On Thursday, October 29, PJM filed its amended Reliability Backstop Procurement with the Federal Energy Regulatory Commission, proposing a bid window opening March 1, 2027, to buy the 6,831 MW its July capacity auction failed to procure. Days later, the arithmetic underneath that filing moved again. In an October 28 note reported by Utility Dive on November 3, Jefferies analysts estimated that the procurement could shrink to roughly 630 MW, after utilities including PPL Electric, FirstEnergy, and AES Ohio lowered their near-term large-load forecasts by 6.5 GW. The backstop was designed as a safety net for a forecast-driven shortfall. The forecast itself is now moving underneath it, and the net is shrinking faster than the case for running it.
Start with the scale of the shrink, because it is easy to understate. A 630 MW procurement is about 9 percent of the 6,831 MW shortfall the July auction measured. At the $555/MW-day maximum the refiled proposal retains, buying the full 6,831 MW would cost roughly $1.38 billion a year, or about $20.7 billion over 15-year terms. Buying 630 MW at the same ceiling would cost about $128 million a year, or about $1.9 billion over 15 years. Those are maximum-price figures, not expected costs — the clearing price is unknown until bids are taken — but the ratio is the point: more than nine-tenths of the procurement’s expected cost has evaporated before a single bid has been solicited, because the load forecast that justified it has been revised down.
What would 630 MW actually buy? Put it against the numbers this market has already produced. The July auction, clearing at its $325/MW-day cap, summoned about 525 MW of new and upgraded supply — roughly the same order of magnitude as the shrunken backstop. The December auction for the 2029/30 delivery year, which closes December 15, would need procured capacity to rise roughly 9,530 MW above the 2028/29 total to meet its reliability requirement. And Charles River Associates’ independent forecast puts the gap between the data-center load that wants to arrive and what the system can power at 1-in-10 reliability at 11.5 GW by 2030. A 630 MW backstop is about one large power plant’s worth of capacity, bought through a one-time emergency procurement, against gaps measured in the thousands of megawatts on every horizon that matters.
None of that makes 630 MW the number. It is an analyst estimate, not PJM’s, and PJM has been careful to say the final target will be recalculated. Jeffrey Shields, a PJM spokesman, told Utility Dive that if the final large-load forecast for 2028/29 comes in lower than this year’s long-term load forecast, it could contribute to a reduction in the backstop target — but that other factors will also matter, including updated accreditation numbers for resources that cleared the July auction and megawatts placed through bilateral contracts. The January release of PJM’s 2027 long-term load forecast is the next hard input, and the final target is not set until April 28. What is already fixed is the direction of travel: every revision since July has pointed down.
What the refile actually proposes
The amended filing keeps the backstop’s core intact and changes the three things FERC objected to. The schedule is now definite, if still only proposed: bidding opens March 1, 2027, closes March 28, the final target and supply amounts are determined April 28, and results are posted May 10, for capacity serving a delivery year beginning June 1, 2028. PJM has asked FERC for an order by January 29, 2027. Nothing in the filing is approved yet, and the original five-month suspension runs to February 28, 2027 — the procurement cannot run until FERC acts on the amended terms.
The commercial terms are unchanged from the September order: commitments of up to 15 years, at a megawatt-weighted average maximum of $555/MW-day — the figure FERC accepted in September, and the figure PJM’s own uncapped simulation produced for the July auction, against the $325 the collar actually allowed. The target will be recalculated with updated information before bidding, and reduced by megawatts procured through executed bilateral contracts, so private deals shrink the public purchase rather than duplicating it.
What changed is the machinery around the purchase. Cost allocation is now based on large-load growth, measured from the actual 2026 summer peak to the forecast for the 2028/29 delivery year, so zones pay in proportion to the firm large load they are adding rather than the load they already serve. The collateral methodology for load-serving entities has been revised. And the filing recognizes that a backstop obligation and its associated collateral can move with the load-serving obligation when a large customer changes suppliers — the transferability question that had left providers of last resort exposed to collateral for load they might not keep.
Maria Scheller, vice president of energy and planning at the consulting firm ICF International, told Utility Dive the revisions sharply reduce risks for providers of last resort and default-service providers, the utilities that supply customers who do not buy from power marketers. But she added that the proposal reduces the direct risk to residential and other existing customers without eliminating it: administrative costs, collateral and financing costs, regulatory lag, customer default, and stranded-cost risk can still create indirect exposure if state mechanisms do not keep the obligation tightly attached to the large load that caused it. Some state regulators, she said, may need to craft new rules on customer-specific allocation, cost recovery, and what happens if the underlying load shrinks, defaults, or disappears. The federal filing hands the residual risk to the states — and the states have not yet written the rules that would contain it.
Why FERC only partly approved it
The September 29 order, 196 FERC ¶ 61,245, is best read as a split decision, and the split is what the refile is answering. FERC accepted the core of the procurement — including the $555 ceiling — but suspended it for five months and opened a paper hearing on three issues: cost allocation, transmission-owner exit rules, and collateral requirements for load-serving entities. The Commission’s concern, in substance, was that the original allocation method would have spread the cost of new capacity driven by large-load growth across existing customers who did not cause it.
