI/SAIFI and improve CEMI, critical outcomes for the performance of the grid. ## **VPPs are Real—and Scaling Quickly** In California, the Demand Side Grid Support (DSGS) program shows what’s possible when customer assets are aggregated with market signals. During a July 2025 event, more than 100,000 home batteries delivered \~500–535 MW to the system - the equivalent of a mid‑size power plant mobilized from rooftops. Independent analysis suggests statewide VPPs could save Californians roughly $550 million per year by 2035 and avoid more than $750 million annually in traditional system costs. With rising costs of electricity across the U.S., these solutions can deliver meaningful savings to consumers and businesses. DOE’s Liftoff work quantifies national potential, while recent policy debates (e.g., California vetoes) underscore that regulatory alignment matters as much as technology. The opportunity is very real; the main barriers are effective policy design, appropriate cost allocation, and market effectiveness. ## **AI for Outage Prediction and Restoration** AI is no longer a lab demo for predictive analytics. Utilities are marrying dense weather data, LiDAR/imagery, and asset histories to predict where and when faults will occur- enabling pre‑positioning of crews, staging of materials, and prioritization of circuits before the first call comes in. Public examples include PG&E’s enhanced outage prediction models that combine thousands of weather stations with AI/ML, and utilities reporting >80% accuracy in predicting storm‑driven component failures in some use cases. Duke Energy has invested heavily in self‑healing and analytics that avoid or shorten millions of customer interruptions. Vegetation management - a leading cause of outages and wildfire risk- is also becoming data‑driven. Using data to be targeted in utilizing limited O&M dollars in the right places can make a huge difference in outage restoration times and reducing overall risk of the system. EPRI’s maturity framework outlines a progression from patrol‑based cycles to remote sensing, risk‑based targeting, and AI‑assisted prioritization. ## **Customer Load Management = Flexible Capacity** DOE’s Grid‑interactive Efficient Buildings (GEB) initiative formalizes the value: buildings can be flexible grid assets when equipped with controls, automation, and time‑varying rates. The roadmap aims to triple energy efficiency and demand flexibility in buildings by 2030 versus 2020 levels - driving both affordability and reliability. That flexibility is the fuel for AI‑driven VPPs. Brattle’s analyses of California’s programs show substantial net savings and reliability benefits when customer devices (batteries, thermostats, EVs) respond to peak events. The lesson for other states: align incentives, simplify enrollment, ensure devices can be orchestrated across programs, and pay for performance. ## **A Roadmap to the Intelligent Utility** 1. Establish a baseline system‑wide data strategy. Build a unified data fabric across AMI, SCADA, OT, GIS, weather, imagery, and markets. Govern metadata, quality, lineage, and cyber posture from day one. Prioritize features that materially improve reliability and economics (e.g., topology‑aware forecasts, locational hosting capacity, feeder‑level net‑load prediction). NREL’s DERMS/ADMS research points to observability as the first constraint to remove. 2. Integrate ADMS and DERMS with clear control authority. Define who does what, when—with ADMS as system‑safety authority and DERMS as the optimizer of DER fleets within network limits. Use standard interfaces and joint testbeds; stage rollout feeder‑by‑feeder with digital twins to validate dispatch under contingencies. OEM platforms and NREL guidance can provide reference architectures. 3. Operationalize AI for resilience - not dashboards. Move beyond pilots by embedding outage prediction into crew scheduling, spares logistics, and vegetation work orders with service‑level agreements. Monitor model drift and false‑positive/negative costs, and ensure interpretability and audit trails for safety‑critical calls. Utilities and industry groups are publishing pragmatic playbooks for doing this safely. 4. Stand up market participation under FERC 2222\. Build or partner for aggregation capabilities, telemetry, baselines, and settlement. Start with one ISO/RTO and two or three homogenous device types to learn fast. FERC’s rule provides the legal runway; the task is aligning interconnection, metering, and distribution‑level review so VPP dispatch doesn’t create local violations. 5. Scale customer programs that pay for flexibility. Pair time‑of‑use/critical‑peak pricing with automated device control. Offer simple “good‑better‑best” enrollment (e.g., thermostat only; thermostat + EV; thermostat + EV + battery). Use pay‑for‑performance to ensure resources deliver when called. California’s results indicate what’s possible even as program design evolves. 6. Measure what matters. Keep leading and lagging metrics few and relevant. Track avoided peak MW; event performance (availability, response time, persistence); system reliability (SAIDI/SAIFI/CEMI); and customer NPS by program. Tie internal incentives to both grid and customer outcomes so flexibility doesn’t become one‑way extraction. 7. Talent, change, and partners. Upskill grid operators and planners on probabilistic thinking and AI‑assisted workflows; recruit DER/VPP product managers who think like both system operators and marketers. Where appropriate, partner with OEMs and aggregators to accelerate time‑to‑value rather than building everything in‑house. ## **What Good Looks Like in 24 Months** - VPPs as first responders: 2–3% of system peak under contract via residential/commercial portfolios with verified delivery in extreme heat or cold. - AI‑assisted restoration: Pre‑storm staging guided by model outputs; measurable reductions in outage minutes and faster Estimated Restoration Time (ERT) accuracy. - Tight ADMS↔DERMS loop: Location‑aware dispatch that respects feeder constraints, coupled with automated voltage support from inverters during events. - Customer value unlocked: Clear bill savings and incentive pathways; simple enrollment; transparent privacy and cybersecurity posture to build trust. ## **The Bottom Line** Huge capital investments are planned to address the access, reliability, and resilience needs of the future grid. To be clear, AI, DERMS, and VPPs don’t replace wires or new generation. They buy time and optionality while we build the right new systems. The intelligent utility of the future is orchestrated, data-driven, and customer-powered. If we integrate the control stack, operationalize AI for resilience, and pay customers for flexibility—under the market access FERC 2222 enables - we can meet peak reliably and affordably, even as demand grows. The industry doesn’t need more pilots- it needs precision at scale. The next decade will reward those who operationalize intelligence, not just instrument it. The grid we build now will determine whether reliability becomes a constraint—-or our greatest competitive advantage. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. *AIxEnergy article author Hannah Kaplan is a senior energy executive and AI strategist with over 17 years of experience at General Electric, Honeywell, and Entergy. She builds organizations at the intersection of technology, markets, and human connection. Having lived in seven countries and visited over 120, she brings a global perspective to the future of energy and intelligence.* ### From Control to Cognition: Training the Grid for the Age of AI URL: https://www.aixenergy.io/from-control-to-cognition-training-the-grid-for-the-age-of-ai/ Last updated: 2026-07-29T19:43:00.000Z Article Summary 0:00 /60.003265 1× For more than a century, electric power systems have operated under a single organizing principle: control. Utilities designed the grid to obey human instruction, not to think. Engineers defined operational parameters—voltage, frequency, load margins—and the machinery complied. The logic was simple, deterministic, and profoundly effective. Electricity flowed because every part of the system, from the turbine to the transformer, knew its place in a chain of command. But the world that built that grid no longer exists. What was once a predictable architecture of load and supply has become a living system of volatility. Solar and wind power introduce intermittent rhythms, electric vehicles generate stochastic demand, and data centers—driven by the exponential hunger of artificial intelligence—have begun to redraw the geography of electricity itself. In Virginia, for example, hyperscale campuses now consume more power than entire cities, prompting Dominion Energy to warn of capacity shortfalls within the decade.¹ Across the country, utilities face similar dilemmas: how to serve loads that move faster than their planning cycles can adapt. The traditional model—human command, machine compliance—is reaching its cognitive limit. The grid must now evolve from something we manage to something that learns. ### The Limits of Control Command-and-control logic worked when the grid was linear. Engineers could model the system as a series of known relationships: if X happens, apply Y. Frequency deviates? Adjust governor setpoints. Load spikes? Dispatch spinning reserve. But these rules presume stability—that weather is seasonal, demand is cyclical, and failure is rare. None of those assumptions hold true anymore. Static line ratings, for instance, still dominate transmission planning even though they assume “average” weather conditions that scarcely exist. FERC Order 881, issued in 2021, recognized this mismatch and required utilities to adopt Ambient Adjusted Ratings—capacity values that flex with real-time temperature and wind conditions.² Yet implementation lags. Decision latency remains high; switching sequences depend on operator judgment and the finite bandwidth of human attention. Control rooms are flooded with alarms—tens of thousands per day in large balancing authorities—creating a signal-to-noise problem that erodes situational awareness.³ Meanwhile, distributed energy resources (DERs) are proliferating faster than operators can model them. Rooftop solar, behind-the-meter batteries, and electric vehicle fleets are no longer passive loads—they are dynamic, interactive nodes at the edge of the grid. Each introduces new behaviors that break old assumptions. Traditional forecasting tools, designed for aggregate loads, struggle to capture these nonlinearities. A single AI data center can draw 300 megawatts continuously, its load shape governed not by human demand but by machine inference cycles.⁴ The result is structural stress: longer interconnection queues, rising congestion costs, and a creeping erosion of reliability margins. We are entering a period where the grid’s complexity exceeds our ability to script it. ### Toward a Learning System A self-learning grid is not the same as an autonomous one. It is not about handing over control to algorithms; it is about constructing adaptive intelligence around human judgment. The shift resembles what aviation experienced when autopilot systems began to augment, rather than replace, pilots. In power systems, this means embedding continuous learning loops into the fabric of operations. The data already exists. Supervisory Control and Data Acquisition (SCADA) systems capture operating states in millisecond intervals. Phasor Measurement Units (PMUs) record oscillation signatures that reveal stability margins. Advanced Metering Infrastructure (AMI) traces household demand and EV charging behavior. When cross-referenced with weather and geospatial data, these signals form a sensory network capable of perception.⁵ The next step is cognition: algorithms that can generalize from patterns, anticipate rather than react, and tune thresholds dynamically. Imagine a transmission corridor that “learns” its own safe operating limits. When ambient temperature drops or wind increases conductor cooling, the system automatically raises its line rating within verified bounds. Or consider a substation that recognizes early indicators of transformer insulation degradation—detected as subtle harmonics in voltage waveforms—and schedules maintenance before a fault occurs. These are not speculative futures; they are pilots already under way at utilities from National Grid in the United Kingdom to Oncor in Texas.⁶ ### The Architecture of Learning To build a cognitive grid, we must combine disciplines that have long operated in silos: power engineering, data science, and systems psychology. Machine learning provides the statistical scaffolding, but human expertise defines the boundaries. Engineers still set protection zones, safety margins, and acceptable risk levels. AI’s role is to explore the vast parameter space in between—millions of combinations too complex for deterministic modeling. Dynamic Line Ratings are the first manifestation of this philosophy: capacity becomes a moving variable, tuned to real-world conditions rather than fixed assumptions. Predictive maintenance follows naturally, as sensors feed degradation models that evolve with each operational cycle. Learning dispatch agents—trained via reinforcement learning—can balance voltage and power factor faster than rule-based optimization, discovering new control strategies through simulation. And adaptive protection systems, already in development by researchers at Pacific Northwest National Laboratory, adjust relay settings in response to network reconfiguration while maintaining NERC compliance.⁷ The pattern is clear. Every operational domain that once depended on static logic can be recast as a learning problem. ### Human-Centered Governance The transition from control to cognition is not a surrender to automation—it is an invitation to redesign the partnership between humans and machines. In a cognitive grid, operators do not vanish; they ascend. Instead of manually approving every action, they supervise learning outcomes. Protection engineers validate adaptive settings rather than recalculating them. System planners become curators of models that evolve continuously instead of authors of studies frozen in time. This human-in-the-loop architecture is essential for both ethical and practical reasons. AI can optimize variables, but it cannot articulate values. Deciding how much reliability is “enough,” or what level of risk is acceptable for a community, remains a human prerogative. As philosopher Hubert Dreyfus warned decades ago, intelligence without embodiment lacks context.⁸ The grid, as a socio-technical system, must retain its human conscience even as it acquires synthetic cognition. The governance challenge, then, is to codify transparency, explainability, and accountability. Every adaptive decision must be traceable—why it was made, under what data conditions, and with what confidence. The energy sector can learn from aviation’s “flight recorder” principle: every algorithmic action leaves an audit trail. Without such frameworks, learning systems risk drifting into ungoverned autonomy, where optimization supersedes judgment. ### Why Timing Matters The urgency of this transformation cannot be overstated. The AI economy is rewriting electricity demand faster than any prior technological wave. Between 2020 and 2025, U.S. household electricity prices rose roughly twenty-four percent,⁹ partly reflecting infrastructure strain from data-center expansion. Developers now approach utilities not with price queries but with megawatt deadlines: *Can we get 400 MW by 2026?* The answer increasingly depends on whether the grid itself can learn. Building new transmission is necessary but insufficient. Permitting timelines stretch across a decade; interregional lines remain politically fraught. Intelligence, however, can be deployed faster than steel. By applying learning models to existing assets—raising throughput through topology optimization, dynamic ratings, and predictive coordination—utilities can unlock latent capacity equivalent to years of capital buildout. DOE’s 2023 Grid Deployment Office estimated that advanced grid management could defer $50 billion in infrastructure costs by 2035.¹⁰ The stakes extend beyond economics. As climate volatility intensifies, resilience will hinge on adaptability. A self-learning grid can self-correct under stress, rerouting power during wildfires or hurricanes in ways no static script could anticipate. In this sense, cognition becomes a resilience strategy—a form of collective intelligence distributed across conductors, substations, and sensors. ### The Cognitive Frontier If today’s power grid resembles the nervous system of industrial civilization, tomorrow’s will resemble its brain. Synapses form where data streams converge: PMUs, inverters, and weather nodes exchanging information at sub-second speeds. Feedback loops strengthen or weaken depending on performance, much like neural plasticity. Over time, the system develops memory—an archive of operational experience that informs future decisions. The analogy is more than poetic. Researchers at the Electric Power Research Institute are experimenting with reinforcement learning agents that “remember” past grid states to improve response during contingencies.¹¹ European operators are testing digital twins that simulate millions of potential configurations, learning optimal dispatch sequences before deploying them in the field. Each iteration moves the grid closer to cognition. Yet this evolution also raises philosophical questions. What does it mean for infrastructure to learn? Can a system designed for obedience acquire agency without destabilizing trust? The answer lies in design intent. A learning grid is not an independent intelligence; it is a collaborative one. Its purpose is not to decide but to assist—to extend human foresight across scales and timeframes we cannot otherwise perceive. ### Policy as Architecture Policy must evolve alongside technology. FERC 881 marked a turning point by embedding learning logic—dynamic adaptation—into regulation. Future rulemakings could go further, mandating data transparency, standardized model validation, and AI governance frameworks akin to medical device certification. At the state level, regulators can incentivize learning-based reliability metrics rather than purely capital expenditure. Internationally, the European Union’s AI Act offers a preview of sector-specific governance: classifying grid control systems as “high-risk” AI applications subject to audit.¹² The United States lacks such coherence, but momentum is building. DOE’s recent “AI for Energy” initiative emphasizes explainability and safety in critical infrastructure applications, signaling an awareness that intelligence must be accompanied by accountability.