PJM's Capacity Auction Came Up 6.8 Gigawatts Short. Its Fix Could Cost Maryland Alone $562 Million.

On September 30th, PJM Interconnection — the grid operator for roughly 65 million people across 13 states and the District of Columbia — runs an auction it has never had to run before. Its regular capacity auction for the 2028/2029 delivery year, held earlier this year, came in 6.8 gigawatts short of the reserve margin the grid needs to stay reliable. PJM's fix is a one-time "Reliability Backstop Procurement": up to $20 billion authorized to spend, a price cap of $555 per megawatt-day — nearly 70 percent above the $325 cap that governed the auction that just came up short — and 15-year commitments for whatever new supply shows up, with a deadline to be online by June 2032.

Where the shortfall actually came from

PJM's own independent market monitor has been saying for over a year what the backstop auction now confirms with a dollar figure attached. Joseph Bowring, president of Monitoring Analytics, found that data centers were responsible for $6.3 billion — 38 percent — of the $16.4 billion in total charges from PJM's most recent base capacity auction. Zoom out to the last four base auctions and the number gets starker: $29.4 billion of $63.6 billion in total capacity charges, 46 percent, traced back to data center load growth. "PJM is continuing to act like it's business as usual," Bowring said of the grid operator's planning approach. "You have to open your eyes and recognize that it is really a paradigm shift, and failing to do that imposes costs on other customers."

The math behind the September 30th auction is what happens when that paradigm shift outruns the planning process built for the old one. Generation retirements, permitting timelines for new plants, and transmission buildout all move on multi-year clocks. Data center interconnection requests, in PJM's queue as in every other grid operator's, have been arriving faster than any of those clocks can turn. A regular annual auction assumes supply roughly keeps pace with demand. This year it didn't, by 6.8 gigawatts, and the backstop is PJM buying its way out of that gap at a price nearly 70 percent above what the market itself was willing to clear at just months earlier.

Who's actually supposed to pay for it

This is where the auction stops being a market-design story and becomes a cost-allocation fight, and five state ratepayer advocates got to FERC before the September 30th start date to say so. Consumer advocates from Maryland, Delaware, the District of Columbia, Illinois, and New Jersey filed a joint protest arguing PJM hasn't demonstrated the backstop procurement is even necessary, and that its design does nothing to protect existing customers if the data center load it's built around gets delayed, downsized, or never shows up. Maryland's Office of People's Counsel ran its own numbers: if PJM procures its full target at the maximum $555 price, Maryland ratepayers alone could be on the hook for as much as $562 million over the life of the 15-year contracts — for two utility zones initially assigned a combined 135.4 megawatts of the total procurement.

Maryland's advocate isn't just objecting, it's offering an out. OPC argues that utilities whose data center customers participate in demand-response programs — cutting electricity use during genuine peak-demand hours rather than running flat out year-round — should be able to opt their zone out of the backstop procurement and the costs that come with it. Utilities that can report that kind of demand flexibility to PJM by October 21st, OPC says, could reduce or eliminate the share of the $555-per-megawatt-day bill assigned to their existing customers. It's a bet that the "must-run at full nameplate load, all the time" assumption baked into a lot of data center interconnection requests is more conservative than the actual operating pattern — and that testing that assumption is cheaper than a $20 billion emergency auction.

The diligence angle

We've written before about the federal Ratepayer Protection Act and Texas's new interconnection deposit rule as two different attempts to answer the same question: who pays when a data center's interconnection triggers grid costs bigger than the project itself. PJM's backstop auction is a third version of that fight, running at a different altitude. The federal bill and the Texas rule both operate at the interconnection-agreement level — a specific project, a specific utility, a specific deposit. PJM's shortfall is systemic: the grid operator is short 6.8 gigawatts of capacity it needs regardless of which specific data center projects end up real, and the $20 billion backstop gets allocated across every load-serving entity in the footprint, data center or not, unless a state successfully carves its ratepayers out the way Maryland is trying to.

For a developer evaluating a site inside PJM's footprint, that's a different kind of commercial-readiness question than the ones we usually flag in a Kill Test. It's not "can this sponsor produce the interconnection deposit" — it's whether the utility a site would connect through is one PJM assigned backstop procurement obligations to, and whether that cost exposure moves if the FERC protest succeeds or the October 21st opt-out reporting changes which zones are on the hook. That's a live number through at least that date, and probably well past FERC's eventual ruling on the joint protest. GridPermit doesn't work inside PJM's footprint today, but the shape of the fight is the one worth tracking everywhere large loads are interconnecting: the closer a grid operator gets to its own reserve margin under stress, the more the cost of fixing that stress becomes a diligence question, not just a market one.