• A confidential “speed to market” pathway, now called custom contracts, would get a 120-day review and need not carry the tariff’s protections.
• One intervenor reads Xcel’s own workpaper as leaving a $29 million annual transmission shortfall at 2,000 MW, a gap that term length and collateral do not close.
• At an illustrative $22 million monthly bill, 60 months of credit support covers 50% of the exit claim on a 10-year term and 25% on a 20-year term.
• A standard tariff with a conditional lane is appearing in several states and at FERC; speed holds up where eligibility terms are published and local upgrade costs are assigned directly to the customer.
Xcel Energy filed its Colorado large-load tariff on April 2. The Public Utilities Commission opened evidentiary hearings 202 days later, on October 21, and has scheduled five hearing days through October 28 without naming a decision date.1 Proceeding 26AL-0137E would govern new customers of 50 megawatts or more, and most of the argument concerns familiar terms: how long the contract runs and how much collateral stands behind the exit fee.2 One provision sits outside those terms and may matter more. Xcel's "speed to market" pathway, since renamed custom contracts, would let the utility and a developer negotiate a confidential agreement that need not carry the tariff's safeguards and would face a 120-day review.2
Utilities in many states are building this structure, a standard tariff for most customers and a negotiated route for the one in a hurry. The idea has merit. A utility that can serve 50 megawatts from existing headroom, or a customer that arrives with its own generation, should not have to wait for a rate case. The difficulty lies in what the negotiated route does to protections that a tariff states in advance: they become terms to be bargained, and the commission sees them last, under a clock. This article draws on the Colorado record, which is unusually detailed, to consider what a fast lane would need to carry before a commission could approve it quickly.
What the pathway promises
Xcel's witnesses put the case for the pathway in two sentences. Jack Ihle, the company's vice president for data centers and large loads, testified that the concept "is necessary to provide an option for large load customers who require greater commercial flexibility." Michael Pascucci, director of regulatory policy, testified that a custom contract would remain "consistent with the company's core commercial terms, including cost responsibility, minimum commitments, and appropriate risk protections."2
Courtney Fieldman, testifying for Rewiring America and the Southwest Energy Efficiency Project, disputes that a promise of equivalent protection amounts to a mechanism that delivers it. In her account the tariff and its form service agreement spell out the minimum monthly bill, the capacity reduction payment, the exit fee and the credit support, and a special contract is not automatically bound by any of them. She finds nothing in the record that subjects the contract capacity in a speed-to-market deal to a verification standard, a non-refundable payment or a letter of credit before the utility commits capital.3 A compressed schedule leaves less time for that verification.
Xcel points to its Pine Island agreement with Google in Minnesota as an analog.3 The agreement is still under review in Minnesota, so it offers only a partial test of the idea. Xcel announced in February that Google would fund 1,900 megawatts of new clean resources and cover new grid infrastructure costs, subject to approval by the Minnesota commission.4 On October 1 the Minnesota Attorney General's office told that commission that Xcel had not shown Google would pay all attributable costs or that other customers were protected against stranded costs, put the value to other ratepayers closer to $1 billion in net costs than the $1.1 billion in net benefits Xcel claims, and objected that almost all dollar amounts were redacted.5 These are advocates' positions and not findings, but they point to where bespoke deals stall. Even when criteria exist, review depends on whether the numbers can be seen.
The arithmetic a side deal can rewrite
The tariff's protections depend on one another, which is also why a side deal can loosen them. The minimum monthly bill is set at 80 percent of contract capacity, and the exit fee equals the sum of the minimum bills remaining in the term. A customer can reduce its contract capacity on 24 months' notice by paying the difference in minimum bills multiplied by the months remaining, after which both the minimum bill and any future exit fee are recalculated on the lower number.3 Xcel proposes a 15-year term, including an optional ramp of up to five years.3 (Xcel's announcement describes termination charges as recovering the remaining costs of project-specific upgrades, while the tariff language Fieldman quotes defines the fee through remaining minimum bills. I use the latter here.)6
Fieldman objects that the protections reset downward while the physical plant stays in place. A customer that never reaches its maximum, or later steps down, leaves the utility holding facilities sized for the higher number, and the capacity reduction payment recovers only the difference in minimum bills, not the remaining revenue requirement of those facilities.3 She then turns to Xcel's own workpaper for a 2,000-megawatt data center scenario, in which network reinforcement costs $1.24 billion and carries an annual revenue requirement of $124 million, while standard transmission revenue at full demand is $95 million. By her reading the gap is about $29 million a year even for a customer fully compliant with the tariff.3 The gap equals about 23 percent of the revenue requirement. Because I could not review Xcel's rebuttal, this remains an intervenor's reading of a company workpaper and should be weighed as one.