¹³ The transition will also reshape workforce development. Future operators will need fluency not only in power flow equations but in data ethics, probabilistic modeling, and human-machine interface design. Training the grid, in this sense, means training ourselves—to think symbiotically with systems that think back. Every technological shift carries an ethical dimension. The command-and-control era reflected the industrial mindset of its time: mastery over nature through calculation and authority. The cognitive era, by contrast, invites humility. It asks engineers to recognize that complexity can no longer be contained by control alone. Learning, not domination, becomes the organizing principle. This is not a loss of rigor; it is an evolution of it. To build a grid that learns is to acknowledge that intelligence—human or artificial—is always bounded, always provisional. Resilience arises not from perfection but from adaptation. In this sense, the future of the grid parallels the future of humanity in the age of AI: survival through cognition. ### Conclusion The grid we inherited was built to obey. The grid we need must be able to learn. If the twentieth century’s triumph was engineering stability into an unruly force of nature, the twenty-first must engineer adaptability into an unstable world. The command-and-control model brought us electrification; the learn-and-adapt model will bring us decarbonization, digitalization, and resilience. We cannot program the future by adding more rules to an increasingly dynamic world. We must train it—patiently, transparently, and with moral intent. The future grid will not be managed. It will be taught. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### Notes 1. Dominion Energy, *Integrated Resource Plan 2023*, Virginia State Corporation Commission filing, May 2023. 2. Federal Energy Regulatory Commission, *Order No. 881: Managing Transmission Line Ratings*, December 2021. 3. North American Electric Reliability Corporation (NERC), *Situational Awareness Technical Reference Document*, 2022. 4. U.S. Department of Energy, *Data Center Energy Consumption Trends*, 2024. 5. Electric Power Research Institute (EPRI), *Grid Learning Systems: Foundations and Pathways*, 2023. 6. National Grid UK, *Dynamic Line Rating Pilot Report*, 2022; Oncor Electric Delivery, *AI-Enhanced Asset Management*, 2023. 7. Pacific Northwest National Laboratory (PNNL), *Adaptive Protection and Learning Relays*, 2024. 8. Hubert L. Dreyfus, *What Computers Can’t Do: The Limits of Artificial Intelligence* (New York: Harper & Row, 1972). 9. U.S. Energy Information Administration, *Electric Power Monthly*, August 2025. 10. U.S. Department of Energy, *Grid Modernization Benefits Report*, 2023. 11. EPRI, *Reinforcement Learning for Grid Optimization*, 2024. 12. European Commission, *Artificial Intelligence Act*, 2024. 13. U.S. Department of Energy, *AI for Energy: Responsible Intelligence Framework*, 2025. ### The Supply-Chain Alarm: AI Data Centers Are Ready—But the Gear Isn’t URL: https://www.aixenergy.io/the-supply-chain-alarm-ai-data-centers-are-ready-but-the-gear-isnt/ Last updated: 2026-06-30T02:26:58.000Z AI-driven data center growth is colliding with a global shortage of transformers and grid equipment, causing years-long delays and soaring costs. The digital boom now depends on the slowest link in the chain: the industrial capacity to deliver power hardware. _This post is for paying subscribers only._ ### When Data Centers Learn to Bend: Why Flexibility Is the Next Frontier in Decarbonizing Hyperscale Infrastructure URL: https://www.aixenergy.io/when-data-centers-learn-to-bend-why-flexibility-is-the-next-frontier-in-decarbonizing-hyperscale-infrastructure/ Last updated: 2026-06-30T02:26:49.000Z Real hyperscaler decarbonization depends on data center flexibility—shifting, routing, and balancing workloads in real time to match cleaner grid power and reduce reliance on fossil generation. _This post is for paying subscribers only._ ### Leapfrogged by the Future: AI and the Fragile Moats of Energy Innovation URL: https://www.aixenergy.io/leapfrogged-by-the-future-ai-and-the-fragile-moats-of-energy-innovation/ Last updated: 2026-06-30T02:26:37.000Z AI is reshaping energy: grids once built in decades now leapfrog with models, agents, and twins. Speed brings opportunity but erodes moats—durability now lies in adaptability, governance, and momentum. The grid is becoming a cognitive system, raising urgent questions of ownership and control. _This post is for paying subscribers only._ ### The Age of Compute: How AI and Data Centers Rewired Global Energy Investment URL: https://www.aixenergy.io/worldinvestment/ Last updated: 2026-07-14T02:10:32.000Z Article Summary 0:00 /62.72 1× When the International Energy Agency released its [*World Energy Investment 2025*](https://iea.blob.core.windows.net/assets/1c136349-1c31-4201-9ed7-1a7d532e4306/WorldEnergyInvestment2025.pdf?ref=aixenergy.io) report, one number captured global attention: 3.3 trillion dollars. That is the total sum the world will invest in energy this year, marking a modest rise despite economic headwinds and geopolitical uncertainty. Yet beyond the headline, the report reveals something more profound—a structural reordering of the global energy economy. Two-thirds of that investment, roughly 2.2 trillion dollars, now flows toward clean energy: renewables, nuclear, grids, storage, and end-use electrification. Fossil fuels, long the anchor of industrial power, have been relegated to the minority share. 0:00 /1:36 1× Within this transition, a new kind of demand is reshaping the investment map—one not born of industrial expansion or household growth, but of computation. Artificial intelligence, and the data centers that sustain it, have emerged as decisive forces in global power dynamics. In the IEA’s words, the *AI-led acceleration in data center investment* could have far-reaching consequences for the power sector. The world has entered what the report calls the beginning of the *Age of Electricity*, yet within that age, the rise of compute power marks a deeper shift: intelligence itself has become an energy-intensive commodity. ### The Great Reallocation: From Fuels to Flows The IEA’s investment data tell a story of historic inversion. For decades, fossil fuel development exceeded investment in electricity infrastructure. Now, electricity is the gravitational center. In 2025, power-sector investment will reach nearly 1.5 trillion dollars—about fifty percent higher than total upstream spending on oil, gas, and coal combined. The reason is twofold: decarbonization and digitization. While clean-energy policies and technology costs continue to drive capital toward renewables, it is the rapid rise in electricity demand—largely from electrified transport, industrial heating, and data processing—that sustains the surge. The IEA notes that data centers, artificial intelligence, and cryptocurrency mining together could consume as much as 1,000 terawatt-hours of electricity globally by 2030, roughly equivalent to Japan’s total demand. In the United States alone, data centers may account for 10 percent of national electricity consumption by 2030, up from just 4 percent today. This represents one of the steepest single-sector demand increases in modern history—comparable to the rise of air conditioning or aluminum smelting in the twentieth century. In the past, energy systems evolved to serve physical industry: factories, smelters, refineries. Today, they are evolving to serve cognitive industry—clusters of chips performing trillion-parameter calculations. Each incremental leap in AI capability, whether training a language model or simulating climate systems, brings with it a corresponding pulse in electricity load. The IEA estimates that meeting this demand could require USD 170 to 340 billion in additional generation, grid, and storage investment in the United States by 2030\. It is, in effect, the rebuilding of an entire power system—not to heat or move matter, but to move information. ### The New Geography of Load This transformation is geographical as much as technological. The IEA’s regional deep dive highlights how U.S. technology firms are leading a global surge in energy procurement. Since 2015, they have contracted roughly 86 gigawatts of renewable capacity through corporate power purchase agreements (PPAs)—a number that rivals the installed wind capacity of Germany and Spain combined. These contracts are not symbolic gestures toward sustainability; they are risk hedges against volatility in both energy price and public scrutiny. Data centers require constant, high-capacity, low-latency power—and that means long-term, firm supply. But the quest for clean electricity is forcing companies to look beyond conventional renewables. The IEA cites data center operators as key early adopters of small modular reactors (SMRs) and advanced geothermal technologies. As of late 2024, about 26 gigawatts of nuclear capacity—mostly SMR—and 265 megawatts of geothermal capacity had been contracted through agreements with technology companies. These numbers may seem small in global context, but their significance lies in the signal they send to investors and developers. The digital sector is now functioning as a laboratory for next-generation power technology. Consider the pattern: hyperscale data centers are no longer passive consumers of electricity; they are becoming *co-developers* of generation assets. In markets like the United States, where capital and venture ecosystems are deep, AI-driven power demand is catalyzing private investment into early-stage energy technologies. Companies such as TerraPower, X-Energy, and Fervo Energy are cited as direct beneficiaries of this feedback loop. As AI capacity expands, it creates a financial incentive to diversify the clean-energy mix beyond solar and wind—to technologies that can provide around-the-clock, dispatchable power. ### The Bottleneck of Wires Yet this surge in private investment meets a sobering constraint: the grid. The IEA describes a widening gap between the pace of digital infrastructure buildout and the speed of grid expansion. Constructing a large-scale data center typically takes three to six years; building new grid infrastructure can take five to fifteen years. The mismatch is not merely a logistical inconvenience—it is a structural barrier. As of 2024, more than 205 gigawatts of advanced-stage solar and wind projects in the United States were waiting in interconnection queues, many of them in the same regions where AI-driven demand is most concentrated. Power availability has become the top concern for over 90 percent of data center operators, according to the IEA. Nearly half now identify grid upgrades as the most critical mitigation strategy for ensuring reliability. The supply chain for electrical equipment, particularly transformers, has become a chokepoint: lead times stretch up to six years, and prices rose by 26 percent in 2024\. These bottlenecks ripple outward—delaying renewable integration, slowing interconnection, and increasing project costs across the power sector. The IEA estimates that accommodating the U.S. data center boom alone will require over USD 16 billion in new grid investment by 2030, not counting upstream generation. This includes not only physical expansion but modernization—digital controls, flexible transmission capacity, and high-voltage direct current (HVDC) corridors capable of handling volatile loads. In short, the cognitive age requires a cognitive grid. ### From Power Consumers to Power Shapers The rise of AI is not merely increasing electricity demand; it is changing its shape. Traditional load profiles follow predictable daily and seasonal patterns—daytime peaks, nighttime troughs, summer air conditioning spikes. AI and data centers operate differently. Their workloads are often flat, constant, and geographically concentrated. The IEA notes that such high load factors are reshaping how utilities plan capacity additions. Instead of peaking demand curves, grid operators are now facing a steady, non-cyclical demand floor that compresses margins for flexibility. This shift has profound implications for market design. In regions where wholesale electricity markets depend on demand volatility to signal investment (for example, through price spikes), the flattening of load could dampen price signals—necessitating new mechanisms to recover fixed costs. Conversely, in systems with abundant renewables, steady data center load can function as a stabilizing force, absorbing excess generation during off-peak hours and enabling higher penetration of variable energy resources. The IEA does not prescribe solutions but highlights this tension as a defining feature of the energy-AI nexus. It is a paradox: AI threatens to strain power systems with new demand, yet also provides a pathway to optimize them. The same machine-learning tools that drive electricity consumption are also being deployed to manage it—from predictive maintenance in transmission networks to dynamic scheduling of distributed resources. The feedback loop between intelligence and infrastructure is tightening. ### Venture Capital and the Digital Energy Frontier In a year when most venture investment in energy technologies declined, artificial intelligence stood out as the lone growth sector. The IEA documents that AI-related start-up funding rose by 50 percent in early-stage rounds and over 200 percent in growth-stage financing during 2024\. This divergence underscores AI’s unique position as both an enabler and a disruptor of the energy transition. While capital tightened across clean-tech segments such as hydrogen and carbon capture, investors flooded into AI start-ups promising optimization, forecasting, and automation tools for power and grid management. The convergence is not coincidental. As power systems grow more complex—intermittent generation, distributed assets, prosumer behavior—the value of intelligence grows exponentially. The IEA’s data suggest that AI is becoming integral to every stage of the energy investment chain: siting renewables, managing assets, predicting demand, and monetizing flexibility. In that sense, the expansion of data centers is both symptom and catalyst. They are the physical embodiment of digital capital formation—massive, energy-hungry infrastructures built to train the very systems that will eventually govern energy itself. ### Fossil Echoes: The Reliability Gap Paradoxically, the report acknowledges that the AI boom may spur a temporary resurgence in fossil-fired generation. Despite record clean-energy investment, dispatchable capacity remains a limiting factor. Gas turbine orders, which had been declining for years, saw an uptick in 2024\. The IEA attributes this to the *AI-led data center spending spree* and the resulting need for reliable, fast-ramping backup power in advanced economies, particularly the United States. This does not signal a reversal of decarbonization trends, but rather the exposure of a reliability gap. Until advanced storage, nuclear, and grid reinforcements scale, fossil generation will remain the bridge between ambition and uptime. In effect, data centers are forcing utilities to reconcile two imperatives: zero-carbon commitments and zero-downtime operations. The former defines long-term policy; the latter defines short-term feasibility. ### The Global Asymmetry of Investment While AI and data centers dominate attention in advanced economies, the IEA cautions that much of the world still lacks the financial capacity to join the digital-energy race. Nearly one-third of all clean-energy investment now occurs in China, which continues to balance massive renewable deployment with steady coal expansion to ensure reliability. India, too, is accelerating investment in both solar and coal capacity, viewing the latter as flexibility insurance for an electrifying economy. In contrast, much of Africa remains on the margins. The continent attracts only two percent of global clean-energy investment despite housing twenty percent of the world’s population. The IEA emphasizes that without increased concessional finance, regions like Sub-Saharan Africa will miss the opportunity to build modern digital and energy infrastructure in tandem. This asymmetry could entrench a new kind of divide—not between fossil and clean economies, but between connected and disconnected ones. ### The Age of Electricity Meets the Age of Intelligence The IEA’s *World Energy Investment 2025* ultimately reads as both ledger and mirror—a balance sheet of capital flows and a reflection of deeper systemic change. The report shows that more than half of total energy investment now goes to electricity and electrified end uses. Yet electricity itself is no longer merely a utility; it is becoming the substrate of cognition. Every watt delivered to a data center powers not just servers, but learning systems—digital entities that, in turn, shape how energy is produced, distributed, and consumed. This recursive dynamic defines the new frontier of energy policy. Traditional distinctions between supply and demand are blurring. AI models require stable energy access, while grid operators require AI to manage that access efficiently. The two have become symbiotic—each dependent on the other’s expansion. The report’s linkage between its *Energy and AI* analysis and *World Energy Investment* series signals that the IEA now treats artificial intelligence not as a niche technology, but as a macroeconomic driver of energy transformation. ### The Cognitive Grid: A New Design Principle The notion of a *cognitive grid* emerges naturally from the IEA’s data. As AI demand accelerates, the power system must evolve from static architecture to adaptive organism. The infrastructure of the twentieth century was built for predictable, analog loads; the infrastructure of the twenty-first must accommodate intelligent, variable, and self-optimizing ones. Achieving that requires new coordination between utilities, regulators, and digital platforms. For policymakers, this means rethinking permitting and interconnection timelines. If a single data center can require 500 megawatts of firm capacity—equivalent to a mid-sized power plant—then transmission planning must shift from reactive approval to proactive design. The IEA’s data imply that without systemic reform, grid delays could erode the economic gains of the clean-energy transition. For utilities, it means engaging data centers as partners rather than adversaries. Many hyperscalers are now experimenting with *grid-integrated operations*: adjusting compute loads to follow renewable availability, colocating generation and processing assets, and even funding substation upgrades. Such models turn AI facilities into flexible nodes rather than rigid endpoints. For investors, the lesson is clarity. The next wave of returns will favor those who can bridge the cognitive and physical layers of infrastructure—where electrons meet algorithms. The IEA’s investment tables hint at this convergence: solar, storage, grid digitalization, and AI all share the same underlying economics of learning curves and scale effects. ### Conclusion: Powering the Mind of the Machine *World Energy Investment 2025* is not a treatise on artificial intelligence. Yet in documenting where trillions of dollars now flow, it inadvertently captures the energy dimension of the digital revolution. AI and data centers are the newest, most capital-intensive form of industrial demand—one that operates at the intersection of electrons, silicon, and information. The world’s energy system, once organized around the extraction and combustion of fuels, is reorganizing around the management of flow: of electrons, data, and capital. Each reflects the same underlying law—that intelligence, whether human or synthetic, requires energy to exist. In that sense, the IEA’s latest report is not just about investment; it is about inheritance. The infrastructure we build today will power the minds of tomorrow. Behind every algorithm lies a transformer—not just the neural kind, but the steel-and-copper kind. And as the IEA makes clear, the challenge of our time is to build both fast enough, clean enough, and wisely enough to ensure that the next revolution in intelligence does not outrun the grid that sustains it. