Hero image for "PJM Hit Its Price Cap Two Years Running. The Real Fight Is Who Pays the $16.4 Billion Tab."

PJM Hit Its Price Cap Two Years Running. The Real Fight Is Who Pays the $16.4 Billion Tab.


Here's the number that should stop every grid planner in their tracks: PJM's capacity price went from $28.92 per megawatt-day for the 2024/25 delivery year to a hard cap of $325.00 for 2028/29, and even at that ceiling, the auction still came up 6,831 megawatts short of what the grid operator says it needs for reliability (Energy News Beat). The total capacity bill for the system jumped from $2.2 billion to $16.4 billion a year over that same stretch. That's not a market finding equilibrium. That's a market telling you, in the bluntest language it has, that it ran out of supply before it ran out of demand.

I've been tracking this capacity crunch since September, when I wrote about PJM's auction crisis costing ratepayers $29.4 billion and again in October when FERC punted the fix for five months. What's changed since then isn't the scarcity — that's structural now — it's the fight over who absorbs it. And that fight just produced two consequential rulings and a counter-filing, all inside the last five weeks.

FERC Said No to Curtailing New Data Centers First

On October 9, FERC ruled on PJM's large-load package, the set of proposals meant to govern how new data centers and other massive loads get connected and billed. The Commission accepted two pieces — a 50-megawatt definition of "Large Load" and a registry requiring every site at or above that threshold to disclose its ramp schedule, backup generation, and supply contracts — effective October 12 (PSV). But it rejected the more consequential pieces: the Interim Resource Adequacy Service that would have curtailed new large loads ahead of everyone else, the requirement that new large loads bring their own capacity, and the carve-out exempting them from the capacity demand curve starting with the 2029/2030 delivery year.

The Commission's reasoning matters as much as the outcome. FERC found PJM's curtailment-first framework reached into territory reserved for state regulators over retail service, and PJM hadn't demonstrated the plan was nondiscriminatory. In plain terms: you can't just decide the newest, biggest customers get cut off first without showing your work on why that's fair relative to everyone else drawing power from the same wires. PJM got an invitation to refile, not a blessing.

What survived — the registry — is visibility, not cost allocation. Every site at 50 megawatts or more now has to tell PJM what it actually plans to consume and when. That's valuable information for grid planning. It is not, by itself, a pricing mechanism, and conflating the two would be a mistake.

The Bigger Fight: Who Funds 6.8 Gigawatts of New Backstop Power

The curtailment ruling is a subplot. The main event is PJM's Reliability Backstop Procurement, the mechanism meant to plug that 6,831-megawatt hole left by the last Base Residual Auction. FERC accepted PJM's original RBP proposal on September 29 but suspended it for five months and ordered a paper hearing on three specific questions: cost allocation, collateral requirements, and what happens to load-serving entities that try to exit the market rather than pay in (Energy News Beat). FERC's own language in that order was unusually direct for a regulatory body: PJM's cost allocation "may be unjust and unreasonable."

The Department of Energy weighed in too, filing a statement at FERC arguing that large loads — not households — should fund the capacity their own growth is forcing onto the system. Utility Dive reported this appears to be DOE's first filing of this kind at FERC in at least five years, according to the same reporting (Energy News Beat). That's a meaningful signal from a federal agency that otherwise spends its public energy pushing to accelerate AI infrastructure buildout — DOE's own data center resource hub frames "winning the AI race" as a matter of national priority and promises to shield families from the cost of getting there through a "Ratepayer Protection Pledge" that requires tech companies to help fund the infrastructure they need (Department of Energy). Those two postures — build faster, bill the builders — aren't contradictory. They're the same instinct applied to two different levers.

PJM responded on October 29 with an amended RBP filing that restructures exactly the three things FERC flagged: cost allocation, credit requirements, and exit obligations for load-serving entities (Interface HD). Under the revised structure, procurement charges would be allocated based on each zone's large-load growth between the actual 2026 summer peak and the forecast 2028/2029 delivery year — a formula that tries to make the customers driving the shortfall pay in closer proportion to how much shortfall they're driving. That's a proposal now sitting in front of FERC again, not an adopted rule. The distinction matters: a filed cost-allocation formula is not a bill anyone has actually received yet.

What 14.4% Versus a 20% Target Actually Tells You

It's worth sitting with how PJM describes its own shortfall, because the number reveals the mechanics better than any headline price. Even clearing at the $325 cap, the 2028/29 auction bought 149,181.6 megawatts against a target that would have required roughly 156,000. The resulting reserve margin landed at 14.4%, against a 20% target (Energy News Beat). A reserve margin isn't a measure of how much power is flowing; it's a cushion against the worst days — the heat waves and cold snaps when everything that can run needs to run. PJM's own language, cited by FERC, is that the market "has failed to keep pace with the unprecedented influx of large load." That's a capacity planning target missed by a wide margin, not a grid actively failing today. Conflating the two — treating a thin reserve margin as proof of imminent blackouts — overstates what the data actually shows. What it reliably shows is that the price signal needed to attract new supply has arrived two auction cycles late, and the backstop procurement is PJM's attempt to buy time while that supply gets built.

And new supply is, in fact, getting built — or at least queued. GE Vernova reported a 122-gigawatt backlog of gas-power equipment orders and reservation slots, alongside a $9 billion acquisition to expand its position in high-voltage electrical equipment (Interface HD). That's a backlog, not generation on the grid — a point I keep hammering because the gap between an order book and a commissioned plant is measured in years, not quarters, and I've written before about how that equipment queue is becoming the real constraint on the whole buildout. OpenAI, for its part, announced a national flexibility program for its computing campuses, building on a Georgia contract allowing up to 1,000 megawatts of demand reduction (Interface HD). An announced demand-response program is a hedge against scarcity pricing, not a substitute for the gigawatts PJM is still short.

This Is What Energy Abundance Actually Costs to Build

Here's where I'll push back on the framing that scarcity pricing is purely a crisis to be managed away. A capacity price that goes from $29 to $325 per megawatt-day is the market screaming for more steel in the ground — more gas turbines, more nuclear uprates, more transmission. That's the price signal doing exactly what it's supposed to do: making new generation investable again after a decade of capacity markets that paid too little to justify building anything. The fight over cost allocation is legitimate and worth getting right — households shouldn't subsidize a hyperscaler's training run — but don't mistake the fight over who writes the check for evidence that we're building too much. We are, if anything, finding out in real time how far behind the curve the last decade of underinvestment left us. The UK offers a parallel worth watching: Drax, Brookfield, and Radiant have proposed a gigawatt-scale data center development that Brookfield estimates could eventually draw more than £20 billion in investment (Interface HD) — another signal that capital is racing toward the exact bottleneck PJM is fighting over, not away from it.

The next marker is FERC's ruling on PJM's amended RBP filing, plus the 30-day compliance filing PJM owes the Commission on the large-load registry details from the October 9 order. Watch whether FERC accepts the zone-by-zone, growth-weighted cost allocation or sends PJM back again — a third rejection would be the clearest sign yet that nobody in the room has found a formula that satisfies both state regulators and federal reliability math. Either way, the price cap isn't the ceiling on this story. It's the floor under a much bigger argument about who finances the grid the next decade actually needs.