On October 9, the Federal Energy Regulatory Commission rejected most of PJM’s proposal for connecting large loads during periods of insufficient generating capacity. It accepted a large-load definition and registry but rejected the proposed Interim Resource Adequacy Service framework and associated bring-your-own-new-capacity provisions. The same day, FERC approved Duke Energy Carolinas’ transmission agreement changes, including a fifteen-year minimum revenue commitment for qualifying large loads. The decisions leave developers facing a consequential uncertainty: the right to build a data center can advance well ahead of the right to operate it at the intended scale.1
These were different proceedings, and neither established a national settlement for AI infrastructure. Taken together, they show why the next wave of development will depend on agreements that reconcile federal transmission and market rules with state retail authority. FERC has retained its responsibilities, while states and utilities must resolve obligations that a regional grid operator cannot simply impose on retail customers. The commercial question is whether those separate decisions produce a dependable service commitment that can support investment.
The underlying difficulty is the mismatch between a customer’s ability to change its plans and a utility’s ability to reverse its investment. A developer can delay a campus, reduce its size, or move computing elsewhere after the utility has committed to facilities intended to serve it. If the customer retains that flexibility without paying for the exposure it creates, other customers effectively insure the development plan. Large-load tariffs should make the allocation of that risk explicit before construction begins.
PJM’s rejected proposal addressed a genuine timing problem. Eligible loads would have connected without matching new capacity while accepting curtailment as system conditions deteriorated. FERC objected to provisions that appeared to impose obligations directly on retail customers, decide which retail customers would be interrupted, and establish compensation resembling a retail rate. The commission’s rejection concerned the proposed legal and commercial arrangement; it did not establish that flexible service is technically infeasible.2
| Timing | Milestone | What to watch |
|---|---|---|
| November 16, 2026 | Five regional operators, including PJM, respond to FERC’s large-load show-cause proceedings. | How proposed connection terms assign costs and reconcile regional requirements with state authority. |
| November 20, 2026 | Southwest Power Pool’s response is due. | Whether SPP offers approaches to connection rights and cost recovery that other regions could adopt. |
| 30 days after filings | Answers are due under the timetable described by FERC’s Office of Public Participation. | Which objections identify unresolved cost exposure, jurisdictional conflicts, or unworkable operating obligations. |
| December 2026 | PJM’s planned capacity auction for the 2029–30 delivery year. | Whether procured capacity meets the reliability requirement. The clearing price alone cannot establish supply adequacy. |
| January 1, 2027 | Dominion’s GS-5 class takes effect. | How applicable service commitments and minimum payments protect other customers if demand falls short. |
| Throughout the period | Further backstop proceedings and PJM scheduling updates. | When additional supply could be procured, when it could enter service, and which customers would pay. |
Schedule reflects information available October 21, 2026. Subsequent notices may revise filing deadlines or auction timing.
The market-design objection was equally consequential. Commissioner David Rosner questioned curtailing these loads before calling on voluntary demand response and excluding their demand from the capacity-market demand curve. His concern was that the market could signal adequate supply while operators still depended on involuntary interruptions. Any revised approach must reconcile the amount of capacity procured with the service customers are actually entitled to receive.3
A customer that knowingly purchases a limited service has accepted an operating constraint with a potential economic benefit. Assigning an interruption priority to retail customers through a regional tariff raises additional questions about authority and discrimination. A workable agreement must establish the customer’s rights, identify who can order a reduction, and make the resulting obligation consistent with the grid operator’s planning assumptions. Without that consistency, the system can either procure too little capacity or charge customers for capacity that their accepted service limits make unnecessary.