If the reading is right, the source of the gap matters. Standard rates average cost across a system, while the upgrades are concentrated where the load lands. A legal-news summary of the proceeding, which I have not confirmed against the docket, reports a $1.24 billion figure for shared-grid upgrades concentrated in the Denver metro area, apparently the same number, and says staff asked for location-based surcharges or credits.7 The gap recurs every year the customer is served, whatever the contract length, so a longer term or more collateral does not close it; only direct assignment of the upgrade costs would. Virginia's commission took that route when it ordered Dominion on July 31 to require a mandatory contribution in aid of construction for the substations and lines that exist only to connect a large load. Google and Amazon had asked for voluntary contributions; the commission made them mandatory.8 A custom contract that weakens the capacity reduction payment would widen the gap, which makes the cost-causation terms the ones to hold fixed in any fast lane.
The dispute over term length concerns a different risk, the one that arises when a customer leaves. The Sun reports Xcel's illustration of a $22 million monthly bill for a 500-megawatt campus and credit support of 60 months of bills.2 The earlier AIxEnergy analysis of the October hyperscaler deals used those inputs to estimate roughly $1.3 billion of collateral. Applied across the proposed terms, the same inputs show a fixed number of months covering a smaller share of the claim as the term lengthens.
| 10-year term | 15-year term | 20-year term | |
|---|---|---|---|
| Maximum exit claim | $2.64 billion | $3.96 billion | $5.28 billion |
| 60 months of bills ($1.32 billion) as share of claim | 50 percent | 33 percent | 25 percent |
| Years at full load if a five-year ramp sits inside the term | 5 | 10 | 15 |
The table treats the $22 million bill as the floor in every month of the term, an upper bound because the tariff sets the minimum at 80 percent of contract capacity and a ramp can lower it early on. The shares in the second row do not depend on that choice, since the claim and the security scale together. Xcel proposes 15 years, commission staff recommends 20 and the Data Center Coalition proposes 10.2 Intervenors add that a 15-year term with a five-year ramp leaves as little as ten years at full load, and the Conservation Coalition would start the clock after the ramp.9 Google proposed capping the exit fee at the shorter of five years of minimum demand charges or the remaining term, which would make 60 months of credit support equal to the whole claim at any term beyond five years. Staff would pair its 20-year term with credit lines of at least 60 months of minimum bills and a bank guaranty of 150 percent of remaining rate base, revolving annually.9 The positions diverge over years six onward, and the sizing of credit support determines who pays for them. Security keyed to rate base grows as upgrades are energized and falls as they depreciate, so it follows the utility's exposure more closely than a fixed count of bill-months, which scales only with the contract.
What earning speed looks like
The most useful idea in the Colorado record for a fast lane is Fieldman's proposal for two tracks. A special contract that adopts the tariff's residual-cost and flexibility terms without material deviation would get streamlined approval. One that seeks weaker terms would trigger full findings and a quantified comparison, measured against a floor based on the net present value of committed cost, with the same standard applied across all arrangements serving a customer or its affiliates, protective-order access for intervenors and annual cumulative reporting.3 That addresses the 120-day problem more directly than lengthening the clock would, since review time ought to scale with how much of the contract departs from the standard. The tariff case took 202 days to reach hearing, with many parties testifying, because its terms were open. A deal that adopts published terms and lists its deviations should need far less, and one that reopens every term cannot be reviewed responsibly in 120 days, however confidential its commercial terms are.
Fieldman also asks that any speed-to-market contract show a binding commitment, such as a non-refundable payment or letter of credit sized to a meaningful share of projected infrastructure cost, before the utility treats the customer's capacity figure as the basis for investment.3 The earlier AIxEnergy piece on community-scale contracts described AEP Ohio's tariff, which pairs a service agreement that starts at energization with reimbursement of 100 percent of buildout costs if a customer cancels or delays by more than 12 months, and a term of the ramp plus eight years. Starting the clock at energization and putting money at risk before construction addresses stranded cost earlier than adding years to the term afterward.
Flexibility carries value in a fast lane only when it can be relied on. On June 5 FERC approved the Southwest Power Pool's Conditional High Impact Large Load Service, which is non-firm, curtailable and usable for up to seven years while firm service is built.10 The sunset matters as much as the curtailment, since the speed is explicitly a bridge. Duke's Nicholas Institute estimated that the largest 22 balancing areas could accommodate 76 gigawatts of new load if it can be curtailed for 0.25 percent of its maximum uptime, about 85 hours a year, though the study did not consider transmission constraints.11 The caveat applies directly in Colorado, where the binding limits are local and aggregate headroom does not help a load that lands on a constrained corridor. The Energy Office there proposed a flexible-service schedule outside the standard term and exit rules, and Fieldman proposed a performance credit against the minimum bill for verified demand response and storage, with measurement protocols and public reporting.3, 9 Either design requires flexibility to be demonstrated before it is credited.