The U.S. Department of Energy agreed, in an October 7 filing supporting FERC — its first FERC filing in at least five years — arguing that large electric loads, not households or other business ratepayers, should fund the generation and infrastructure the backstop is designed to secure. The refile’s growth-based allocation is the direct answer to that objection, and it is the provision FERC will weigh most heavily when it decides whether to hold the paper hearing in abeyance. If the Commission accepts the new formula, the backstop proceeds on the March schedule; if not, the hearing continues and the March dates slip — or the procurement does not happen at all.
The design question the shrink raises
Here is the question the 630 MW estimate forces, and it is not whether the backstop should be smaller. It is whether a procurement of that size justifies the machinery built to run it. The backstop is a one-time emergency purchase with 15-year contracts, a bespoke cost-allocation formula, a paper hearing, a federal filing, and a state-level implementation burden that Scheller says may require new rules in multiple jurisdictions. At 6.8 GW, that machinery was proportionate to the problem: a shortfall large enough to threaten reliability, bought once, at a price the regular market could not reach. At 630 MW, it is a great deal of market design for roughly one plant’s worth of capacity — and the fixed costs of the design, the litigation, and the allocation fights do not shrink with the target.
The allocation fight shrinks but does not disappear. Under the growth-based formula, zones whose forecasts fell pay less, which is the design working as intended — but the measurement is now visibly contested, with near-term forecasts down 6.5 GW by the utilities’ own filings and CRA’s independent forecast reaching 82 GW by 2047. Zones will argue about measurement, not method, and the argument will be fought over a procurement one-tenth the size it was designed to be. That is not a reason to abandon the formula — but it is a reason to ask whether a one-time procurement is the right vehicle for a number this small.
The counterargument deserves its full weight. The backstop is insurance, and insurance is judged by the loss it prevents, not the premium it costs. CRA’s hard-constraint case is a genuine reliability warning: if demand follows the soft trajectory and supply follows the status quo, 11.5 GW of load cannot be served at 1-in-10 reliability by 2030, and the cost of being short — load shedding on a hot day in a system serving 67 million people — is worse than the cost of being long. A small backstop still buys something the regular auction cannot: 15-year terms at uncapped prices, for resources that will not commit on a one-year contract at $337.83. But insurance priced for a 6.8 GW fire and bought for a 630 MW one is a different purchase, and it should be underwritten as one.
The forecast moved. The auction didn’t — yet.
Why the near-term forecast fell is now well documented, and the mechanism matters more than the number. PJM’s Independent Market Monitor calculated that 4,017 MW — 42 percent — of the forecast increase in data-center load between the 2026/27 and 2028/29 delivery years had been canceled or delayed as of September 21. The utilities’ lowered forecasts that Jefferies cites are, in part, the same phenomenon reaching the official forecast through Load Analysis Subcommittee filings rather than through the monitor’s audit. The long-run pipeline can keep growing — CRA’s 30 GW by 2030 stands — while the near-term forecast falls, because that is where cancellations and delays land first.
That asymmetry is now driving a structural challenge to the auction itself. On November 3, Exelon filed a complaint with FERC, in docket EL27-11, arguing that PJM’s three-year forward Base Residual Auction is no longer just and reasonable and asking the Commission to replace it with a prompt auction held two months before the delivery year, starting with the 2030/31 delivery year — an auction currently scheduled for May 2027 that would move to approximately March 2030. Exelon asked FERC to decide by January 31, 2027, the day before PJM posts planning parameters for the 2030/31 auction.
Exelon’s argument is, in substance, the backstop’s own evidence turned against the auction that created it. The complaint says the risk-balancing analysis the Commission relied on in 2006 to accept the load-forecast inaccuracies inherent in a multi-year forward period has fundamentally collapsed in 2026, and that capacity resources are now functionally incapable of being constructed within the current three-year cycle — turbine backlogs measured in years mean the auction buys existing capacity, not new entry, so buying three years early purchases forecast error without purchasing supply. The complaint includes a figure by Emma Nicholson of Charles River Associates correlating shorter forward periods with smaller changes in peak-load forecasts between the base auction and the third incremental auction, held about four months before the delivery year.
PJM has opposed fast-track processing and asked for until December 21 to answer on the merits. Its Board is scheduled to consider the question at its November 17 meeting, after a November 13 workshop on more than 30 responses to PJM’s request for papers. The complaint explicitly does not apply to the December 2029/30 auction or to the backstop — both proceed on their current schedules whatever FERC decides. But the filing reframes the backstop’s shrink as a symptom rather than an accident: the three-year forward auction is the instrument that converts a forecast into a purchase, and if the forecast is wrong by gigawatts within a single year, the instrument itself is the problem.
The calendar, in order
December 15: the 2029/30 auction closes, capped at $337.83/MW-day. January: PJM releases its 2027 long-term load forecast, the first hard input to the recalculated target. January 29: the date PJM has asked FERC to rule on the amended backstop proposal. January 31: the date Exelon has asked FERC to decide its complaint. February 28: the original five-month suspension expires. March 1 to 28: the backstop bid window, if FERC has approved the refile. April 28: the final target. May 10: the results.
What to watch, in that order: whether FERC holds the paper hearing in abeyance or lets it run past the January dates; how much new supply December summons against last time’s roughly 525 MW; what the January load forecast does to the 6.5 GW reduction — whether it confirms it, deepens it, or reverses it; and what number PJM posts on April 28 as the backstop’s final target. The gap between 630 MW and 6,831 MW is not a rounding error. It is the distance between two forecasts of the same system, and whichever number April produces will tell us which forecast PJM believed when it mattered.