Responsibility remains distributed. FERC oversees interstate transmission service and wholesale market rules, while state regulators govern much of retail service and rate design. Utilities develop service plans and execute customer agreements; regional operators study system impacts and operate the bulk power system under approved rules. State and local authorities also exercise important siting powers. A project needs the relevant approvals to work together, because a zoning approval cannot establish transmission deliverability and a retail contract cannot override regional operating requirements.4
Duke’s approval provides a useful example of how a financial commitment can address part of this problem. For qualifying new or transferred loads of at least 100 megawatts, its framework requires the transmission customer serving the load to support a committed revenue contribution based on 75 percent of scheduled megawatts over fifteen years. Shortfall payments apply when actual contributions fall below the commitment. This is a transmission revenue obligation, rather than a guarantee to purchase 75 percent of projected electricity consumption or to reimburse every cost associated with the project.5
That distinction prevents an attractive headline number from substituting for a cost analysis. A minimum payment reduces exposure to underutilization only to the extent that the resulting revenue covers the relevant investment. The value of the commitment also depends on the creditworthiness of the obligor and the security available after construction. Commissioner Lindsay See supported Duke’s proposal while questioning whether its protections went far enough; Commissioner Judy Chang dissented over financial security and the treatment of network-upgrade costs in shared transmission rates.6
Three separate exposures deserve attention in every large-load proceeding. The first is investment that becomes stranded because demand fails to arrive. The second is the allocation of shared network costs even when the customer uses all the electricity it promised to consume. The third is the effect of additional demand on regional capacity and energy costs. A minimum bill may address the first exposure while leaving the other two largely untouched, so compliance with a tariff cannot by itself establish that existing customers are financially unaffected.
Regulators should also distinguish an attributable infrastructure cost from a market price effect. A dedicated substation can often be associated with an identifiable customer, while an increase in regional capacity prices reflects interactions among demand, available supply, auction rules, and expectations. Assigning every resulting price increase to the newest customer requires a defensible counterfactual and an administrable method. Treating those effects as irrelevant would be equally unsatisfactory; the policy response may require supply procurement or market reform alongside customer-specific charges.
PJM’s capacity results show why these distinctions matter now. The July auction for the 2028–29 delivery year cleared at the approved ceiling of $325 per megawatt-day. Even after including resources procured through the Fixed Resource Requirement approach, committed supply fell 6,831 megawatts below PJM’s reliability requirement. PJM characterized the result as a smaller reserve cushion and greater risk, rather than a determination that the system would necessarily fail to serve demand.7
December’s auction should not be presented as the immediate expiration of price protection. The extended price collar covers the 2029–30 delivery year as well. The difficult condition is that constrained prices coexist with a reserve shortfall and long construction lead times. The collar can moderate exposure to auction prices, but the remaining resource requirement still has to be met through additional supply, dependable demand reductions, or changes in the load that ultimately takes service.8
PJM’s proposed Reliability Backstop Procurement was intended to help address the supply gap. FERC’s September 29 order accepted and suspended the proposal, assigning a February 28, 2027, effective date subject to further proceedings. PJM canceled the planned September 30 start and said the procurement timetable would be revised. February 28 is therefore not a confirmed auction date, and the December capacity auction approaches without the originally planned backstop sequence having occurred.9
This creates different investment conditions for existing and proposed generation. An operating plant may benefit from scarcity, subject to its hedges, performance, and the applicable market rules. A new plant must obtain a revenue commitment sufficient to finance construction while carrying the risk of equipment, fuel, permitting, and interconnection delays. Higher short-term prices and a financeable long-term project are related, but they are not interchangeable. A customer seeking earlier service needs to understand which of those problems its proposed supply agreement actually solves.
Flexible service, additional supply, and firm service supported by financial commitments offer three broad commercial approaches. Projects can combine them, and none creates an automatic entitlement to connect. Their value depends on the constraint at the proposed location and on whether the agreement changes the utility’s or grid operator’s actual obligations.
For flexible service, the central question is how much dependable reduction is available during the conditions that determine investment and reliability requirements. A facility able to reduce demand for two hours may provide limited relief from a constraint lasting several days. A reduction available during a regional capacity shortage may not address a local transmission contingency. The agreement needs measurable limits on response time, duration, frequency, notice, and recovery, along with clear authority to call an event and consequences for nonperformance.
The portfolio matters as much as the individual facility. Several customers may depend on the same alternate computing location, weather-sensitive supply, or fuel infrastructure, making their claimed flexibility less independent than it appears. Moving workloads can also transfer demand to another constrained part of the system. EPRI’s DCFlex demonstrations are relevant because they test operational capabilities that could support future service agreements. Their results must still be evaluated against the local constraint and translated into terms that utilities can administer and lenders can assess.10
Financial design can reinforce or undermine those operating terms. A customer may remain obligated to support sunk network costs while earning compensation for a verified reduction that avoids an additional capacity requirement. Those payments cover different obligations and can coexist. But an agreement that charges for firm service, requires uncompensated reductions, and gives no identifiable benefit for flexibility leaves the customer little reason to invest in the capability. Equally, a discount should not be justified by a claimed avoided cost that the utility continues to incur.