Eligibility raises a further problem. A Denver witness warned that loads can game the 50-megawatt threshold by subdividing, and Fieldman recommends treating any entities with a common parent, more than 50 percent common ownership or shared control over interconnection decisions as a single set.3, 9 The same gap is wider in a custom contract, where each deal is reviewed alone. Queue order raises a related issue: United Power has a pending 600-megawatt interconnection request with Xcel and described "delay and a lack of transparency" in how it has been handled.9 A discretionary lane makes the order in which customers are served a regulatory matter, and the materials I reviewed do not say how Colorado would decide it.
One tariff, one lane, repeated
The same structure appears in federal and state dockets. On June 18 FERC issued section 206 show-cause orders to the six regional grid operators, giving them 60 days to justify their tariffs or reform them across five areas that include flexible transmission service and co-location. Its fact sheet says nothing in the orders displaces state authority over retail rates and that states police retail cost shifting.12 The federal answers will take longer than the 60 days suggest, because FERC granted abeyance requests in August; responses from CAISO and its transmission owners are due November 16 and answers to them December 16.13 After the commission rejected most of PJM's connection plan on October 9, as the earlier AIxEnergy analysis of that order describes, the retail terms moved further toward the states.
Pennsylvania's commission voted 4 to 0 on October 1 for a tentative order that would curtail eligible new large loads lacking sufficient new capacity first during PJM emergency events, keep data centers out of the critical-load class and create a registry and a model tariff for loads of 50 megawatts or more. A final order is anticipated January 28, 2027, after a November 17 technical conference on cost allocation.14 The tentative order is a conditional lane in the SPP sense: connect sooner, accept curtailment first. Its curtailment trigger was written around PJM's proposed interim service, which FERC rejected nine days later, so the trigger will need a different definition before the final order. That is my inference from the two dates. By EEI's October 9 list, 25 states had approved at least one large-load tariff and eight more were pending.15
Colorado sits outside that federal frame. FERC's orders reach regional transmission organizations and independent system operators, and Xcel's Colorado utility is in neither. It has been approved for the SPP Markets+ day-ahead market, which the commission has said cannot substitute for the full RTO that state law directs Xcel to join or explain why it cannot, under a 2029 deadline the commission restored in January.16 With no federal overlay, the commission's retail terms will serve as the rulebook for a large part of the non-RTO West. Which utility's tariff applies to a given campus is already a screening question that can cross jurisdictions. The AIxEnergy Map carries large-load tariff applicability beside its queue and constraint layers, and the order Colorado writes will change that layer for the interior West.
What a commission can approve quickly
A commission can approve a contract in weeks when it incorporates published terms and lists its departures, and cannot when every term is open. Utilities that want 120-day reviews should publish eligibility criteria and standard terms, along with a format for the deviation sheet, before the first deal arrives, and file each sheet publicly even if the dollar terms stay confidential. Minnesota shows what happens otherwise: the Attorney General is asking the commission to require Xcel to justify each redaction in the Google filing.5
Developers and hyperscalers can move a deal toward the fast category by bringing what a commission would otherwise have to ask for. That means security sized to the assets being built, direct payment of local upgrade costs, a flexible block with a stated quantity and a sunset, and a term that starts at energization. A fixed number of months of credit support costs little to offer and does less as terms lengthen. Security tied to remaining rate base is harder to offer, and it would carry more weight with a staff that has asked for 150 percent of it.
Xcel says custom contracts will be consistent with its core terms, and the commission has not said how it will verify that. The scheduled hearings end today, the commission has set no decision date, and I could not review the full e-filed record, including Xcel's rebuttal. FERC's grid-operator filings begin arriving in mid-November, and Pennsylvania's cost allocation conference follows on November 17.
