PJM was meant to open the offer window for its Reliability Backstop Procurement on 30 September. It did not. On 29 September the Federal Energy Regulatory Commission accepted the proposal, suspended it for five months to an effective date of 28 February 2027, made it subject to refund, set three issues for a paper hearing and opened a parallel section 206 proceeding in Docket EL26-108. PJM confirmed the next day that the procurement would not start as planned and that the timing is to be determined.
Most early coverage has treated this as a setback for reliability. Our view is different. The pause is an opportunity, and the most useful thing PJM can do with it is not only to fix the three defects FERC identified but to re-open the size of the purchase. The 6,831 MW target was set from an auction whose data centre forecast has already moved materially, and under the cost allocation FERC is steering PJM towards, every megawatt bought will be charged to someone's forecast. Getting the quantity right is now a consumer protection question as much as a reliability one.
How PJM got here
The backstop is a response to two capacity auctions that failed to clear the region's reliability requirement. The FERC order sets out the history. The 2027/2028 Base Residual Auction, held in December 2025, cleared 6,623 MW short of the requirement after accounting for Fixed Resource Requirement resources, the first time a BRA had cleared significantly short. The largest driver was 5,250 MW of forecast load growth, of which nearly 5,100 MW was attributed to data centres. That auction cleared at its temporary price cap of $333.44 per MW-day. The 2028/2029 auction, announced on 14 July 2026, again cleared short, at its temporary cap of $325 per MW-day, leaving the 6,831 MW gap the backstop is designed to fill.
The political push was explicit. In January 2026 the White House National Energy Dominance Council and the governors of all 13 PJM states issued a statement of principles urging a backstop auction offering 15-year price certainty for new capacity, to begin no later than September 2026, with costs allocated to load-serving entities with data centres that had not self-procured or agreed to be curtailable. The PJM Board then ran a fast-track stakeholder process. PJM's own design received only 30 per cent support in the Members Committee, against a competing subscription model sponsored by the Data Center Coalition and a group of electric distributors. The Board adopted its own design in late July with three changes: two limited opt-outs, a weighted-average offer cap of $555 per MW-day and tougher collateral rules for load-serving entities.
PJM's fact sheet describes the result as a one-time action. New resources only, commercial operation by 1 June 2032, commitments of up to 15 years, and an initial target equal to the 2028/2029 shortfall, reduced by signed bilateral contracts and other showings of new supply.
What FERC accepted, and what it did not
The Commission found parts of the design just and reasonable. According to Utility Dive's account of the order, those include the $555 per MW-day weighted-average cap and the supplier collateral rules. Three areas were set for hearing: cost allocation, transmission owner exit rules and load-serving entity collateral.
On cost allocation, FERC suggested allocating costs on updated load forecasts rather than the forecast that sat behind the July auction, and said PJM should use the most up-to-date load and resource forecasts to set the initial procurement target. On exits, it said costs should follow the load-serving entities in a zone if a transmission owner leaves PJM before the costs are recovered. On collateral, it found the proposed demand-side requirements did not strike a reasonable balance; Northern Virginia Electric Cooperative said it would have had to post roughly $2 billion. The Commission also rejected the opt-out for qualifying cooperatives and municipal utilities as discriminatory against other load-serving entities with data centres in their footprint.
The procedural clock is short if PJM wants it to be. Initial briefs are due 45 days from the order, which takes us to 13 November, with responses 20 days later. The order allows the hearing to be held in abeyance if PJM makes a new section 205 filing within 30 days, that is by 29 October, and it explicitly encourages PJM to do so. Bloomberg Law reported that FERC was strongly encouraging a new proposal by the end of February.
The concurrences show where the Commission's weight lies. Chairman Laura Swett, as quoted by Utility Dive, said the Commission would not be forced into accepting a deeply flawed, eleventh-hour mechanism with billion-dollar implications for consumers. Commissioner Lindsay See wrote that customers who drive new costs should bear appropriate responsibility for them and called for a serious focus on load forecasting. Commissioner David Rosner said the downside risk of inaction far outweighs the risk of an imperfect solution, while urging states to set large-load retail tariffs and curtailment priorities.
The number that should move
The strongest argument for re-sizing comes from the record itself. The Independent Market Monitor told FERC that the data centre forecast behind the 2028/2029 auction was built on load adjustments PJM received before 5 September 2025. Using project tracking from Industrial Info Resources, the Monitor counted 4,017 MW of data centre load cancelled or delayed as of 21 September 2026. That is 42 per cent of the 9,481 MW increase in data centre load between the 2026/2027 and 2028/2029 auctions, which rose from 11,993 MW to 21,474 MW. Dominion's zone accounts for 51 per cent of the cancellations and delays, and AEP Ohio for 29 per cent. On the Monitor's arithmetic, the 2028/2029 shortfall would fall from 6,831 MW to 3,055 MW.
The Monitor's figure is not beyond dispute. It says plainly that it does not have project-level data behind the distributors' load adjustments, and it counts only projects originally due in service by 2029. PJM's own January 2026 forecast had already cut its summer 2027 and 2028 peaks by 4,000 MW and 4,400 MW, citing tighter vetting of load adjustment requests. Forecasts are moving in one direction, and quickly.
That matters because of how the bill is meant to be paid. The fact sheet allocates costs first to zones in proportion to their share of the target, then to load-serving entities within each zone. FERC's preferred approach ties that allocation to updated forecasts. If the target stays at 6,831 MW while the forecast that justified it shrinks, the zones with the most cancelled projects will be asked to fund capacity for load that is no longer coming, under commitments that can run for 15 years, with accepted offers capped at a weighted average of $555 per MW-day. That is a weak position to defend before state commissions, and it is the kind of outcome Commissioner See's concurrence warns against.
What PJM should do with the five months
We would argue for four steps, in this order.
First, file the narrow fixes quickly. A section 205 filing by 29 October that adopts FERC's suggested approaches on exits and demand-side collateral, and removes the co-operative and municipal opt-out, would put the hearing into abeyance and keep a winter procurement possible.
Second, re-base the target on a refreshed load adjustment review rather than on the July auction result. FERC has already said the target should use the most current forecasts. A fresh review of the distributors' large load adjustments, tested against project status, is the natural vehicle, and doing it before the purchase avoids a refund fight afterwards.
Third, keep the offsets mechanism generous. Signed bilateral contracts for new supply, integrated resource plans and large-load demand-side arrangements all reduce the target. Bilateral matchmaking has been running since a June request for proposals, and every megawatt contracted directly between a data centre and a new plant is a megawatt that does not need a regional backstop or regional cost allocation.
Fourth, separate the 2029/2030 auction from the backstop. PJM had intended to announce backstop results before the December base auction. With the backstop now unlikely to run before then, the base auction should proceed on its own forecast, and the backstop should be sized net of whatever that auction clears.
What to watch
PJM is due to give an administrative update at the Market Implementation Committee on 7 October. The 29 October window for a new filing will show whether PJM chooses speed or a fuller redesign. The 2029/2030 auction in December will reveal whether the supply response to three years of capped prices is arriving without a backstop at all.
A backstop is meant to catch what the market misses. Bought at the wrong size, it becomes a 15-year commitment to the forecast of September 2025. FERC has given PJM time to avoid that. The quantity, not just the rulebook, should be on the table.