Supporting additional supply addresses a different exposure. A long-term contract can help finance new generation or storage, but an annual energy purchase does not necessarily provide dependable capacity during system stress. A contract with an existing generator may reassign output without expanding available supply. The analysis should establish what is additional, what capacity is accredited, whether it is deliverable, and when it becomes available relative to the customer’s demand.
The interval between the two service dates deserves its own commercial terms. If a data center starts before its contracted resource, the agreement should identify the interim supply obligation and the consequences of a delayed generator. If generation arrives first and the customer cancels, the financing still needs an adequate source of repayment. A matched quantity of annual megawatt-hours does little to resolve either exposure unless the contract addresses timing and performance.
Firm service can instead require substantial minimum charges, contributions toward infrastructure, collateral, and exit payments. Willingness to accept those obligations may improve a project’s credibility, but it cannot create unavailable equipment or eliminate an engineering constraint. Developers should compare sites on the total cost of obtaining the required service, including delay and the obligations that remain if their business plan changes. A lower posted electricity rate can be outweighed by a later connection or a larger commitment to unused capacity.
The state responses provide useful evidence of how these obligations can be structured. AEP Ohio’s data center tariff combines minimum billing-demand provisions with a contract term comprising up to four ramp years plus eight years, along with financial protections. Its minimum-demand formula varies with contract capacity, so the commonly cited 85 percent figure is an incomplete description. Ohio’s next challenge is to determine how any additional flexible-service arrangement would interact with those existing commitments and PJM’s requirements.11
Pennsylvania’s PUC has adopted a model framework to guide utility filings on cost responsibility, collateral, ramp schedules, exit provisions, and study expectations. The framework creates a basis for more consistent treatment, but individual utility proceedings still have to establish enforceable terms. For developers and financiers, the distinction between commission guidance and an effective utility tariff remains material to the certainty of a service commitment.12
In Virginia, Dominion’s GS-5 class takes effect January 1, 2027, with extended service commitments and minimum-payment protections for applicable large loads. Implementation will test whether the customer commitments, expected demand, and construction schedules remain consistent as projects develop. In the Carolinas, Duke Energy Carolinas’ approved federal transmission framework must operate alongside the relevant retail arrangements. Neither approach should be described as a uniform rule for every utility in its state or states.13
Across these jurisdictions, forecast discipline should begin before a utility makes an irreversible commitment. A study request, an executed service agreement, and an operating load provide different evidence of demand, and planning should distinguish them. Deposits and milestones can require a developer to support its request as the utility incurs greater costs. Transfer or release provisions can preserve some customer flexibility where another qualified customer can use the capacity without imposing additional uncompensated costs.
Contract design also has to account for bargaining incentives. A developer has reason to preserve options across potential sites; a utility has reason to obtain sufficient commitment before ordering equipment. A progressively binding agreement can accommodate both interests by increasing the customer’s exposure as the utility’s recoverable alternatives narrow. Security should follow the remaining exposure through construction and operation, with reductions tied to demonstrated cost recovery or an approved reassignment of obligations.
For utility executives, the immediate priority is to connect each major expenditure to the customer milestone and contractual remedy that support it. Regulators should require a clear account of the costs still borne by other customers under credible delay, downsizing, and default scenarios. Developers should bring the service agreement into site selection before land and equipment commitments make relocation expensive. Investors should distinguish a forecasted connection date from an enforceable service commitment and an energized facility when valuing contracted megawatts.
Government economic-development teams have a parallel responsibility to assess investment that can actually be served on a credible schedule. Announced demand alone establishes neither deliverability nor net public benefit. Counsel and consultants can add particular value by reconciling retail contracts with transmission obligations, operating protocols, and financing documents. The same reduction in demand should not receive inconsistent treatment across those agreements or be counted twice as a reliability contribution.