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. *Sources: International Energy Agency, World Energy Investment 2025 (Paris: IEA, 2025); International Energy Agency, Energy and AI (Paris: IEA, 2025); Renewables 2024 (IEA, 2024).* ### The Utility-AI Playbook: A Practitioner’s View from the Grid Edge URL: https://www.aixenergy.io/the-utility-ai-playbook-a-practitioners-view-from-the-grid-edge/ Last updated: 2026-07-29T19:43:05.000Z Article Summary 0:00 /62.223673 1× When SCADA first entered the substation, engineers met it with folded arms. Many of us who grew up around relays, analog meters, and switchyards learned to trust the physical world—the smell of ozone, the vibration of a transformer, the needle of an ammeter. Letting a remote signal tell us what was happening miles away required not just new technology, but a new mindset. We didn’t call it “digital transformation” back then—we just called it *change*, and it took pilots, patience, and a lot of coffee to make it stick. That was half a century ago. Today, the grid stands on the brink of another transformation—one defined not by electrons alone, but by intelligence. Artificial intelligence is no longer confined to tech labs or data centers; it’s entering substations, control rooms, and utility planning departments. And much like SCADA or AMI before it, AI is meeting the same old skepticism: “Can we really trust it?” The answer, as before, lies not in the technology itself but in *process*. Utilities have always advanced through structured, stepwise adoption—pilot, verify, scale. What separates a failed innovation from a lasting one is not the brilliance of the idea, but the rigor of its execution. If AI is to become truly *utility-grade*, it must be built on that same foundation: reliability, accountability, and governance. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2025/10/From-Insight-to-Implementation--1-.png) ### Step 1: Start with the Use Case, Not the Hype Every tool that lasts in a utility starts with a problem, not a pitch. SCADA didn’t begin as an experiment in automation—it was a visibility solution. AMI wasn’t about smart meters; it was about smarter billing and outage management. AI must follow that logic. The first step is defining *why* you need AI. What decision does it accelerate? What risk does it mitigate? What cost does it reduce? If those answers aren’t clear, the technology becomes a hammer searching for nails. The best AI deployments I’ve seen begin with humble, well-bounded questions: “Can we improve DER hosting analysis accuracy?” “Can we automate switching sequence validation?” “Can we predict asset failure earlier?” These questions may sound small—but so did the first SCADA experiments that only monitored a single substation. Start small, but design for scale. Every success story in utility modernization—from distribution automation to demand response—followed the same trajectory: prove it locally, then scale system-wide. ### Step 2: Structure the Data Before You Train the Model Utilities are rich in data but poor in structure. We have terabytes of load profiles, breaker operations, outage reports, and regulatory filings, but they live in silos—each with its own taxonomy, timestamping, and engineering shorthand. Before any AI model can be trained, this fragmented ecosystem needs order. The process begins with *data literacy*: understanding what data exists, who owns it, and how it’s used. The next step is harmonization. Create common schemas that define terminology across departments—whether it’s voltage levels, equipment IDs, or naming conventions. Once the dictionary is consistent, AI can begin to learn meaningfully. Think of structured data as the foundation of AI literacy. When we rolled out GIS integration twenty years ago, utilities spent years cleaning up asset registries before seeing value. The same principle applies here. A model trained on messy, inconsistent data will generate elegant nonsense. AI doesn’t need more data—it needs *better* data. Contextual, structured, and relevant. ### Step 3: Fine-Tune Responsibly You don’t need to train a billion-parameter model from scratch to unlock value. Tools like Low-Rank Adaptation (LoRA) now allow utilities to fine-tune large models with manageable datasets. This is where the real engineering discipline comes in. The fine-tuning process should unfold in three deliberate stages: 1. Domain Adaptation: Train the base model on the utility’s internal corpus—technical manuals, line diagrams, maintenance logs, and operating procedures. This ensures the model “speaks utility.” 2. Instruction Tuning: Introduce human reasoning through instruction–response pairs. For instance: - *Input:* “Evaluate the voltage impact of adding 500 kW of solar at Feeder 14.” - *Output:* “Voltage rise exceeds ANSI limits; recommend Volt-VAR inverter settings.” This step transforms expert logic into machine-readable intelligence. 3. Alignment and Validation: Incorporate safety and compliance guardrails. The model must know what *not* to do—just as a relay knows when *not* to close. Fine-tuning responsibly mirrors commissioning equipment: design, test, align. We don’t energize a transformer without inspection; we shouldn’t deploy an AI model without validation. ### Step 4: Roll Out in Phases Utilities thrive on phased implementation. Whether it’s a new substation automation system or a tariff redesign, the formula is the same: pilot, evaluate, scale. - Phase 1 – Pilot (3–4 months): Select a focused use case. Build a dataset of 1,000–2,000 examples. Test outputs with senior engineers in the loop. The goal is not perfection but trust. - Phase 2 – Expansion (6–8 months): Add adjacent use cases—load forecasting, reliability modeling, voltage optimization. Integrate the model with existing tools (CYME, Synergi, OMS, CIS). Capture user feedback for retraining. - Phase 3 – Enterprise Integration (12+ months): Deploy AI models across departments. Introduce continuous learning pipelines that refine performance over time. Create department-specific model variants tuned to planning, operations, or customer service. Each phase builds credibility. The process should be transparent, auditable, and paced for institutional absorption. Remember, trust in utilities isn’t built overnight—it’s earned, one accurate prediction at a time. ### Step 5: Measure What Matters Utilities don’t measure success by novelty—they measure it by performance. For AI, that means rigorous evaluation frameworks. - Technical Accuracy: Are voltage calculations correct? Are outputs consistent with IEEE and NERC standards? - Process Efficiency: How much time does AI save in report generation or analysis cycles? - Reliability Metrics: Are outage forecasts improving? Are maintenance interventions more targeted? Benchmarks should be contextualized. Comparing AI to human performance isn’t about replacement—it’s about augmentation. If engineers spend less time wrangling data and more time solving problems, that’s success. AI evaluation, like regulatory rate cases, should balance quantitative rigor with contextual judgment. A model may not always be right—but if it makes people more effective, it’s doing its job. ### Step 6: Govern Relentlessly In the power sector, governance is the difference between reliability and risk. We have checklists for everything—switching procedures, relay settings, safety tags. AI needs the same. Governance should operate on three levels: 1. Regulatory Compliance: Ensure models respect regional standards and reporting requirements. Train them on actual rulebooks—NERC CIP, IEEE standards, state regulations. 2. Data Security: Treat models as critical infrastructure assets. Store them on-premises where feasible. Protect inputs and outputs as you would SCADA data. 3. Engineering Oversight: Keep humans in the loop. No model should autonomously issue control actions without operator verification. Document every AI-assisted decision for traceability. AI without governance isn’t innovation—it’s exposure. A well-governed model doesn’t just make good recommendations; it makes defensible ones. ### Step 7: Think Long Term Utility innovation has always been evolutionary, not revolutionary. SCADA didn’t change the grid overnight—it matured through thousands of incremental upgrades. AI will be the same. The goal isn’t to build a flashy prototype; it’s to establish a durable capability. Long-term planning means designing systems for extensibility. Models must adapt to new data sources, standards, and use cases. Feedback loops should turn every correction into training material. Cross-utility collaboration—anonymized and aggregated—will accelerate industry-wide learning. Imagine models that improve as neighboring grids learn, creating a network of collective intelligence. This vision isn’t far off. It’s the natural progression of an industry that already understands feedback, control, and resilience. ### The Human Factor After three decades in T&D, I’ve learned that every great system, no matter how advanced, depends on the people who operate it. AI is no different. It will not replace human expertise—it will amplify it. The operator remains the ultimate authority. Every time a human adjusts a model’s recommendation based on experience, the system grows smarter. That is the partnership we are building: a feedback loop between cognition and experience, between machine inference and human intuition. The self-learning grid isn’t about autonomy; it’s about collaboration. Reliability has never been just an algorithm—it’s a culture. And in that culture, trust is earned slowly, but once earned, it endures. AI must live by those same rules. Let the grids learn for themselves—but never without us. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. #### Notes 1. IEEE Standard 1547-2018, *Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces* (IEEE, 2018). [https://standards.ieee.org/standard/1547-2018.html](https://standards.ieee.org/standard/1547-2018.html?utm%5Fsource=chatgpt.com) 2. Hu, Edward J., et al. “LoRA: Low-Rank Adaptation of Large Language Models.” *arXiv preprint* (2021). [https://arxiv.org/abs/2106.09685](https://arxiv.org/abs/2106.09685?utm%5Fsource=chatgpt.com) 3. Edison Foundation / Institute for Electric Innovation, *Electric Company Smart Meter Deployments—Foundation for a Smart Energy Grid* (Edison Foundation, 2013). [https://www.edisonfoundation.net/-/media/Files/IEI/publications/Final-Electric-Company-Smart-Meter-Deployments--Foundation-for-A-Smart-Energy-Grid.pdf](https://www.edisonfoundation.net/-/media/Files/IEI/publications/Final-Electric-Company-Smart-Meter-Deployments--Foundation-for-A-Smart-Energy-Grid.pdf?utm%5Fsource=chatgpt.com) 4. U.S. Department of Energy, *Advanced Metering Infrastructure and Customer Systems: Summary Report* (DOE, 2016). [https://www.energy.gov/sites/prod/files/2016/12/f34/AMI%20Summary%20Report\_09-26-16.pdf](https://www.energy.gov/sites/prod/files/2016/12/f34/AMI%20Summary%20Report%5F09-26-16.pdf?utm%5Fsource=chatgpt.com) 5. U.S. Department of Energy, “Grid Modernization Initiative,” Energy.gov (DOE). [https://www.energy.gov/gmi/grid-modernization-initiative](https://www.energy.gov/gmi/grid-modernization-initiative?utm%5Fsource=chatgpt.com) 6. Federal Energy Regulatory Commission, *Commissioner-Led Reliability Technical Conference* (FERC, 2024). [https://www.ferc.gov/media/2024-commissioner-led-reliability-technical-conference](https://www.ferc.gov/media/2024-commissioner-led-reliability-technical-conference?utm%5Fsource=chatgpt.com) 7. North American Electric Reliability Corporation (NERC), *Reliability Standards – Standards & Compliance* (NERC). [https://www.nerc.com/pa/Stand/Pages/Default.aspx](https://www.nerc.com/pa/Stand/Pages/Default.aspx?utm%5Fsource=chatgpt.com) 8. National Association of Regulatory Utility Commissioners (NARUC), *AI & Utility Regulation Resources* (NARUC). [https://pubs.naruc.org/pub/165824D5-A04C-0A98-6CC4-2F91B57ADA7E](https://pubs.naruc.org/pub/165824D5-A04C-0A98-6CC4-2F91B57ADA7E?utm%5Fsource=chatgpt.com) 9. Electric Power Research Institute (EPRI), *Monitoring and Advanced Data Analytics Program* (EPRI). (EPRI member resource; publicly viewable via EPRI program descriptions.) ### The Digital Dividend: How Data Centers Are Rewriting the IEA’s Renewables 2025 Outlook URL: https://www.aixenergy.io/renewables2025/ Last updated: 2026-07-14T02:10:39.000Z Article Summary 0:00 /76.982857 1× The International Energy Agency’s [*Renewables 2025*](https://iea.blob.core.windows.net/assets/48eccb83-984c-45d2-bf78-67a61e88d241/Renewables2025.pdf?ref=aixenergy.io) report arrives at a pivotal juncture—a moment when the clean energy revolution is both maturing and mutating. The numbers are staggering: by 2030, global renewable electricity capacity will surpass 4,600 gigawatts (GW)—more than twice the 2022 total—and renewables will generate nearly 43 percent of the world’s electricity. This is not a marginal transition; it is the defining infrastructure buildout of the century. At the core of this acceleration lies solar photovoltaics (PV), responsible for almost 80 percent of new renewable additions. In 2025 alone, the world is expected to install more than 750 GW of new renewable capacity—an unprecedented leap. Wind power, the stalwart of past decades, continues as the second pillar, though recent headwinds—supply chain friction, financing stress, and offshore cost inflation—have tempered short-term expectations. Offshore wind additions have been revised downward by roughly one-quarter from *Renewables 2024*, yet the long-term trajectory remains upward. China remains the gravitational center of renewable growth, projected to account for nearly 60 percent of new global capacity through 2030\. Europe advances steadily under the twin imperatives of energy security and decarbonization, while India embarks on a historic scaling of solar and wind, bolstered by domestic manufacturing and grid reform. In the United States, long-term policy momentum from the Inflation Reduction Act persists, though interconnection delays and local opposition have trimmed near-term deployment. By the decade’s close, solar and wind together will generate nearly 28 percent of the world’s electricity. Yet amid the familiar optimism, one understated clause in *Renewables 2025* signals a deeper inflection: the report attributes rising electricity demand in part to “services and commercial sectors linked to digitalization and cooling needs.” In that phrase lies a quiet revolution—the fusion of two great forces of the 21st century: clean energy and computation. 0:00 /1:36 1× ### Electricity Demand Growth Projections The IEA projects global electricity demand to grow about 3 percent annually through 2030, a modest but meaningful revision upward. In 2024, electricity demand surged 4.3 percent, the steepest increase in peacetime history, adding roughly 1,080 terawatt-hours (TWh) to the grid. Nearly 60 percent of that rise originated in buildings—driven not by lighting or appliances, but by air-conditioning and digital infrastructure. To understand the magnitude of these new forces, we translate narrative into arithmetic. **Step 1: Data Center Growth**: According to the IEA’s *Energy & AI* report, global data center electricity use rose from \~460 TWh in 2022 to an expected 1,000 TWh by 2026\. A smooth trajectory suggests an annual increment of roughly 135 TWh; conservatively, 100 TWh of new demand is attributed to 2024. **Step 2: Cooling Demand**: The IEA attributes roughly 600 TWh of 2024’s electricity growth to the buildings sector, emphasizing cooling as the main driver. If 30–60 percent of that increase stems from air-conditioning, that equates to \~180–360 TWh. **Step 3: Combined Impact**: Digitalization (data centers + AI) and cooling together contributed \~280–460 TWh, or 26–43 percent of 2024’s global electricity growth. **Step 4: Counterfactual Growth**: Subtracting this uplift, global demand growth would have been 3.1–3.6 percent, not 4.3 percent—an adjustment of 0.7–1.2 percentage points. | **Driver** | **2024 Increment (TWh)** | **Share of Global Growth (%)** | | ------------------------------------ | ------------------------ | ------------------------------ | | Data centers (AI, cloud) | \~100 | \~9% | | Cooling (air-conditioning) | \~180–360 | \~17–33% | | **Total (Digitalization + Cooling)** | **\~280–460** | **\~26–43%** | **Step 5: Projected Continuation** Assuming data center demand grows \~15 percent annually, and cooling normalizes, the IEA’s forecasted growth rates—3.3 percent (2025) and 3.7 percent (2026)—would drop to 2.3–2.8 percent and 2.7–3.2 percent, respectively. | **Year** | **IEA Headline Growth (%)** | **Ex-Digitalization & Cooling (%)** | **Demand Adjustment (TWh)** | | -------- | --------------------------- | ----------------------------------- | --------------------------- | | 2024 | 4.3 | 3.1–3.6 | 280–460 | | 2025 | 3.3 | 2.3–2.8 | 240–400 (proj.) | | 2026 | 3.7 | 2.7–3.2 | 300–450 (proj.) | These calculations reveal that nearly one percentage point of modern demand growth is now attributable to “AI + AC.” ### The Renewables Impact: Quantifying the Digital Uplift In *Energy & AI*, the IEA explicitly links 110 TWh of additional renewable generation (2024–2030) to data center electricity use—AI being the dominant contributor. Converting this to capacity requires assumptions about technology mix and capacity factors. | **Technology** | **Share (%)** | **Generation (TWh)** | **Capacity Factor** | **Estimated Capacity (GW)** | | -------------- | ------------- | -------------------- | ------------------- | --------------------------- | | Solar PV | 70 | 77 | 0.20 | 44 | | Onshore wind | 25 | 27.5 | 0.35 | 9 | | Offshore wind | 5 | 5.5 | 0.45 | 1–1.5 | | **Total** | **100** | **110** | — | **\~54 GW** | Thus, absent AI and data centers, *Renewables 2025* would forecast roughly 54 GW less renewable capacity by 2030—equivalent to the entire installed solar fleet of France and Spain combined. The IEA identifies the United States as the largest locus of digital load growth, followed by China and Europe. Allocating capacity accordingly: | **Region** | **Incremental Generation (TWh)** | **Estimated Capacity (GW)** | **Technology Mix** | | ------------- | -------------------------------- | --------------------------- | --------------------------------------- | | United States | 55 | 27 | Solar-heavy; moderate wind | | China | 33 | 16 | Solar + wind balance | | Europe | 22 | 11 | Solar, onshore wind, small offshore | | **Global** | **110** | **\~54** | **70% solar, 25% onshore, 5% offshore** | This distribution reflects where AI and data