Notes
- Colorado Public Utilities Commission, “Electric Rate Cases,” Proceeding No. 26AL-0137E schedule, accessed October 28, 2026, https://puc.colorado.gov/electric-rate-cases. ↩
- Mark Jaffe, “Colorado Utility Regulators to Decide Whether Xcel’s Data Center Tariff Proposal Protects Customers,” Colorado Sun, October 21, 2026, https://coloradosun.com/2026/10/21/xcel-data-centers-energy-colorado-puc/. ↩
- Courtney Fieldman, Answer Testimony on behalf of Rewiring America and the Southwest Energy Efficiency Project, Hearing Exhibit 1101, Colorado PUC Proceeding No. 26AL-0137E, August 19, 2026, as posted by Earthjustice, Earthjustice copy. This is intervenor testimony; the figures it attributes to Xcel workpapers and tariff exhibits come through the witness and I have not checked them against Xcel’s filings. ↩
- Xcel Energy, “Xcel Energy to Power New Google Data Center in Minnesota,” news release, February 24, 2026, https://newsroom.xcelenergy.com/news/xcel-energy-to-power-new-google-data-center-in-minnesota. Terms are those announced by the parties and remain subject to Minnesota Public Utilities Commission review. ↩
- Minnesota Attorney General’s Office, press release on its comments regarding the Xcel and Google Pine Island agreement, October 1, 2026, https://www.ag.state.mn.us/Office/Communications/2026/10/01_XCEL-Energy.asp. The positions described are the Attorney General’s, not findings. ↩
- Xcel Energy, “Xcel Energy Proposal Protects Customers from Higher Bills as Data Center Demand Grows,” news release, April 2, 2026, https://newsroom.xcelenergy.com/news/xcel-energy-proposal-protects-customers-from-higher-bills-as-data-center-demand-grows. ↩
- “Colorado Regulators Hear Xcel’s Proposed Tariff for Large Load Customers,” Compute Law, October 23, 2026, https://computelaw.blog/brief/colorado-regulators-hear-xcel-s-proposed-tariff-for-large-load-customers/. A secondary summary; I did not confirm its figures against the docket. ↩
- “Dominion Energy Must Charge Data Centers for Transmission Costs, Virginia SCC Rules,” Utility Dive, August 12, 2026, https://www.utilitydive.com/news/dominion-energy-scc-transmission-costs-data-center/827693/. ↩
- Shaina Cole, “Xcel’s Large Load Tariff Fight: Who Will Pay for Colorado’s Data Centers?,” Rocky Mountain Voice, September 9, 2026, https://rockymountainvoice.com/2026/09/09/xcels-large-load-tariff-fight-who-will-pay-for-colorados-data-centers/. Party positions are as summarized by the outlet from filed testimony. ↩
- “FERC Approves SPP Conditional High Impact Large Load Service,” Utility Dive, June 8, 2026, https://www.utilitydive.com/news/ferc-spp-chills-large-load-transmission-service/822211/ (FERC Docket No. ER26-1323, approved June 5, 2026). ↩
- “Existing US Grid Can Handle ‘Significant’ New Flexible Load: Report,” Utility Dive, 2025, https://www.utilitydive.com/news/us-grid-headroom-flexible-load-data-center-ai-ev-duke-report/739767/, describing Nicholas Institute for Energy, Environment & Sustainability, Duke University, Rethinking Load Growth, https://www.powermag.com/wp-content/uploads/2025/02/rethinking-load-growth.pdf. The study’s figures exclude transmission constraints. ↩
- Federal Energy Regulatory Commission, “FERC Launches Aggressive, Targeted Action to Speed Large Load Integration,” news release, June 18, 2026, https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration; and “Fact Sheet: FERC Takes Action to Supercharge America’s Grid Efficiency & Reliability,” https://www.ferc.gov/news-events/news/fact-sheet-ferc-takes-action-supercharge-americas-grid-efficiency-reliability-and. ↩
- California Independent System Operator, “Regulatory Report, October 2026,” Docket No. EL26-71 summary, https://www.caiso.com/documents/regulatory-report-oct-2026.pdf. ↩
- Pennsylvania Public Utility Commission, “PUC Advances Consumer Safeguards as Data Center Growth Adds New Demands to Electric Grid,” news release, October 1, 2026 (Tentative Order, Docket No. M-2026-3064961), https://www.puc.pa.gov/press-release/2026/puc-advances-consumer-safeguards-as-data-center-growth-adds-new-demands-to-electric-grid-10-1-26. ↩
- Edison Electric Institute, “Large Load Tariffs and Projects,” updated October 9, 2026, https://www.eei.org/-/media/Project/EEI/Documents/Issues%20and%20Policy/List%20of%20Large%20Customer%20Projects%20and%20Tariffs.pdf. Counts are EEI’s. ↩
- Advanced Energy United, “Colorado PUC Denies Reconsideration of Xcel Entry into SPP Markets+, Signals Increased Scrutiny of Future Xcel Market Decisions,” news release, January 2026, https://advancedenergyunited.org/press-releases/colorado-puc-denies-reconsideration-of-xcel-entry-into-spp-markets-signals-increased-scrutiny-of-future-xcel-market-decisions/. ↩
