PJM's Backstop Auction Addresses the Known Gap While the New-Load Problem Remains Unsolved
The board's curtailment plan and bilateral matchmaking buy time, but interconnection data and surging AI demand suggest the capacity deficit extends well beyond 2032.
PJM Interconnection's board on Monday (2026-07-28) proposed a backstop capacity auction and a new curtailment framework for data centers, its latest move to address a reliability shortfall that has grown for three consecutive years.7
The number behind the urgency is 6.8 gigawatts. That is how short PJM's most recent capacity auction fell against the grid's reliability requirements, the third straight year the market failed to secure enough forward supply commitments. PJM delivers power to 67 million people across 13 states and Washington D.C. Persistent deficits at that scale are a systemic problem.5
The board's plan sets June 1, 2032 as the eligibility deadline for resources entering the backstop auction. PJM has also engaged Charles River Associates to oversee a bilateral matchmaking process between data centers and generation developers, launched through a request for proposals issued on June 9 (2026). Stakeholders approved the two-part framework on Tuesday (2026-06-30).7,3
Markets have taken these steps as evidence of progress. PJM Western Hub real-time power cleared $62.49 per megawatt-hour in early Friday (2026-07-31) trading. The consensus on PJM real-time has tilted bearish, with signals running roughly 4-to-1 against the bulls.
But critics cited in the board's own proposal warned that a backstop auction addressing only the deficit from a prior base capacity auction fails to "solve the crux of PJM's issue — new large loads which have yet to materialize," according to documents reviewed by Utility Dive. The auction fills the known hole. The loads expected to arrive after 2032 sit well outside the scope of what is being proposed now.7
That caveat sharpens considerably against what PJM's interconnection queue actually delivers. Of the capacity that submitted interconnection requests from 2000 to 2019, only 13% had reached commercial operations by the end of 2024, the International Energy Agency reported. Seventy-seven percent had been withdrawn. The queue presents a large supply pipeline on paper; in practice, it behaves more like a waiting room that most applicants eventually leave.1
Demand is accelerating faster than the grid's procurement tools were designed to handle. Data centers now account for roughly half of all incremental electricity demand growth in the United States, the IEA found; global electricity consumption from data centers rose 17% in 2025, while AI-specific consumption surged 50% in the same year. PJM felt the pressure directly on Thursday (2026-07-02), when load climbed to roughly 163 gigawatts during a heat dome that pushed heat indices past 110 degrees from Washington to New York, stopping just short of the 2006 all-time peak of 165,563 megawatts, even as PJM had forecast load could exceed 166,000 MW that day.4,1
The curtailment plans for data centers give the grid a demand-side tool it has not previously held over large industrial customers. But curtailment does not add generating capacity, and it depends on data center operators honoring interruptibility commitments that have not been stress-tested anywhere near the scale now projected.7
The bilateral channel overseen by Charles River Associates may matter more than the auction itself in the near term. Canary Media reported in July (2026-07-17) that direct contracts between individual data centers and generation or storage developers represent a potentially larger procurement mechanism than the centralized process, one that moves outside PJM's regulatory timelines. Still, bilateral deals carry counterparty risk, lack the firm-capacity commitment structure of a capacity auction, and PJM has no established track record running a bilateral program of this scope.6,7
PJM moved the backstop auction up from 2027 to September of this year (2026), E&E News reported in May (2026-05-21), a sign the operator itself did not believe waiting was viable. The July board proposal confirms the direction. If September's auction attracts committed capacity at scale and the interconnection attrition rate falls as data centers bring contracted generation, the bearish case on PJM capacity pricing becomes harder to sustain. If historical withdrawal rates hold and the new large loads arrive faster than the 2032 window anticipates, the 6.8-gigawatt shortfall from this past auction will look modest against what follows, and the current bearish consensus will look inadequately positioned.2,57