center buildout is most aggressive—and where renewables are most responsive. ### The Interdependence: Intelligence Meets Infrastructure The once-linear relationship between electricity and GDP has fractured. Today, data centers are no longer passive consumers but active agents shaping generation portfolios and grid operations. Early data centers drew criticism for energy intensity; hyperscale facilities now drive clean energy procurement. Through 24/7 carbon-free energy (CFE) commitments, companies like Google, Microsoft, and Amazon have become anchor customers for renewable projects, contracting more than 50 GW of clean power globally since 2020\. The sector has effectively created a new class of offtaker—one that buys time-aligned energy, not just annual offsets. AI training loads, while massive, are flexible. Unlike industrial processes, many can be shifted across hours or even days. This has given rise to “grid-interactive compute,” where data centers function as controllable demand assets. Emerging pilots in Virginia, Dublin, and Finland already curtail or time-shift compute to absorb renewable variability. The next frontier is co-location: pairing data centers with solar, wind, and battery storage to create integrated microgrids. In Texas, Nevada, and Inner Mongolia, gigawatt-scale campuses are being designed with dedicated generation and 4–8 hour batteries, capable of islanded operation during grid stress events. This architecture turns data centers into stabilizers, not stressors—nodes of flexibility embedded within a renewable grid. ### Strategic Implications **Policymakers** Governments should treat data centers as strategic infrastructure, integrating them into grid planning and incentivizing renewable co-development. Including digitalization in national demand models will prevent chronic underestimation of load growth. Aligning permitting and interconnection reforms with AI expansion is now a necessity. **Utilities and Grid Operators:** Data centers represent both risk and reward. Clustering can overload transmission corridors, yet their flexibility and credit strength make them ideal partners for storage integration, demand response, and voltage support. Regulators should enable tariff structures that reward real-time flexibility and 24/7 carbon matching. **Renewable Developers and Supply Chains:** The hyperscale boom is creating a durable floor under renewable demand. Corporate PPAs, once peripheral, now drive a third of U.S. solar procurement. Developers that integrate battery and compute co-siting will find new financing efficiencies and smoother interconnection pathways. **Data Center Owners and Developers:** The new competitive frontier is energy intelligence. The winners will not simply consume energy but choreograph it—balancing AI workloads with renewable variability, leveraging storage, and trading flexibility into grid markets. ### Conclusion: The Energy of Intelligence In *Renewables 2025*, the IEA chronicles a world electrifying at speed. Yet the story between the lines is even more profound: the machines driving the digital revolution are also accelerating the clean energy transition. Without data centers and AI, global electricity demand in 2024 would have risen 3.1–3.6 percent, not 4.3 percent. Without their pull, the world’s renewables forecast would shrink by roughly 110 TWh of generation—about 54 GW of clean capacity. But with them, the grid is evolving—not just expanding, but learning. The once-separate domains of energy and computation are converging into a single system of intelligence, where electrons and algorithms reinforce one another. Data centers are no longer endpoints of electricity—they are instruments of renewal. In that symbiosis lies the blueprint for the next energy age: not power *for* intelligence, but power *with* intelligence. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### References 1. International Energy Agency. *Renewables 2025: Analysis and Forecasts to 2030.* Paris: IEA, 2025\. [https://www.iea.org/reports/renewables-2025](https://www.iea.org/reports/renewables-2025?ref=aixenergy.io). 2. International Energy Agency. *Energy and AI.* Paris: IEA, 2025\. [https://www.iea.org/reports/energy-and-ai](https://www.iea.org/reports/energy-and-ai?ref=aixenergy.io). 3. International Energy Agency. *Global Energy Review 2025.* Paris: IEA, 2025\. [https://www.iea.org/reports/global-energy-review-2025](https://www.iea.org/reports/global-energy-review-2025?ref=aixenergy.io). 4. International Energy Agency. *Electricity 2025.* Paris: IEA, 2025\. [https://www.iea.org/reports/electricity-2025](https://www.iea.org/reports/electricity-2025?ref=aixenergy.io). 5. S&P Global 451 Research. *European Data Center Power Market Analysis.* London: S&P Global, 2025. 6. Brookfield Renewable Partners. “Microsoft and Brookfield Renewable Announce 10.5 GW Clean Energy Agreement.” Press release, May 1, 2024. 7. Reuters. “Forget AI: Keeping Cool Is a Bigger Power Sector Problem.” *Reuters*, June 23, 2025. ### The OpenAI–AMD Accord in Context: From Silicon Optionality to Grid Sovereignty URL: https://www.aixenergy.io/the-openai-amd-accord-in-context-from-silicon-optionality-to-grid-sovereignty/ Last updated: 2026-06-30T02:26:13.000Z OpenAI’s 6-GW deal with AMD marks AI’s growing dependence on real power. It’s a milestone linking silicon to substations, forcing planners and policymakers to treat compute demand as part of the energy system, not apart from it. _This post is for paying subscribers only._ ### From Fortress to Federation: Flexible Data Centers and the Future of Energy URL: https://www.aixenergy.io/from-fortress-to-federation-flexible-data-centers-and-the-future-of-energy/ Last updated: 2026-06-30T02:26:02.000Z Flexible data centers are reshaping the grid—shifting from rigid energy consumers to dynamic partners. By embedding flexibility into software, markets, and regulation, they promise cleaner, cheaper, and more resilient power while redefining the balance between digital and energy systems. _This post is for paying subscribers only._ ### Beyond the PPA: Flexibility as the Real Test of Hyperscaler Decarbonization URL: https://www.aixenergy.io/beyond-the-ppa-flexibility-as-the-real-test-of-hyperscaler-decarbonization/ Last updated: 2026-06-30T02:25:52.000Z Hyperscale data centers are reshaping the grid. Beyond renewable contracts, their true test is flexibility—using batteries, orchestration, and demand response to replace fossil redundancy. The future of AI depends on whether hyperscalers embed adaptability into reliability. _This post is for paying subscribers only._ ### America’s AI Action Plan: An Energy and Infrastructure Appraisal URL: https://www.aixenergy.io/americas-ai-action-plan-an-energy-and-infrastructure-appraisal/ Last updated: 2026-07-14T02:10:45.000Z In July 2025, the White House released its [*America’s AI Action Plan*](https://www.whitehouse.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf?ref=aixenergy.io), a blueprint intended to secure U.S. leadership in the artificial intelligence revolution. The document arrives at a moment of inflection: hyperscale data centers are consuming power on a scale once reserved for metropolitan grids, interconnection queues are bloated with gigawatts of speculative projects, and the public conversation oscillates between AI as savior and AI as destabilizer. Against this backdrop, the plan’s three pillars—Accelerate AI Innovation, Build American AI Infrastructure, and Lead in International AI Diplomacy and Security—invite both admiration and critique. From the perspective of AIxEnergy, a platform devoted to mapping the convergence of computation and electricity, the plan represents a monumental recognition of reality. Yet it also reveals fault lines that could determine whether the United States enters the AI age as master builder or captive consumer. ### The First Strength: Unflinching Recognition of AI’s Energy Appetite Rarely does a policy document state so bluntly that artificial intelligence is “the first digital service that compels a step-change in U.S. energy generation.” This is more than rhetoric—it is an admission that bits and electrons have fused into one economy. Where the early internet could piggyback on existing infrastructure, AI compels new wires, new substations, and new generation. The plan wisely calls for streamlined permitting under NEPA, FAST-41 coverage for data centers and power plants, and categorical exclusions for low-impact expansions. For developers stalled in queues or waiting years for environmental reviews, this signals the possibility of accelerated buildouts. For utilities, it provides political cover to retain dependable assets that might otherwise retire prematurely. But recognition is not the same as resolution. While the plan directs agencies to “stabilize, optimize, grow,” the sequence betrays a bias: keep the old fleet online, squeeze existing corridors, then—only then—pursue frontier technologies. For an administration otherwise keen to tout decarbonization, this prioritization of stability over transformation risks baking in fossil incumbency. ### Strength in Workforce Strategy Equally notable is the plan’s treatment of workforce policy as infrastructure. Treasury guidance to expand tax-advantaged AI upskilling, creation of an AI Workforce Research Hub, and a national skills framework for electricians, advanced HVAC specialists, and controls engineers reflect a long-overdue acknowledgement: without human hands, no server rack hums and no cooling loop closes. In positioning trades as strategic assets, the plan echoes mid-twentieth-century mobilizations where welders and machinists underpinned shipyards and airfields. States and utilities now have a federal invitation to tie economic development packages to training consortia—a competitive lever for those vying to host hyperscale campuses. Yet the workforce vision stops short of addressing the geographic mismatch between where skills reside and where campuses sprout. Apprenticeship programs in the Midwest do little for a cluster in northern Virginia unless labor mobility is supported. Nor does the plan address the attrition risk posed by older workers retiring faster than replacements are trained. A hub-and-framework is laudable, but execution will hinge on state-level implementation. ### Safety and Standards: NIST’s Expanding Role On evaluation and safety, the plan expands the National Institute of Standards and Technology’s role. The emphasis is no longer on abstract principles but on measurement science, model interpretability, robustness, and mission-specific testbeds. This is a pragmatic shift, aligning safety with engineering discipline rather than political debate. By linking safety to metrics, the plan strengthens the hand of developers seeking predictable evaluation criteria. Still, the tilt toward measurement carries risks of its own. The plan largely sidesteps the social impacts that animated earlier debates: algorithmic bias, disinformation, and labor displacement. For a document that aspires to be comprehensive, the omission of social trust could erode legitimacy. Moreover, NIST’s expanded remit comes without clear funding commitments. Scaling measurement science to frontier models requires resources as vast as those being poured into chip fabs and cloud campuses. Without them, standards may lag practice. ### Diplomatic Ambition: Exporting the Full Stack The third pillar, international diplomacy and security, envisions exporting a “full-stack AI package”—chips, models, software, and standards—while tightening compute export controls. The ambition is to make the U.S. not just a leader but an architect of the global AI order. For allies, this is reassurance: access to U.S. technology and standards promises interoperability. For rivals, it signals resolve: location verification and end-use monitoring aim to close loopholes in semiconductor tooling. Yet ambition can shade into overreach. Allies may bristle at controls that feel less like partnership and more like dependency. The plan also risks conflating economic advantage with diplomatic cohesion. Just as OPEC learned that dominance breeds defection, the U.S. may discover that too tight a grip invites parallel supply chains. Technology diplomacy is most effective when it fosters resilience, not monopoly. ### Weakness in Grid Prescription Perhaps the most consequential weakness lies in the grid prescription itself. By privileging dispatchable resources in interconnection, the plan offers clarity—but at the expense of innovation. Enhanced geothermal and long-duration storage are nodded to, yet the near-term bias tilts unmistakably toward natural gas. For hyperscalers desperate for reliable power, this may be welcome news. For climate commitments, it is a precarious gamble. The risk is twofold. First, privileging quick-to-grid resources risks crowding out investment in technologies that need nurturing. Second, by aligning markets around resource adequacy rather than emissions reduction, the plan could exacerbate the very climate vulnerabilities it claims to address. The danger is that the United States enters the AI age with secure servers but a destabilized climate. ### Permitting Acceleration: A Double-Edged Sword The promise of streamlined NEPA reviews and categorical exclusions for low-impact projects is undeniably attractive. Developers have long lamented multi-year permitting cycles that render projects obsolete before shovels break ground. Acceleration can unleash capital and meet surging demand. Yet history warns that acceleration without accountability breeds backlash. The interstate highway program, celebrated for its speed, also left scars of dislocation and inequity. Without safeguards, fast-tracked data centers and power projects may ignite local resistance, from water usage concerns to land-use disputes. Speed, in other words, must be balanced with legitimacy. ### Public Lands and Federal Siting The plan’s invitation to expand federal siting of data centers and generation assets marks a radical departure. Traditionally, federal lands were reserved for mining, drilling, or conservation. To imagine them dotted with AI campuses is to reconfigure the geography of the Republic. The opportunity is vast: access to transmission corridors, renewable resources, and federal oversight could de-risk development. But here too lie hazards. Concentrating critical infrastructure on public lands risks making them geopolitical targets. It also raises questions of federalism: will states acquiesce to federal siting decisions that bypass local control? A strategy that accelerates siting may simultaneously ignite sovereignty disputes. ### The Plan’s Industrial Philosophy Stepping back, the plan represents a philosophy of industrial mobilization. Its ethos is less about nudging markets and more about commanding them. By calling AI an industrial, information, and renaissance revolution, it casts the moment as one akin to railroads or electrification. The ambition is refreshing, especially after decades of policy drift. Yet the mobilization frame also courts risk. Industrial mobilization has historically thrived under conditions of consensus—wartime, existential competition, or crisis. Whether the American public perceives AI as such a cause remains uncertain. If consensus falters, so too may the political durability of the plan. ### Implications for Hyperscalers and States For hyperscalers, the plan offers three actionable levers: permitting acceleration, priority interconnection for dispatchable capacity, and alignment with federal procurement standards. Each of these can be operationalized in corporate strategy. A hyperscaler planning a 1 GW campus can pair behind-the-meter gas turbines with expedited permitting, then advertise compliance with federal security standards to attract enterprise clients. For states, the plan reframes competitiveness. Economic development is no longer just about tax breaks; it is about workforce pipelines, public land siting, and alignment with federal grid directives. States that can bundle these elements will secure campuses and the jobs they bring. Those that cannot will watch demand migrate elsewhere. ### The Unfinished Business of Decarbonization Perhaps the most glaring omission is a coherent pathway to reconcile AI’s energy appetite with climate imperatives. The plan does not explicitly mention climate change or carbon targets—an omission that underscores its political character. By focusing on reliability and resource adequacy while sidestepping decarbonization, the plan leaves the most consequential question unanswered: how to power AI without undermining climate commitments. Here lies the unfinished business. Wind, solar, and batteries—technologies now proven at scale—are capable of being deployed at timelines and costs that match, and in many cases outpace, fossil incumbents. In 2024 alone, the U.S. added nearly 30 gigawatts of wind and solar capacity and over 10 gigawatts of battery storage, with project lead times often measured in months rather than years. These are no longer experimental add-ons; they are the backbone of new capacity additions. The plan’s failure to give them equal priority to dispatchable fossil capacity is not merely a policy oversight but a strategic blind spot. The accelerated deployment schedule for renewables is especially critical in the AI era. Data centers, though demanding in load, are also predictable and locationally concentrated. That predictability dovetails with renewable development zones and the rapid build-out of batteries capable of shifting solar and wind output into evening peaks. Indeed, the same permitting acceleration the plan envisions for fossil plants could supercharge renewable integration if applied symmetrically. An objective view recognizes that reliability remains paramount. Fossil plants today deliver firm capacity with high confidence. But a forward-looking strategy would place wind, solar, and batteries on a glidepath to equal footing—using incentives, procurement mandates, and market design to ensure that the cleanest resources also become the firmest. Without such measures, the United States risks meeting the AI challenge with yesterday’s fuels rather than tomorrow’s ingenuity. ### Conclusion: A Political Document with Practical Implications The *America’s AI Action Plan* is an important and useful document, but it is also undeniably political in nature. Its framing, emphases, and omissions reveal as much about what policymakers are willing to confront as what they are eager to avoid. It recognizes AI as both economic engine and energy disruptor, and it offers practical levers—permitting reform, workforce investment, safety standards, and international diplomacy—that can accelerate U.S. leadership. Yet its weaknesses are equally notable: a bias toward fossil incumbency, an underdeveloped decarbonization strategy, and potential overreach in diplomacy. These gaps show the limits of political consensus more than the limits of technical imagination. Its success will depend less on the words of July 2025 than on the choices of the next decade: whether the United States can convert recognition into reinvention, whether hyperscalers embrace flexibility rather than redundancy, and whether the grid evolves into an architecture of optionality rather than a scaffold for fossil dependence. In the end, the plan’s strengths and weaknesses converge on a single question: will America’s AI future be powered by yesterday’s fuel or tomorrow’s ingenuity? For AIxEnergy, that is the fulcrum upon which both competitiveness and climate balance will turn. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### The Megawatt Mirror: How Data Center Finance Is Rewiring the Energy Landscape URL: https://www.aixenergy.io/the-megawatt-mirror-how-data-center-finance-is-rewiring-the-energy-landscape/ Last updated: 2026-06-30T02:25:10.000Z _This post is for paying subscribers only._ ### The Utility-AI Value Framework URL: https://www.aixenergy.io/the-utility-ai-value-framework-2/ Last updated: 2026-07-29T19:43:09.000Z Throughout history, the electric utility has been a crucible of transformation. The birth of centralized power plants in the late nineteenth century was followed by the slow march of electrification, connecting urban cores and rural farms alike. Later came the rise of Supervisory Control and Data Acquisition (SCADA) systems in the 1960s and 1970s, introducing the first digital nerve endings into a previously analog industry. Each shift required utilities to rethink not only their technologies but also their operating philosophies. Article Summary 0:00 /44.643265 1× Today, we stand at a similar inflection point. The forces reshaping the sector—renewable integration, electrification of transport, surging data-center demand, decarbonization mandates, and climate resilience—coincide with the maturation of artificial intelligence (AI) and generative AI (GenAI). This confluence invites a new way of thinking: how can utilities systematically evaluate, prioritize, and implement AI-driven applications to maximize both business value and social good? What follows is a conceptual value framework for aligning AI and GenAI applications with utility objectives. It is not a prescriptive model but a compass: a way for utilities to rank initiatives, balance short-term wins with long-term transformation, and adapt to shifting regulatory and market realities. ## **Why AI, and Why Now?** The urgency arises from converging pressures. Electricity demand in the United States is projected to grow at roughly 5 percent annually through 2030, a marked reversal from the flat demand of the previous two decades, largely due to electric vehicles and hyperscale data centers. \[1\] Renewable penetration in some regions already exceeds 50 percent during peak hours, challenging grid stability. Meanwhile, utilities face rising capital costs, intensifying extreme weather, and regulators demanding both decarbonization and affordability. AI offers leverage across this complexity. Predictive maintenance algorithms can reduce downtime and extend asset life. Demand forecasting models can balance intermittent renewables with battery dispatch. Generative AI can synthesize regulatory filings, accelerate customer engagement, and even generate optimized control code for distributed resources. Yet the spectrum of possibilities also creates a dilemma: which initiatives matter most, and how should they be ranked? The sheer diversity of potential AI and GenAI applications in the utility sector makes prioritization a formidable task. From predictive maintenance and vegetation management to customer-facing chatbots and regulatory automation, each use case promises some combination of efficiency gains, resilience benefits, or enhanced customer value. Yet utilities operate in a dynamic environment where pressures vary—an IOU facing competitive markets will not weigh the same priorities as a municipal utility striving for community trust, nor will a wildfire-prone California utility mirror the calculus of a Midwestern co-op. The abundance of possibilities underscores the need for structured frameworks that can sort, score, and rank initiatives in a way that reflects both universal industry goals and the unique imperatives of each utility. ## **Top Utility AI Applications** The sheer breadth of AI applications now emerging for utilities reflects just how deeply these technologies can touch every layer of the business. Some are operationally focused—like predictive maintenance, fault detection, and vegetation management—aimed at keeping assets healthy and minimizing outage risk. Others target system-level optimization, from dynamic line rating and demand forecasting to virtual power plants and grid digital twins, reshaping how electricity flows across increasingly complex networks. Still others push into the commercial and regulatory frontier: energy trading algorithms, GenAI for compliance filings, or customer-facing chatbots that redefine engagement. The diversity underscores a central tension: with so many viable use cases spanning reliability, decarbonization, cost control, and customer experience, utilities must make deliberate choices about where to begin, and how to sequence investments in an environment where the technological frontier shifts faster than the regulatory one. Table 1 presents a curated list of 25 high‑potential AI and GenAI applications tailored to electric utilities in the United States—drawn from recent reports, industry pilots, expert analyses, and academic developments. **Table 1** __Utility AI Applications (Top 25)__ | # | Application | Description | | -- | ------------------------------------------------ | ------------------------------------------------------------------------------------------------ | | 1 | Predictive Maintenance | AI monitors transformer networks and equipment to anticipate failures before they happen. | | 2 | Computer Vision Asset Inspection | Automated image-analysis of power lines and substations to expedite field monitoring. | | 3 | AI-Driven Risk Mapping for Climate Threats | Platforms like Rhizome map pole-level wildfire and storm risks to guide preventive investments. | | 4 | Generative AI Field Assistants | Tools like Avangrid’s “First Time Right Autopilot” support real-time technician troubleshooting. | | 5 | Dynamic Line Rating (DLR) | Real-time weather-based line capacity forecasting, boosting transmission throughput 30%. | | 6 | Transformer Failure Prediction | AI sensors forecast failures to maintain continuous operations. | | 7 | Virtual Power Plants (VPPs) | Aggregating storage, EVs, and DERs to supply grid services efficiently. | | 8 | Autonomous Power Plant Operations | Real-time hazard detection through sensor networks reduces human risk. | | 9 | Demand Forecasting & Load Management | AI predicts energy use and dynamically controls DERs to ease peak loads. | | 10 | Smart Grid & Fault Detection | AI enhances grid awareness for quicker fault isolation and restoration. | | 11 | Energy Trading Optimization | AI algorithms optimize procurement timing and pricing in volatile markets. | | 12 | Regulatory Automation & Document Drafting | GenAI auto-generates regulatory filings and compliance reports. | | 13 | Enhanced Call Center AI | GenAI improves customer support via human-like chat, reducing handle time. | | 14 | Technician Onboarding & Training via GenAI | AI offers contextual training and manual access to technicians in the field. | | 15 | Energy Theft Detection & Fraud Monitoring | AI flags anomalies indicating unauthorized usage. | | 16 | Grid Simulation & Digital Twins | Digital models for scenario planning and resilience testing. | | 17 | Agentic AI Workflows for Distribution Planning | Systems like PowerChain enable AI to architect complex grid analysis autonomously. | | 18 | Prosumers & Blockchain-Backed Energy Trading | Combining AI with blockchain to enable decentralized peer-to-peer transactions. | | 19 | Energy Usage Disaggregation for Buildings | AI pinpoints device-level consumption to optimize efficiency. | | 20 | Digital Twin Risk Simulation | Platforms like Neara help forecast vegetation and weather risk impacts on grid infrastructure. | | 21 | Vegetation Management via AI & Satellite Imagery | AiDash and others use AI to reduce costs and wildfire risks in tree trimming. | | 22 | Cloud-based DER Integration Platforms | SaaS solutions for integrating and monitoring distributed energy resources. | | 23 | Building-level Energy Insights for Utilities | Tools like Verdigris monitor facility energy usage and inform utility programs. | | 24 | Smart Meter Analytics & Demand Response | Companies like Itron use AI to manage load and drive customer engagement. | | 25 | Hydrogen Network and Future Scenario Planning | Simulations for integrating hydrogen into future energy mixes using AI-enhanced models. | This proliferation of possibilities is precisely why a new structured evaluation model is needed to enable utilities to prioritze: without a consistent framework to score and compare initiatives, utilities risk chasing novelty rather than value, missing the chance to align AI adoption with their core business priorities and regulatory mandates. ## **The Value Framework** The Utility-AI Value Framework is designed to answer that dilemma. It evaluates AI and GenAI applications across six dimensions. Each focus area is provided with an initial base weight here to reflect its relative importance to utility performance and public mandate: - **Grid Efficiency and Reliability (25 percent)** Grid reliability is the cornerstone of the utility business model. In the United States, outages already cost the economy an estimated $150 billion annually. \[2\] AI promises to reshape this domain. Predictive analytics applied to sensor data can forecast transformer failures before they occur. Machine vision drones can autonomously inspect transmission lines, reducing reliance on manual patrols. Advanced reinforcement learning models can optimize real-time dispatch decisions, increasing efficiency in congested networks. Scoring high in this focus area requires demonstrable reductions in outage duration, frequency, or system losses. For instance, Florida Power & Light’s early use of AI-assisted drones reduced inspection costs and accelerated post-storm restoration—a clear step toward higher resilience. \[3\] Reliability gains translate directly into regulatory goodwill, customer satisfaction, and avoided economic losses. - **Integration of Renewable Energy (20 percent)** The second focus area reflects perhaps the defining challenge of the twenty-first-century grid: integrating variable renewables. Unlike the dispatchable coal and gas plants of old, solar and wind require sophisticated balancing mechanisms. AI excels here. Deep learning models can predict wind and solar output with sub-hourly precision, enabling more efficient dispatch of flexible resources. \[4\] AI-enabled distributed energy resource management systems (DERMS) are another frontier. They orchestrate rooftop solar, behind-the-meter batteries, and EV charging fleets, treating them as aggregated virtual power plants. When layered with generative AI interfaces, operators can ask natural-language questions—“What is the expected ramp at 6 p.m.?”—and receive optimized control strategies. Utilities scoring high in this category are those that use AI not merely to tolerate renewables but to actively unlock higher penetrations. - **Operational Cost Reduction (20 percent)** Utility economics remain under relentless pressure. Inflationary capital costs, interest rate volatility, and fuel price uncertainty make cost efficiency paramount. AI offers relief through predictive maintenance, optimized scheduling, and workforce augmentation. For example, predictive models applied at Duke Energy reduced unplanned maintenance costs by up to 25 percent across gas turbines. \[5\] Generative AI adds another dimension: automating document preparation for regulatory filings, rate cases, or compliance reporting. While not as glamorous as real-time grid control, these applications chip away at overhead, freeing staff to focus on higher-value activities. The framework rewards applications that produce measurable cost savings without sacrificing service quality. A high score here requires both scale (how much cost is addressed) and sustainability (whether savings persist beyond pilot stages). - **Customer Experience and Engagement (15 percent)** Historically, utilities have viewed customer engagement as secondary to infrastructure management. But in the age of prosumers, electrified mobility, and smart homes, customer experience is central. AI-driven personalization enables utilities to tailor rate plans, suggest efficiency upgrades, or nudge EV charging behavior. Generative AI chatbots, if trained responsibly, can resolve customer inquiries with higher accuracy and empathy than legacy IVR systems. Pacific Gas & Electric, for instance, piloted AI-driven text alerts that improved participation in demand response events, yielding both customer satisfaction and peak load relief. \[6\] Applications scoring high in this category not only reduce call center burden but also build trust and engagement—critical for adoption of dynamic rates or distributed programs. - **Regulatory Compliance and Risk Management (10 percent)** Utilities are among the most heavily regulated industries. Compliance with NERC reliability standards, cybersecurity mandates, and decarbonization goals consumes enormous resources. AI can automate compliance monitoring, flag anomalies in real time, and even generate draft filings for regulator review. Equally critical is risk management. Climate-driven extreme weather requires predictive analytics to map fire risk, flood exposure, or asset vulnerability. California utilities now use AI to assess wildfire ignition probability along transmission corridors, enabling pre-emptive shutoffs and vegetation management. \[7\] High scores in this area derive from demonstrable reductions in compliance cost, audit frequency, or risk exposure. While these benefits may be less visible to customers, they are vital for regulatory relations and reputational resilience. - **Innovation and Future-Readiness (10 percent)** The final category looks beyond the immediate horizon. How well does an AI initiative prepare the utility for future disruption—hydrogen blending, distributed microgrids, or AI-driven demand surges from quantum computing? Scoring here reflects strategic optionality. Examples include pilot programs that test generative AI for grid-planning scenario analysis or sandbox environments where operators can simulate extreme future events. European utilities such as Enel have invested in AI-powered digital twins of entire networks, offering a proving ground for future-readiness. \[8\] Though weighted only at 10 percent, this category matters disproportionately for leadership positioning. It ensures utilities remain explorers, not laggards. Each proposed application is scored on a scale of 1 to 5, where 1 equals minimal impact and 5 equals transformative impact. A weighted total value score is then calculated: ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2025/09/Equation.png) This quantitative rubric helps utilities compare otherwise disparate initiatives. For example, how does an AI-driven outage management system compare to a GenAI-powered regulatory assistant? By forcing both onto the same evaluative plane, the framework creates a rational basis for prioritization. Table 2 shows how each of the top 25 utility AI applications might score by each value dimension. **Table 2** ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2025/09/Table2.png) ## **Utility-Specific Weightings** The initial “base weights” (25/20/20/15/10/10) are really just a default compass—a conceptual starting point. In practice, each U.S. utility will recalibrate those weights depending on its geography, asset mix, regulatory environment, and strategic objectives. Here’s how that might play different utility types operating in different market contexts. - **Investor-Owned Utilities (IOUs) in Competitive Markets** - Higher weight on Operational Cost Reduction (25–30 percent) because shareholder pressure makes O&M efficiency paramount. - Customer Engagement also rises (20 percent) since competitive suppliers often encroach, forcing IOUs to differentiate through service. - Regulatory Compliance may drop to 5–7 percent in states with lighter oversight. *Example: Exelon in PJM territory—cost control and customer programs like demand response are front-loaded.* - **Public Power and Municipal Utilities** - Customer Experience carries heavier weight (20–25 percent) because these utilities are owned by the communities they serve. - Reliability stays high, but Innovation is sometimes less emphasized unless city councils or boards make it a priority. *Example: Sacramento Municipal Utility District (SMUD)—customer-centric with aggressive decarbonization pilots, but balanced by affordability mandates.* - **Rural Electric Cooperatives** - Reliability can be weighted as high as 35 percent given the dispersed geography and vulnerability to outages. - Renewable Integration also rises if local co-ops are pursuing community solar. - Cost Reduction matters but is constrained by limited capital budgets. *Example: Tri-State G&T in the Mountain West—focus on reliability and renewables integration.* - **Utilities in High-Renewables States (e.g., California, New York, Hawaii)** - Renewable Integration may jump to 30 percent or more, since curtailment, storage, and DER orchestration dominate. - Regulatory Compliance also rises (15 percent) due to stringent climate mandates. - Reliability remains critical, but it’s often framed around resilience against wildfires, hurricanes, or storms. *Example: Hawaiian Electric—renewable integration is existential, with AI pilots focused on DERMS and forecasting.