PJM's board has given its data-center power plan operative shape: a mandatory registry of loads at or above 50 megawatts, curtailment of unsupplied new large loads ahead of Pre-Emergency Load Management from June 1, 2027, and a one-time Reliability Backstop Procurement in September–October targeting 6,831 MW at a $555 per MW-day cap. [1] [2]
The July 27 decisional letter, reported in detail July 28 and 29, converts what was the July 28 account of PJM proposing an emergency power auction for data-center demand from a planned filing into a defined framework. It also records what the board rejected: the stakeholder-preferred voluntary alternative. [1]
A framework is not a tariff
The ceilings matter more than the numbers. FERC has approved nothing. No tariff filing existed as of July 29. Cost allocation to data centers requires each of PJM's 13 states to act. [3] A board framework is a proposal; it does not become an approved tariff, a procured megawatt, or a payer until filings, decisions, and delivered capacity exist. None does.
The rejection embedded in the letter deserves its own line. The board considered a voluntary alternative — the structure stakeholders preferred — and chose the mandatory registry and curtailment hierarchy instead. [1] That choice is the strongest signal in the framework: the grid operator does not believe voluntary self-supply commitments from hyperscalers will arrive fast enough, and it said so by picking compulsion over cooperation. It is still a signal, not a rule, but it tells the reader which way the board expects the fight to go.
The frames divide on what the framework means. The trade press — Utility Dive, DCD, E&E — reads a technical reliability fix: a registry, a hierarchy, a backstop auction aimed at closing the data-center power gap through a two-plan structure. [1] [2] Power-markets and data-center-skeptic communities on X read something starker: the first U.S. grid formally telling hyperscalers to bring their own power or get cut first — a two-tier reliability structure. No citable status was recovered for that frame, so it stands unobserved as text, not silent.
The dependency chain
Each frame underplays what the other carries. The technical-fix frame underplays the two-tier structure: under the hierarchy, unsupplied new data-center loads are curtailed before the general emergency measures that touch everyone else. [1] The skeptical frame underplays the dependency chain: nothing binds anyone until FERC approves and thirteen statehouses move on cost allocation. [3]
The 13-state dependency is the quietest and most binding fact in the file. PJM spans thirteen jurisdictions; the framework's cost allocation — the question of who actually pays for the backstop and the buildout — runs through each one. [3] A framework can survive FERC and still die in fragments if some states act and others refuse, leaving the registry universal and the payer question balkanized. The two-tier grid, if it arrives, will arrive one statehouse at a time.
The backstop auction is the framework's moving part and its first real test. Six thousand eight hundred thirty-one megawatts at a $555 per MW-day cap is a target, not a result; the September–October procurement will reveal whether capacity can actually be bought at or under that ceiling, or whether the cap clears nothing and the gap stays open. [1] A failed auction would not kill the registry or the curtailment queue — it would expose them as instruments pointed at a supply that does not yet exist.
The receipts that would arbitrate the frames are calendar items. What tariff text PJM eventually files; whether FERC approves, modifies, or rejects the registry and curtailment hierarchy; what the September–October backstop bids show about clearing prices against the $555 cap. [1] Until then, the honest description is the board's own: a registry and a queue, defined on paper, waiting on Washington and the states.
There is also a jurisdiction discipline to keep. This is a PJM record — the mid-Atlantic grid — and it does not generalize to SPP, ERCOT, or any other operator. The board letter's mechanics are PJM's answer to PJM's interconnection queue; other grids watching the same data-center load growth will write their own letters, and none of them binds anyone yet either. [3] The story's national significance is real but derivative: PJM is the largest grid to put the two-tier idea in writing, which makes it the template everyone else will be measured against — if it survives FERC.
-- DARA OSEI, London