The next ninety days contain several opportunities to make these choices concrete. The federal response deadlines and Virginia’s implementation date are published; PJM’s backstop timetable remains unsettled. The calendar below distinguishes those milestones so that a proposed procurement is not mistaken for an available source of capacity.14
The electricity industry has always invested in assets whose useful lives exceed the planning horizons of individual customers. Large AI loads intensify that familiar problem through their scale, development speed, and ability to change location or operating patterns. The appropriate response is to make the customer’s retained flexibility and the utility’s committed investment visible in the same agreement, with payment and operating obligations that reflect both.
For a proposed campus, the decisive diligence question is what happens if the load, supply, or network arrives later than expected. A credible agreement identifies who provides interim service, who can reduce demand, and which party absorbs the unrecovered cost. Projects that resolve those questions before construction give utilities a stronger basis to invest and lenders a stronger basis to finance. The October decisions make that work more urgent by leaving the essential bargains to be completed across institutions that share responsibility for the result.
Notes
1. Ethan Howland, “FERC Rejects Most of PJM’s Large Load Interconnection Plan,” Utility Dive, October 12, 2026; and “FERC Approves Duke Large Load Transmission Rules Despite Cost-Shift Concerns,” Utility Dive, October 13, 2026. ; .
2. Howland, “FERC Rejects Most of PJM’s Large Load Interconnection Plan,” reporting FERC’s October 9, 2026, decision in Docket No. ER26-3515-000. .
3. David Rosner, “Commissioner Rosner’s Concurrence in PJM Interconnection, L.L.C. (ER26-3515-000),” Federal Energy Regulatory Commission, October 9, 2026. .
4. Federal Energy Regulatory Commission, Office of Public Participation, “Cost Transparency and Customer Protection Initiatives in FERC’s June 2026 Show Cause Orders for Large Load Interconnections,” accessed October 21, 2026. The cited page is an educational explanation of the proceedings and jurisdictional responsibilities. .
5. Judy Chang, “Commissioner Chang’s Dissent in Duke Energy Carolinas, LLC,” Docket Nos. ER26-2691-000 and ER26-2691-001, Federal Energy Regulatory Commission, October 9, 2026, background and n. 1, citing Duke Energy Carolinas, LLC, 197 FERC ¶ 61,020, para. 35 (2026). .
6. Lindsay See, “Commissioner See’s Concurrence in Duke Energy Carolinas, LLC,” Docket Nos. ER26-2691-000 and ER26-2691-001, Federal Energy Regulatory Commission, October 9, 2026; Chang, “Dissent.” ; .
7. PJM Interconnection, “PJM Capacity Auction Procures 138,318 MW of Generation Resources as Work Continues to Address Growing Electricity Demand,” PJM Inside Lines, July 14, 2026. .
8. Monitoring Analytics, 2026 Quarterly State of the Market Report for PJM: January through June, sec. 5, August 22, 2026. .
9. PJM Interconnection, “FERC Accepts PJM Reliability Backstop Proposal,” PJM Inside Lines, September 30, 2026. .
10. Electric Power Research Institute, “DCFlex,” accessed October 21, 2026. Discussion of contractual design and portfolio performance represents the author’s analysis, rather than a claim that the demonstrations have validated a particular tariff. .
11. AEP Ohio, “Data Center Tariff,” accessed October 21, 2026, especially “Contract Term,” “Collateral Requirement,” “Minimum Demand Charges,” and “Assigning Retail Capacity to Another Customer.” .
12. Pennsylvania Public Utility Commission, “PUC Releases Final Order Establishing First-of-Its-Kind Large Load Model Tariff Framework,” May 13, 2026. .
13. Virginia State Corporation Commission, “SCC Data Center Initiatives: Facts and Figures,” February 2, 2026; and “SCC Issues Order on DEV Biennial Review 2025,” November 25, 2025; See, “Concurrence.” ; .
14. FERC, Office of Public Participation, “Cost Transparency and Customer Protection”; PJM Interconnection, “PJM Board Directs Action on Resource Adequacy, Affordability and Large Loads,” July 27, 2026; PJM, “FERC Accepts PJM Reliability Backstop Proposal”; Virginia SCC, “SCC Issues Order on DEV Biennial Review 2025.” Dates reflect information available October 21, 2026; subsequent notices may revise procedural or auction schedules. ; ; .