* - **Utilities in Deregulated vs. Regulated Markets** - Deregulated markets: Trading and cost optimization score higher; compliance lower. - Vertically integrated, regulated utilities: Regulatory & risk management (15–20 percent) becomes central, since commissions scrutinize every AI investment for prudence. Table 3 summarizes the proposed weights by utility type and market context. **Table 3** __Utility AI Value Framework Weights by Utility Type__ | Utility Type | Grid Reliability | Renewables | Cost | Customer | Compliance | Innovation | | ----------------------------------- | ---------------- | ---------- | ---- | -------- | ---------- | ---------- | | Base Framework (Conceptual) | 25% | 20% | 20% | 15% | 10% | 10% | | IOU in Competitive Market | 20% | 15% | 30% | 20% | 5% | 10% | | Public Power / Municipal | 25% | 20% | 15% | 25% | 5% | 10% | | Rural Electric Cooperative | 35% | 20% | 15% | 10% | 10% | 10% | | High-Renewables States (CA, HI, NY) | 20% | 30% | 15% | 10% | 15% | 10% | | Regulated Utility | 25% | 15% | 15% | 10% | 20% | 15% | ## **Different Priorities** 💡 The Utility-AI Value Framework reveals that some priorities are universal while others hinge on context. Virtual power plants, predictive maintenance, and dynamic line rating stand out as foundational bets across utility types, while digital twins and customer-focused analytics emerge as differentiators shaped by governance, markets, and resource mix. In short, the core AI moves are clear, but the accents vary with terrain. Translating abstract value scores into actionable priorities requires sensitivity to context. No two utilities face the same pressures: an investor-owned utility competing in deregulated markets, a municipal provider accountable to its community, a rural cooperative balancing reliability and budget, or a regulated monopoly under commission oversight will each judge AI investments through a different lens. To capture these nuances, the framework has been recalculated across six weighting scenarios—one base model and five utility archetypes. Table 4 shows how the same 25 AI and GenAI use cases rise or fall in priority depending on whether the emphasis is cost control, renewable integration, customer engagement, reliability, compliance, or future-readiness. **Table 4** __Final Score for the Top 5 Application by Utility Type__ | Utility Type | Rank | Use Case | Score | Notes | | -------------------------------- | ---- | ------------------------------------- | ----- | ----- | | Base Framework (Conceptual) | | | | | | — | 1 | Virtual Power Plants (VPPs) | 3.65 | | | — | 2 | Predictive Maintenance | 3.55 | | | — | 3 | Dynamic Line Rating (DLR) | 3.50 | | | — | 4 | Grid Simulation & Digital Twins | 3.35 | | | — | 5 | Cloud-based DER Integration Platforms | 3.25 | | | IOU in Competitive Market | | | | | | — | 1 | Predictive Maintenance | 3.65 | | | — | 2 | Virtual Power Plants (VPPs) | 3.55 | | | — | 3 | Smart Meter Analytics & DR | 3.30 | | | — | 4 | Dynamic Line Rating (DLR) | 3.25 | | | — | 5 | Cloud-based DER Integration Platforms | 3.15 | | | Public Power / Municipal | | | | | | — | 1 | Virtual Power Plants (VPPs) | 3.70 | | | — | 2 | Smart Meter Analytics & DR | 3.40 | | | — | 3 | Predictive Maintenance | 3.35 | | | — | 4 | Dynamic Line Rating (DLR) | 3.30 | | | — | 5 | Grid Simulation & Digital Twins | 3.15 | | | Rural Electric Cooperative | | | | | | — | 1 | Dynamic Line Rating (DLR) | 3.80 | | | — | 2 | Virtual Power Plants (VPPs) | 3.75 | | | — | 3 | Predictive Maintenance | 3.70 | | | — | 4 | Grid Simulation & Digital Twins | 3.55 | | | — | 5 | Demand Forecasting & Load Management | 3.35 | | | High-Renewables State (CA/HI/NY) | | | | | | — | 1 | Virtual Power Plants (VPPs) | 3.75 | | | — | 2 | Dynamic Line Rating (DLR) | 3.60 | | | — | 3 | Grid Simulation & Digital Twins | 3.50 | | | — | 4 | Cloud-based DER Integration Platforms | 3.45 | | | — | 5 | Predictive Maintenance | 3.30 | | | Regulated Monopoly Utility | | | | | | — | 1 | Dynamic Line Rating (DLR) | 3.60 | | | — | 2 | Predictive Maintenance | 3.55 | | | — | 3 | Virtual Power Plants (VPPs) | 3.50 | tie | | — | 3 | Grid Simulation & Digital Twins | 3.50 | tie | | — | 5 | Digital Twin Risk Simulation | 3.25 | | Note: “tie” indicates identical scores sharing the same rank. The early read of the Utility-AI Value Framework tells a clear story: some bets travel well, others depend on the terrain. Across archetypes, virtual power plants rise to the top because they knit together renewables, demand flexibility, and customer participation—especially in high-renewables states and municipals where DERs are already part of the fabric. Predictive maintenance is the quiet constant: whether you’re a competitive IOU chasing O&M efficiency, a municipal defending reliability, or a regulated utility answering to commissions, keeping assets healthy and outages rare never goes out of style. Where context bites, it does so decisively. Dynamic Line Rating leads in rural co-ops and regulated monopolies, a least-cost way to unlock capacity when new wires are hard to build. Digital twins emerge as strategic rather than tactical—valued for scenario testing, resilience modeling, and long-range planning, with different emphases by market and governance. Competitive IOUs gravitate to smart meter analytics and demand response to differentiate on customer experience; municipals lift community-facing tools; high-renewables states double down on orchestration platforms; rural systems favor forecasting and efficiency to stretch scarce capital. The pattern that remains after the noise: VPPs, predictive maintenance, and DLR are the foundational moves. Digital twins—and, in some settings, smart meter analytics—become the differentiators, reflecting each utility’s mix of regulation, market pressure, and resource variability. ## **Conclusions** How does this framework compare to existing industry roadmaps? The U.S. Department of Energy’s AI for Science, Energy, and Security strategy emphasizes national competitiveness but lacks a utility-specific scoring model. \[9\] EPRI’s AI Use Case Taxonomy catalogues applications but does not prescribe prioritization. \[10\] European regulators often emphasize carbon intensity metrics. This Value Framework complements these by offering a flexible, utility-centric tool that aligns technology evaluation directly with business and regulatory imperatives. No framework is without caveats. Three limitations stand out: 1. *Data Quality and Bias:* AI models are only as good as their training data. Utilities must confront gaps in sensor coverage, cybersecurity risks, and potential biases in historical outage data. 2. Workforce Transition: Adoption requires reskilling engineers and operators. AI augmentation must avoid alienating seasoned staff. 3. *Regulatory Acceptance:* Not all regulators may trust AI-driven decisions, particularly if explainability is limited. Building confidence will require transparent pilots and rigorous testing. Moreover, generative AI carries additional risks: hallucinations, privacy concerns, and intellectual property ambiguity. Utilities must tread carefully, adopting governance frameworks that ensure responsible deployment. The electric utility has always been more than wires and substations; it is a public trust. AI and GenAI, properly applied, can help utilities honor that trust in an era of profound change. The Value Framework outlined here offers a conceptual starting point. By scoring initiatives across efficiency, renewables, cost, customers, compliance, and future-readiness, utilities can prioritize wisely, balance innovation with prudence, and remain adaptive in a shifting landscape. This is not a finished product but an invitation—to utilities, regulators, and innovators alike—to test, refine, and iterate. As history has shown, those who embrace structured experimentation often emerge as leaders in the next chapter of energy. ## **Join the AI×Energy Community — Before You’re Playing Catch-Up** The digital infrastructure revolution is not just moving fast—it is rewriting the rules of energy, finance, and technology in real time. If this deep dive into data-center finance opened your eyes, imagine having direct access to the next wave of intelligence *before it hits the headlines*. Subscribe to AI×Energy—free—for the same insider-level analysis, exclusive briefings, and system-level roadmaps trusted by leaders in energy, AI, and infrastructure. You will be joining hundreds of executives, investors, and technologists who rely on AI×Energy to anticipate capital flows, decode regulatory shifts, and spot the hidden forces shaping tomorrow’s grid. Do not read about the future secondhand—subscribe to the playbook. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## **Notes** 1. U.S. Energy Information Administration. *Annual Energy Outlook 2024*. Washington, D.C.: U.S. Department of Energy, 2024. 2. U.S. Department of Energy. *Economic Benefits of Increasing Electric Grid Resilience to Weather Outages*. Washington, D.C.: DOE, 2013. 3. Florida Power & Light. “Drone Technology and Storm Response.” *FPL Technical Brief*, 2021. 4. National Renewable Energy Laboratory. *Machine Learning for Solar Forecasting*. Golden, CO: NREL, 2020. 5. Duke Energy. “Predictive Maintenance Case Study.” 2022\. [https://www.dukeenergy.com](https://www.dukeenergy.com/?ref=aixenergy.io). 6. Pacific Gas & Electric. “AI-Driven Customer Engagement in Demand Response.” *PG&E Pilot Report*, 2021. 7. California Public Utilities Commission. *Wildfire Mitigation and AI Analytics*. Sacramento, CA: CPUC, 2022. 8. Enel Group. “Digital Twin Innovation for the Grid.” 2022\. [https://www.enel.com](https://www.enel.com/?ref=aixenergy.io). 9. U.S. Department of Energy. *Artificial Intelligence for Science, Energy, and Security: A Strategic Plan*. Washington, D.C.: DOE, 2020. 10. Electric Power Research Institute (EPRI). *AI Use Case Taxonomy for the Electric Power Sector*. Palo Alto, CA: EPRI, 2021. ### How AI Workload Flexibility Is Shaping the Grid’s Future URL: https://www.aixenergy.io/how-ai-workload-flexibility-is-shaping-the-grids-future/ Last updated: 2026-06-30T02:25:00.000Z AI-driven data centers are reshaping the grid, turning from static loads into flexible assets that support reliability and decarbonization. With Google leading the way, this shift demands new policies, pricing models, and equity safeguards to align digital growth with grid resilience. _This post is for paying subscribers only._ ### The Machine in the Control Room: Reviewing NERC’s White Paper on AI and ML in Real-Time System Operations URL: https://www.aixenergy.io/the-machine-in-the-control-room-reviewing-nercs-white-paper-on-ai-and-ml-in-real-time-system-operations/ Last updated: 2026-07-14T02:10:50.000Z In November 2024, the North American Electric Reliability Corporation released a white paper with a sentence that carried the finality of a bell toll: “the ‘genie’ cannot be put back into the bottle” (p. 1). It was not the first time technology had crept into the control room, but it was the first time the institution tasked with reliability admitted, publicly and without hedging, that artificial intelligence would soon sit alongside human operators. The report, *AI and ML in Real-Time System Operations*, reads as both a warning and a blueprint. It calls to mind earlier moments when machines arrived in spaces once thought irreducibly human: the cockpit, the nuclear control room, even the trading floor. Each time, the results were double-edged—efficiency gained, safety compromised, and only later, after failure and reform, a more stable equilibrium. Author's summary 0:00 /45.766531 1× The stakes are high. The grid is no longer just an engineering system; it is the beating heart of the digital economy. Artificial intelligence has not only become a tool of operators—it has become a source of demand unprecedented in scale. Data centers are already consuming gigawatts, with projections to double or triple that appetite by 2030\. This duality—that AI is both a consumer of power and a potential custodian of reliability—creates a paradox. The machine arrives as both burden and promise. [Artitical Intelligence and Machine Learning in Real-Time System OperationNERC White Paper, November 2024Whitepaper-AI and ML in Real-Time System Operations.pdf1 MBdownload-circle](https://www.aixenergy.io/content/files/2025/09/Whitepaper-AI-and-ML-in-Real-Time-System-Operations.pdf "Download") NERC’s most important insight is that AI cannot be left to improvisation. It requires a contract. Just as in aviation, where autopilot freed pilots from monotony but left them unprepared for crisis, AI in grid operations risks producing “out-of-the-loop” operators (p. 33). The report insists on clarity: is the machine in active control, serving as co-pilot, or offering decision-support? (pp. 30–33). Without such distinctions, the operator is left in a twilight zone, unsure when to trust, when to override, and when to simply watch. History instructs us here. Three Mile Island was not caused by a single failure but by a torrent of alarms that overwhelmed human cognition. Likewise, in 2009, Air France Flight 447 descended into the Atlantic not because of mechanical defect but because pilots, long habituated to automation, lost the thread of their own situational awareness. The NERC report does not recount these tragedies, but it echoes their lessons: machines change the role of humans, often in ways unanticipated until the moment of crisis. To avoid repeating history, the contract must be written, rehearsed, and enforced. This contract is not theoretical. It must be embedded in procedure, procurement, and training. In AIxEnergy, I have described this as a “cognitive contract,” in which roles are not left implicit but spelled out in the same language as switching orders. It is a recognition that reliability is not only electrical but cognitive: a joint system of human and machine, bound together by discipline and trust. ## Learning in Public The report’s survey of 47 entities offers a candid snapshot of an industry in motion. Nearly half are still “learning,” while others have implemented AI for load forecasting, renewable prediction, or cyber defense (p. 41). Some admitted they did not know whether their tools qualified as AI at all (p. 45). It is as if the industry is peering through fog, unsure of the contours of the machine already in its midst. Policy stances diverge widely. Some organizations have banned public large language models outright, citing cybersecurity and reliability risks (p. 44). Others experiment cautiously, allowing AI to draft outage reports or assist in technical writing. The landscape resembles the early days of cloud computing, when utilities treated it first as an intruder, then as a tolerated adjunct, and finally as indispensable infrastructure. AI is poised on the same trajectory. What is banned today will, once hardened by standards and trust, become tomorrow’s baseline. Trust is not assumed; it is engineered. NERC insists that operators must be able to interrogate why a model produces its outputs, likening it to reviewing state estimator residuals (p. 42). Confidence intervals, drift detection, and anomaly alerts are not luxuries but reliability controls (pp. 49–50). Without them, AI systems risk becoming black boxes, and black boxes have no place in real-time reliability. Cybersecurity amplifies the stakes. AI is described in the report as a double-edged blade: a tool for defenders but also a weapon for adversaries. The “Morris II” worm study is cited as evidence of AI’s capacity to propagate attacks (p. 29). NERC recommends extending STRIDE and MITRE ATT&CK into AI contexts and adopting MITRE ATLAS to map adversarial tactics (pp. 26–27). For an industry already steeped in CIP compliance, this mapping is invaluable. It renders AI risk legible, not alien, to existing practices. Yet the challenge extends further. In my judgment, the supply chain of AI models must itself be treated as critical infrastructure. Training data, model provenance, and update management require audit trails. Without them, adversaries will insert poison into the very systems meant to protect reliability. Just as firmware updates are controlled and logged, so too must model updates be scrutinized. ## Training the Operator of Tomorrow Perhaps the most striking emphasis of the report is on training. Operators must not only learn to use AI but to question it, to spot drift, and to override it when necessary (pp. 26–28). High-fidelity simulators are singled out as indispensable, but with a warning: they must replicate the messy ambiguities of reality, not sanitized case studies (p. 34). This resonates deeply with my own observations. Today’s simulators prepare operators for deterministic contingencies. They do not prepare them for probabilistic systems that present confidence intervals and shifting baselines. The operator of tomorrow must be trained not only to respond to contingencies but to interrogate the machine itself. Did the model drift? Did it misclassify? Did its confidence collapse? These questions must become as natural as checking frequency or voltage. In AIxEnergy’s work, we have called for integrating hazard layers, congestion indices, and interconnection delays into training environments. Only by rehearsing against structural stresses—blackouts entangled with misbehaving models, congestion compounded by probabilistic errors—can operators be truly prepared. Reliability has always been about preparing for the worst. AI demands we broaden our imagination of what the worst can be. ## The Road Ahead The NERC report’s strengths are evident. It grounds AI in human factors, offers a shared vocabulary of automation modes, integrates cybersecurity frameworks, and calls for transparency (p. 58). Its silences are also clear. The survey is small. The document avoids prescriptive standards. And most glaringly, it omits discussion of AI’s role as a load driver, despite the obvious relevance of AI-driven data center demand to reliability planning (p. 2). That silence is telling. It reflects a choice to confine scope, but also leaves a gap at the heart of the debate. From my perspective, these omissions underscore the need for follow-on work. NERC has offered a foundation. The task now is to build. Mode contracts must be codified. Explainability dashboards must be standard. AI supply chains must be secured. Competency programs must be institutionalized. These steps align with the EthosGrid™ governance framework I have developed, which insists that machine behavior must be auditable, transparent, and orchestrated according to principles of integrity. Released in November 2024, the NERC white paper is a landmark, not because it solves the AI challenge but because it acknowledges it. Its insistence on human engagement, explainability, and cybersecurity is both timely and necessary. Its silences remind us that the hardest questions remain unanswered. For those of us at the intersection of AI and energy, the lesson is clear. The grid remembers. It remembers 1965 and 2003\. It remembers storms and near misses. It remembers the costs of complacency. It will also remember how we chose to integrate its new machine mind. If we design collaboration, we may gain resilience. If we improvise, we risk fragility. The contract is ours to write. The future will hold us accountable. North American Electric Reliability Corporation. *AI and ML in Real-Time System Operations*. Atlanta: NERC, November 2024. ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ### Twenty-Five Companies Rewiring Power for the Compute Age URL: https://www.aixenergy.io/twenty-five-companies-rewiring-power-for-the-compute-age/ Last updated: 2026-06-30T02:24:05.000Z These 25 companies that span supply creation, grid intelligence, demand flexibility, and hyperscaler strategy are reengineering the global power system to match the unprecedented speed and scale of the AI era. _This post is for paying subscribers only._ ### The AI Wars: An Electricity Arms Race Between Superpowers URL: https://www.aixenergy.io/the-ai-wars-an-electricity-arms-race-between-superpowers/ Last updated: 2026-06-30T02:23:56.000Z By 2030, AI data centers could consume 945 TWh—more than Japan uses today. The U.S. faces grid strain as AI drives demand toward 50 GW by 2028, while China aligns energy, chips, and propaganda under state policy. The real AI war is over electrons, narrative, and infrastructure supremacy. _This post is for paying subscribers only._ ### Fine-Tuning AI for the Grid: How Industry-Specific LLMs Are Transforming Utilities URL: https://www.aixenergy.io/fine-tuning-ai-for-the-grid-how-industry-specific-llms-are-transforming-utilities/ Last updated: 2026-07-29T19:43:15.000Z On the surface, a large language model like GPT-4, Claude, or Gemini can appear omniscient—capable of answering nearly any question with speed and eloquence. These models are trained on vast libraries of public information, from literature to code to news reports. Yet, inside the high-stakes environment of an electric utility control room, that breadth can be a liability. Article Summary 0:00 /66.795102 1× In power system operations, the meaning of a single word can carry consequences. A casual misinterpretation of “trip” could imply a short journey in everyday conversation but, in grid operations, it means the sudden disconnection of a line or generator—a potentially costly or dangerous event. “Islanding” might conjure tropical images for the uninitiated; for a system operator, it signals an isolated section of the grid running independently, a delicate and sometimes risky condition. In such a context, ambiguity is not an option. This is why utilities and their technology partners are turning to industry-specific large language models—fine-tuned versions of general-purpose LLMs, trained on the vocabulary, data, and compliance frameworks unique to the sector. These models are not just better conversationalists; they are trained to think in kilowatts, outages, dispatch signals, and regulatory clauses. ## **From General to Grid-Specific** Fine-tuning starts with data—specialized datasets drawn from SCADA event logs, IEEE research archives, FERC and NERC regulatory documents, maintenance records, and engineering reports. The goal is to build a language model that is fluent in the technical lexicon and operational patterns of the utility world. Security is another pillar. Regulated industries cannot afford to send sensitive operational data to open, uncontrolled environments. Many deployments take place on-premises or in secure, private cloud environments, where access is tightly controlled. Alignment with NERC-CIP standards is a baseline, not an afterthought. These fine-tuned models are also engineered to work within existing utility workflows. They integrate with SCADA, GIS mapping tools, AMI analytics, and load forecasting systems—not as external consultants but as embedded, API-driven team members. And unlike their generalist cousins, they are optimized for specific, high-value tasks: power system optimization, DER integration guidance, asset failure prediction, and compliance reporting. ![](https://storage.ghost.io/c/e3/c9/e3c9e740-20f7-4e9c-9a31-131fc1166819/content/images/2025/08/AI-Specialization-in-Utility-Models.png) ## **Proof in the Control Room** The clearest example to date is eGridGPT, developed by the National Renewable Energy Laboratory in collaboration with the University of Michigan, the Lower Colorado River Authority, and Oregon State University. Designed for the grid control room, eGridGPT interprets operational data, runs physics-based digital twin simulations, and generates decision-support recommendations—always with a human operator in the final control loop.1 The approach is mirrored in the private sector. Schneider Electric has applied AI to load forecasting, energy management, and digital twin creation for infrastructure-intensive operations such as AI data centers. Working with Nvidia, Schneider has achieved a reported 20 percent reduction in cooling energy use and 30 percent faster project development for high-density compute facilities.2 These are measurable, operational gains—precisely the kind of outcome that makes AI adoption in utilities compelling. ## **Barriers to Scale** While the promise is clear, adoption is not without friction. - Data availability remains a hurdle; proprietary utility data is valuable but often siloed, and labeling it requires experts who understand the operational nuances. - Computational cost can be significant—full model training is often infeasible, pushing most toward fine-tuning as a cost-effective compromise. - Trust and explainability are essential; an AI’s recommendation must be auditable, especially in regulated environments where a wrong decision can trigger compliance violations. - Integration complexity with legacy systems means deployment is as much an IT challenge as an AI one. ## **How Utilities Can Proceed** A successful implementation often follows a staged approach: 1. Define use cases with clear business value, such as automating compliance reports or optimizing outage responses. 2. Curate datasets that combine proprietary operational logs with trusted public sources, ensuring regulatory compliance in handling sensitive data. 3. Select the model—either fine-tune an existing LLM or build a custom one for highly sensitive applications. 4. Train and evaluate the model using supervised learning and reinforcement learning with human feedback (RLHF). 5. Deploy securely via private cloud or on-premises systems, embedding AI governance into day-to-day operations. ## **Research Momentum** Multiple initiatives are already expanding the knowledge base: - NREL’s eGridGPT project integrates AI into real-time operational decision-making. - A Texas A&M University study explores LLM applications in power flow analysis, forecasting, and specialized knowledge bases. - ComEd’s Grid Analytics team is developing AI tools to enhance operational efficiency and automation. ## **Emerging Applications** As these models mature, new utility-specific use cases are emerging: - Grid Management: Real-time analytics to anticipate and resolve operational anomalies. - Predictive Maintenance: Asset failure forecasting to reduce unplanned outages. - Energy Demand Forecasting: Advanced prediction models to optimize resource allocation. - Energy Management – Building-level energy optimization for cost and efficiency gains. ## **Looking Ahead** The adoption curve for industry-specific LLMs in utilities may be gradual, but the trajectory is clear. From early pilots to embedded operational tools, fine-tuned AI is beginning to occupy a role in grid management that is as indispensable as SCADA systems or GIS maps. With proper governance, secure integration, and targeted training, these models are on track to become standard components of the utility technology stack—capable of delivering operational efficiency, enhanced reliability, and better customer outcomes. *Shailesh Jain is a Principal Advisor at SCE with 30+ years in T&D, specializing in strategic asset planning and risk-based strategies to enhance grid reliability and resilience. He leads cross-functional efforts advancing energy transition and modern grid solutions. He can be reached at* [*https://www.linkedin.com/in/sjain-smartgrid/*](https://www.linkedin.com/in/sjain-smartgrid/?ref=aixenergy.io)*.* ## **Join the AI×Energy Community — Before You’re Playing Catch-Up** The digital infrastructure revolution is not just moving fast—it is rewriting the rules of energy, finance, and technology in real time. If this deep dive into data-center finance opened your eyes, imagine having direct access to the next wave of intelligence *before it hits the headlines*. Subscribe to AI×Energy—free—for the same insider-level analysis, exclusive briefings, and system-level roadmaps trusted by leaders in energy, AI, and infrastructure. You will be joining hundreds of executives, investors, and technologists who rely on AI×Energy to anticipate capital flows, decode regulatory shifts, and spot the hidden forces shaping tomorrow’s grid. **Do not read about the future secondhand—own the playbook.** ## Sign up for AIxEnergy AI × Energy unpacks how artificial intelligence is revolutionizing energy systems—from accelerating demand growth and optimizing smart grids to enhancing renewables, storage, and market dynamics. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. ## **References** 1. National Renewable Energy Laboratory. *eGridGPT: Trustworthy AI in the Control Room*. Golden, CO: NREL, 2024\. [https://www.nrel.gov/docs/fy24osti/87740.pdf](https://www.nrel.gov/docs/fy24osti/87740.pdf?ref=aixenergy.io) 2. Business Insider. “As AI Surges, Schneider Electric Is Teaming Up with Nvidia to Help Companies Build Data Centers That Can Handle AI’s Energy Demands.” July 16, 2025\. [https://www.businessinsider.com/schneider-electric-nvidia-ai-data-centers-designs-energy-demands-2025-7](https://www.businessinsider.com/schneider-electric-nvidia-ai-data-centers-designs-energy-demands-2025-7?ref=aixenergy.io) 3. Texas A&M Engineering. “Large Language Models May Revolutionize the Electric Energy Sector.” June 24, 2024\. [https://engineering.tamu.edu/news/2024/06/large-language-models-may-revolutionize-the-electric-energy-sector.html](https://engineering.tamu.edu/news/2024/06/large-language-models-may-revolutionize-the-electric-energy-sector.html?ref=aixenergy.io) 4. Utility Analytics Institute. “Large Language Models and Grid Analytics.” 2024\. [https://utilityanalytics.com/large-language-models-grid-analytics/](https://utilityanalytics.com/large-language-models-grid-analytics/?ref=aixenergy.io) 5. Salesforce. “Artificial Intelligence in Utilities.” 2024\. [https://www.salesforce.com/energy-utilities/artificial-intelligence-utilities/](https://www.salesforce.com/energy-utilities/artificial-intelligence-utilities/?ref=aixenergy.io) 6. TRC Companies. “Benefits of AI and Machine Learning for Utility Companies.” 2024\. [https://www.trccompanies.com/insights/benefits-of-ai-and-machine-learning-for-utility-companies/](https://www.trccompanies.com/insights/benefits-of-ai-and-machine-learning-for-utility-companies/?ref=aixenergy.io) 7. Master of Code. “Generative AI in Energy and Utilities.” 2024\. [https://masterofcode.com/blog/generative-ai-in-energy-and-utilities](https://masterofcode.com/blog/generative-ai-in-energy-and-utilities?ref=aixenergy.io) 8. Verdigris Technologies. “Energy Management Platform.” 2024\. [https://www.verdigris.co/](https://www.verdigris.co/?ref=aixenergy.io) ### AI-Powered Sustainability: A Strategic Roadmap for U.S. Corporate Leaders URL: https://www.aixenergy.io/ai-powered-sustainability-a-strategic-roadmap-for-u-s-corporate-leaders/ Last updated: 2026-06-30T02:22:08.000Z AI has evolved from isolated sustainability pilots into a strategic, enterprise-wide engine—driven by regulation, investor demand, climate risk, and technological maturity—that is rapidly becoming essential for competitive advantage in corporate ESG performance. _This post is for paying subscribers only._ ### How Electricity Became the Hidden Frontline in the U.S.–China AI Rivalry URL: https://www.aixenergy.io/how-electricity-became-the-hidden-frontline-in-the-u-s-china-ai-rivalry/ Last updated: 2026-06-30T02:21:59.000Z The AI conversation has long centered on algorithms, data, and hardware—but beneath every inference lies a more vital dimension: electricity. The key question isn’t just whose code is smartest—but whose power stays on. _This post is for paying subscribers only._ ### Embedding Intelligence into the Energy Fabric: AI/ML as a Defining Feature of 6G Networks URL: https://www.aixenergy.io/embedding-intelligence-into-the-energy-fabric-ai-ml-as-a-defining-feature-of-6g-networks/ Last updated: 2026-06-30T02:21:51.000Z _This post is for paying subscribers only._ ### The Algorithm and the Turbine: How AI Re-Engineers Wind Power Performance URL: https://www.aixenergy.io/the-algorithm-and-the-turbine-how-ai-re-engineers-wind-power-performance/ Last updated: 2026-06-30T02:21:43.000Z AI-driven platforms have transformed wind farms into intelligent systems that predict turbine failures, optimize performance, and strategically bid energy, boosting reliability and reducing costs across operations and markets. _This post is for paying subscribers only._ ### AI in Solar Forecasting and Reliability URL: https://www.aixenergy.io/ai-in-solar-forecasting-and-reliability/ Last updated: 2026-06-30T02:21:34.000Z AI is transforming solar energy from a variable resource into a smart, predictive system—enhancing forecasting accuracy, reducing downtime, and enabling more reliable, cost-efficient grid integration worldwide. _This post is for paying subscribers only._ ### How AI is Rewiring the Brain of the Battery URL: https://www.aixenergy.io/how-artificial-intelligence-is-rewiring-the-brain-of-the-battery/ Last updated: 2026-06-30T02:21:25.000Z AI is turning batteries from static energy tanks into smart, real-time decision-makers—maximizing market profits, managing degradation, and stacking value across services. _This post is for paying subscribers only._ ### The Grid Behind the Algorithm: Reviewing the DOE’s Case for AI-Driven Infrastructure Reform URL: https://www.aixenergy.io/the-grid-behind-the-algorithm-reviewing-the-does-case-for-ai-driven-infrastructure-reform/ Last updated: 2026-07-14T02:10:56.000Z In a year marked by historic energy stress and exponential data growth, a new report from the U.S. Department of Energy (DOE) offers a consequential claim: the United States may be unable to power its own artificial intelligence (AI) future. The document, titled *Report on Evaluating U.S. Grid Reliability and Security* (Washington, DC: Department of Energy, July 2025), asserts that the fate of American AI leadership is increasingly tied to the adequacy of its electric grid.¹ The conclusions are far-reaching, and while they reflect a strong point of view about the relationship between energy security and digital innovation, they also raise important questions. The modeling and scenarios presented warrant serious attention, even as readers consider the broader context in which these projections are offered. Listen to the author's summary 0:00 /172.041438 1× ### The AI Load Is Real The report incorporates a national projection of 50 gigawatts (GW) of new data center load by 2030, largely driven by AI workloads. This estimate synthesizes projections from McKinsey & Company, the Electric Power Research Institute (EPRI), Lawrence Berkeley National Laboratory (LBNL), and S&P Global.² These data centers, serving as the operational backbone of the AI economy, require significant and continuous energy input to support model training, inference, and deployment across sectors. AI-driven data centers are expected to account for nearly half of the projected 115 GW total load growth by 2030\. The remainder stems from broader electrification trends, including transportation and manufacturing. AI’s contribution is notable not only for its scale, but for its concentration in specific geographic regions, which may create regional stress even in a nationally adequate grid. ### A Hundredfold Increase in Risk? The DOE’s simulations suggest that, under a scenario where planned power plant retirements proceed and new firm capacity is not brought online at scale, the grid’s reliability could degrade significantly. In the "Plant Closures" scenario, Loss of Load Hours (LOLH)—a measure of outage frequency—increases by a factor of 100 by 2030\. PJM, a major regional grid operator, could experience over 1,000 hours of annual shortfall in a worst-case weather year.³ Normalized Unserved Energy (NUSE), indicating unmet demand, exceeds threshold levels across most regions with concentrated data center growth. While these modeled outcomes present a stark picture, they rest on specific assumptions: that all retirements occur as announced, only Tier 1 generation projects are completed, and the full 50 GW of AI-related load manifests on schedule. The modeling approach, which is deterministic rather than probabilistic, allows for transparency but does not capture adaptive strategies—such as behind-the-meter generation, load shifting, or locational flexibility—that may emerge in practice.⁴ ### From Reliability to Geopolitics Beyond technical modeling, the report places energy security within a broader geopolitical frame. It warns that a failure to scale grid infrastructure to meet AI demand could allow "adversary nations \[to\] shape digital norms and control digital infrastructure," posing long-term strategic risks.⁵ This language elevates energy planning from operational necessity to strategic imperative. This framing reflects an emerging understanding: energy infrastructure underpins not just physical industries but the digital economy itself. As AI systems become more integral to national productivity and daily life, the resilience and scalability of the power grid will increasingly determine the feasibility and location of their deployment. ### Parsing Policy from Prognosis The report advocates a shift in grid planning to accommodate load growth from AI. Specifically, it encourages greater alignment between transmission investment, permitting timelines, and the expected demand from compute-intensive infrastructure. It also calls for more balanced generation additions, arguing that the current pipeline—largely comprised of intermittent wind and solar—may be insufficient to provide firm capacity when needed. Here it is important to distinguish between what can be deployed quickly and what may be needed in the long run. While natural gas is often cited as a reliable firm generation source, manufacturing and infrastructure constraints have created a six-year backlog for new gas-fired power plants. Similarly, coal and nuclear units typically require seven to ten years to permit and construct, making them ill-suited to solve near-term reliability challenges. By contrast, wind, solar, and battery storage systems are the only utility-scale technologies that can be reliably deployed within a 12-month development timeline. These technologies are not only widely available and cost-competitive, but also increasingly capable of providing essential grid services when paired with storage and demand response. Their value lies not only in carbon reduction, but in speed, scale, and flexibility. Thus, while the report rightly underscores the importance of firm capacity, the broader solution set must include accelerated integration of renewables, modernized grid planning, and dynamic resource coordination. Achieving reliability in the face of changing load profiles, increasing electrification, and weather variability will require all available tools—not just conventional generation. ### Conclusion: Signal in the Noise The DOE’s *Report on Evaluating U.S. Grid Reliability and Security* is both a technical artifact and a call to action. Its modeling underscores the scale of the transformation underway in energy and computing—and the consequences of inaction. While interpretations of its findings may vary, the central insight is clear: powering the AI economy requires not only advanced algorithms but robust, forward-looking infrastructure. As such, the conversation must now expand. AI developers, infrastructure investors, regulators, and grid planners all have a role to play in building a system capable of sustaining this next wave of technological growth. In the emerging era of AI, reliable electricity is not merely an input—it is a strategic enabler. ### Join the AI×Energy Community The digital infrastructure revolution isn’t slowing down—it’s accelerating. If you found this deep dive into data-center finance illuminating, you won't want to miss what’s next: expert analysis, exclusive insights, and system-level roadmaps at the intersection of AI, energy, and infrastructure. **Subscribe to AI×Energy** for free—the weekly dispatch you need to stay ahead of capital flows, grid innovation, sustainability strategies, and the hidden forces shaping tomorrow’s energy intelligence. [Subscribe Now](https://www.aixenergy.io/#/portal/signup) --- ¹ U.S. Department of Energy, *Report on Evaluating U.S. Grid Reliability and Security*, July 2025, https://www.energy.gov. ² See Electric Power Research Institute, “Powering Intelligence: Analyzing Artificial Intelligence and Data Center Energy Consumption,” March 2024, https://www.epri.com/research/products/3002028905; Shehabi et al., “2024 United States Data Center Energy Usage Report,” Lawrence Berkeley National Laboratory, https://escholarship.org/uc/item/32d6m0d1; and S&P Global, “US Datacenters and Energy Report,” 2024. ³ DOE, *Grid Reliability and Security*, 7–9. ⁴ Ibid., 10–13, 20–21. ⁵ Ibid., Executive Summary, 1. ### The Billion-Dollar Brain: How Silicon Valley’s AI Arms Race Is Reshaping the Power Grid URL: https://www.aixenergy.io/the-billion-dollar-brain-how-silicon-valleys-ai-arms-race-is-reshaping-the-power-grid/ Last updated: 2026-06-30T02:21:08.000Z I has entered its baroque phase—capital-heavy, power-hungry, and hardware-driven. Data centers rival cities in load, reshaping grids and energy politics. Tech giants pour billions into compute, while VCs chase vertical AI. The race is less about algorithms than land, power, and electricity itself. _This post is for paying subscribers only._ ### AIxEnergy Convergence Map: Technologies, Companies, Deployments, and Capital Flows URL: https://www.aixenergy.io/aixenergy-convergence-map-technologies-companies-deployments-and-capital-flows/ Last updated: 2026-07-29T19:40:38.000Z The convergence of AI and energy is redefining modern power systems. At its core lies a layered AI‑powered grid: intelligent sensors collect real‑time data, ML forecasts load and renewable output, reinforcement learning orchestrates dispatch and market participation, and automated systems. _This post is for paying subscribers only._ ### The Power Beneath the Code: How AI and Data Centers Are Rewiring Global Energy Investment According to IEA URL: https://www.aixenergy.io/the-power-beneath-the-code-how-ai-and-data-centers-are-rewiring-global-energy-investment/ Last updated: 2026-07-14T02:11:03.000Z Artificial intelligence, in its rise from computational experiment to cultural force, has become more than an abstract function of algorithms and hardware. It is an industrial force, with demands that echo the steel mills and textile factories of the nineteenth century. As large language models, generative AI systems, and synthetic cognition scale into every corner of modern infrastructure, they drag with them an insatiable need: reliable, baseload, uninterrupted power. The [*World Energy Investment 2025*](https://www.iea.org/reports/world-energy-investment-2025?ref=aixenergy.io) report from the International Energy Agency is both mirror and map of this transformation. It reveals a new industrial revolution unfolding in the shadows of cloud campuses and undersea cables—a revolution that is reshaping global investment, rewriting energy geopolitics, and reconstructing the electric grid in the image of synthetic minds.1 ## The New Load: AI's Electrification Imperative The numbers alone are enough to command attention. Annual investment in data centers has surged by 67 percent over the past two years. By 2030, global electricity demand from these digital behemoths could double to 950 terawatt-hours. That would make data centers, driven largely by AI inference and training, one of the largest electricity consumers on the planet—exceeding entire national grids. This transformation is not speculative. Companies like Microsoft, Amazon, Google, and Meta are already inking multibillion-dollar power purchase agreements (PPAs) across continents. Their models demand constant uptime; their users expect split-second responsiveness. To deliver that, hyperscalers are becoming utilities in their own right, not only consuming power but commissioning its very production. To support this digital hunger, the IEA estimates that over $170 billion in new power generation investment will be required by 2030 just to meet data center demand. But this is not simply a matter of quantity. It is a fundamental shift in the direction, structure, and philosophy of energy investment. ## The Nuclear-Geothermal Convergence As AI redraws the geography of electricity demand, it is also reconfiguring the technological palette of generation. One of the most consequential results is the return of nuclear energy—not as an ideological lightning rod, but as an industrial necessity. Small modular reactors (SMRs), long regarded as boutique solutions awaiting a business case, now have one. Their alignment with data center load profiles is uncanny: both are capital-intensive, designed for steady-state operation, and yield optimal returns when running near full capacity. What was once theoretical now looks inevitable. In the United States, Japan, India, and South Korea, nearly 27 gigawatts of new or recommissioned nuclear capacity are being mapped to data center growth zones. Behind the curtain of techno-optimism lies hard-nosed economics. Data center operators are not pursuing nuclear partnerships to win environmental awards. They are pursuing them to secure electricity at scale, at consistency, and at cost. Geothermal, too, is rising from obscurity to strategic prominence. Advanced geothermal—leveraging horizontal drilling, closed-loop systems, and subsurface analytics—is now attracting nearly $1 billion in annual investment. Developers like Fervo Energy and Sage Geosystems have inked agreements with Google and Meta respectively, creating a new class of behind-the-meter baseload power tailored to synthetic cognition. These shifts reflect a core truth: AI workloads are not just demanding more power. They are reshaping the character of the generation fleet itself. ## Grid Bottlenecks and the Return of Energy Islanding And yet, the energy transition is not seamless. The Achilles' heel of the AI-powered grid is not generation—it is transmission. While more than $1 trillion is being invested annually in power generation, grid investment remains stranded around $400 billion. That gap is now a chokepoint. Permitting delays, transformer shortages, and aging infrastructure are creating friction across the grid. In response, energy-intensive users are turning to energy "islanding"—the co-location of data centers with private power assets, from solar farms with battery storage to modular reactors and advanced geothermal. This echoes the early industrial age, when factories were built beside rivers and coal seams. Today, the mineral is replaced by cognition, and the river by fiber. The digital factory does not follow labor markets. It follows energy security. ## From Emissions to Economics: The Investment Pivot Contrary to climate conventional wisdom, emissions reductions are not the primary driver of clean energy investment today. The IEA makes this point explicit: the current boom in renewable, nuclear, and storage investment is being driven by energy security, cost competitiveness, and technological alignment—not carbon policy alone. This is particularly true in the AI sector. In Pakistan, falling solar prices allowed households and businesses to import 19 gigawatts of rooftop panels in 2024—nearly half the country's total grid capacity. The result was a 10 percent decline in national grid sales, reshaping the utility business model overnight. That shift was not the result of government mandates. It was the result of retail users acting in their own economic interest. Solar investment will reach $450 billion in 2025, with battery storage topping $66 billion. AI workloads are indirectly underwriting much of this surge, pushing utilities, developers, and financiers to meet the new electrified load with technologies that scale rapidly and operate flexibly. ## Data Centers as the New Anchor Tenants In infrastructure finance, a subtle but seismic change is underway. For decades, utilities were the cornerstone of power project underwriting. Long-term offtake agreements guaranteed returns. Today, data centers have become the new anchor tenants. Banks and institutional investors have reallocated capital accordingly. In 2024, 96 percent of primary project finance from banks and 93 percent from institutional investors flowed into low-emissions generation. AI-enabled infrastructure is now treated as investment-grade collateral. And with good reason. A hyperscale data center operator, with a 20-year clean energy offtake agreement, offers a risk-adjusted return profile that rivals or surpasses traditional utilities. For investors, this is not a tech story. It is a cash-flow story. ## Geopolitical Rewiring: Where the Grid Goes, Power Follows But the AI-energy convergence is not evenly distributed. China now commands over one-quarter of global energy investment, and has leveraged its dominance in solar manufacturing, critical minerals, and grid infrastructure to shape a formidable techno-energy complex. The United States, buoyed by the Inflation Reduction Act, has mounted a response—but policy uncertainty looms beyond the 2026 electoral cycle. Meanwhile, Africa, with 20 percent of the global population, attracts just 2 percent of clean energy investment. Unless structural reforms and international finance mechanisms are mobilized, the energy demands of AI could exacerbate existing inequities. Those without reliable, affordable, and clean power risk being left behind—not just economically, but cognitively. ## The Grid for Minds If the twentieth century built a grid for machines—factories, refineries, motors—the twenty-first must build a grid for minds. These are not metaphorical minds, but synthetic agents: systems that process, predict, compose, and decide at scale. Unlike human workers, these minds do not sleep. They do not rest. They demand electricity continuously, predictably, and with increasing sophistication. They are not merely applications riding on the grid. They are becoming the grid’s primary drivers. The IEA's *World Energy Investment 2025* report is not merely a ledger of capital flows. It is a ledger of philosophical realignments. What we are witnessing is not an evolution of demand. It is a redefinition of what the energy system is for. 1. International Energy Agency. *World Energy Investment 2025*. 10th ed. Paris: IEA, 2025\. [https://www.iea.org/reports/world-energy-investment-2025](https://www.iea.org/reports/world-energy-investment-2025?ref=aixenergy.io). ### Toward Net Zero URL: https://www.aixenergy.io/toward-net-zero/ Last updated: 2026-06-30T02:16:28.000Z As AI drives soaring demand, U.S. data centers race to cut energy and water use. With liquid cooling, 24/7 clean energy, and bold innovation, net-zero is within reach by 2030–2040. Sustainability is becoming reality, not just marketing. _This post is for paying subscribers only._ ### Digital Giants and the Race for Net-Zero Data Centers URL: https://www.aixenergy.io/digital-giants-and-the-race-for-net-zero-data-centers/ Last updated: 2026-06-30T02:16:18.000Z Data centers are surging in energy and water demand. Can we build truly net-zero facilities? Discover the technologies, policies, and strategies reshaping digital sustainability. Read the full "Toward Net Zero" report now at CleanPowerShift.com. _This post is for paying subscribers only._ ### The Intelligence Revolution: How AI Is Rewriting the Rules of Energy URL: https://www.aixenergy.io/the-intelligence-revolution-how-ai-is-rewriting-the-rules-of-energy/ Last updated: 2026-07-14T02:11:09.000Z The fusion of artificial intelligence and energy is no longer a theoretical construct or the stuff of distant future forecasts. It is here, arriving with seismic force, reshaping the way we generate, consume, and innovate energy systems globally. In its groundbreaking *Energy and AI* report, the International Energy Agency (IEA) delivers the most sweeping, data-driven analysis to date of this new reality, capturing both the staggering promise and the profound risks. What emerges is a portrait of a sector on the cusp of a transformation as fundamental as the arrival of electricity itself—and a call to action that the energy and tech sectors must answer together. At the heart of the IEA’s findings lies an unflinching truth: there is no AI without energy. The explosion in artificial intelligence capabilities over the past decade has spawned an equally explosive demand for electricity. The data centers that train and deploy generative models are not incidental infrastructure; they are fast becoming the heavy industries of the digital era. A single AI-focused data center today consumes as much electricity as 100,000 households. The next generation of mega-centers, already under construction, will require twenty times that amount, rivaling the energy needs of entire nations. By 2030, global data center consumption is projected to more than double to 945 terawatt-hours. By 2035, it could exceed 1,200 terawatt-hours—more than the total annual electricity use of Japan. The geography of this growth compounds the challenge. In the United States, nearly half of all data center capacity is packed into five regional clusters, concentrating unprecedented strain on local grids already grappling with electrification pressures. Northern Virginia, the epicenter of America’s data economy, now faces energy demands that rival those of major industrial sectors. Without urgent upgrades to grid infrastructure, the IEA projects that up to twenty percent of new data center projects could be delayed or canceled altogether due to connection bottlenecks. Meeting this rising tide of demand will require a diversified approach to energy sourcing. Renewables—led by solar and wind—are expected to account for roughly half of the new electricity supply needed to power AI expansion, thanks to their cost-competitiveness and rapid deployment times. However, renewables alone cannot shoulder the entire load. Dispatchable power sources—natural gas, advanced geothermal, and a coming wave of small modular nuclear reactors—must be part of the mix to guarantee stability and resilience, particularly in regions where grid flexibility is limited. Without these firm resources, the energy backbone of the digital economy could falter just as it becomes indispensable. Ironically, the same force straining the system also offers the tools to fix it. The IEA details how AI itself is becoming critical to energy sector optimization. Intelligent forecasting of renewable generation, predictive maintenance, real-time fault detection, and dynamic grid balancing are no longer theoretical exercises; they are operational necessities. AI-driven fault detection, for instance, could cut outage durations by up to fifty percent and unlock as much as 175 gigawatts of additional transmission capacity without building new lines—a stunning figure that eclipses the additional load projected from AI data centers by 2030. Yet the most profound impact of AI on the energy sector may not come from efficiency gains alone. It may come from radically accelerating the pace of innovation itself. Historically, the energy sector has operated on painfully long R&D cycles. New technologies like high-efficiency solar cells, advanced batteries, and carbon capture materials often take decades to progress from laboratory to market. The IEA reveals that AI could collapse these timelines, unleashing a new era of rapid discovery and deployment. In fields ranging from photovoltaics to synthetic fuels, from carbon capture to low-carbon cement, AI models are already demonstrating the ability to identify promising materials and configurations at a fraction of the time and cost of traditional research methods. In biomedicine, similar AI applications have slashed discovery timelines by a factor of 45,000\. Transposed into the energy sector, this capability could be nothing short of revolutionary. Only 0.01 percent of next-generation solar materials have been experimentally explored. AI could make the other 99.99 percent accessible in ways previously unimaginable. However, the path to an AI-driven energy renaissance is not automatic. Major barriers remain. The energy sector is behind other industries in digital maturity, hampered by fragmented data access, limited AI expertise, and persistent cybersecurity concerns. Only two percent of venture capital raised by energy startups flows to companies leveraging AI at their core. Without aggressive policy support to expand digital skills, unlock open data, and secure critical infrastructure, these barriers could prevent the sector from realizing the full potential of AI. Security risks loom large. Critical mineral supply chains, particularly for gallium and rare earth elements vital to AI hardware, are dangerously concentrated. China currently controls ninety-nine percent of refined gallium production, a vulnerability that could escalate as AI infrastructure scales. Cybersecurity risks are growing in parallel. The energy sector has witnessed a threefold increase in cyberattacks over the past four years, often enhanced by AI-driven offensive capabilities. Yet AI also offers defensive advantages, allowing companies to detect and neutralize threats far faster and more accurately than human operators alone. Emerging and developing economies find themselves at a precarious crossroads. These nations account for half of global internet users but less than ten percent of global data center capacity. If they can invest in reliable power infrastructure and leapfrog legacy systems, AI could become a catalyst for building smarter, cleaner energy grids. If they fail, the digital divide—and the energy divide—will only widen. Environmental concerns surrounding AI’s electricity appetite are valid, but the IEA analysis suggests they are manageable with smart policy interventions. Even under high-growth scenarios, emissions from data centers are expected to remain below one and a half percent of total energy sector emissions by 2035\. More importantly, AI-driven efficiencies across industrial processes, transportation, and grid management have the potential to deliver net emissions reductions that far outweigh the footprint of the data economy itself. Still, the IEA cautions against magical thinking: AI is a powerful tool, but it is no substitute for strong climate policy. In the end, *Energy and AI* reads as both a warning and an invitation. The energy and tech sectors are now inextricably linked. Their futures will rise—or fall—together. The choices made in the next few years will determine whether AI becomes a driver of sustainability and resilience, or a new source of systemic vulnerability. Collaboration between industries, accelerated policy frameworks, investment in infrastructure, and a shared commitment to innovation will be essential. The intelligence revolution is not waiting for permission. It is already rewriting the rules of energy. The only question is who will adapt quickly enough to lead it—and who will be left behind. **Source** International Energy Agency (IEA). *Energy and AI: World Energy Outlook Special Report*, 2025\. Available at [https://www.iea.org/reports/energy-and-ai](https://www.iea.org/reports/energy-and-ai?ref=aixenergy.io)​.