PJM's Data Center Ultimatum: Bring Your Own Power or Face First-Line Blackouts

PJM Interconnection filed a proposal with the Federal Energy Regulatory Commission to treat new data centers of 50 megawatts or larger as "non-capacity-backed load," meaning they face curtailment before household demand-response programs during grid emergencies^[1][3]. The filing follows two consecutive capacity auctions that fell short, with data centers responsible for roughly 5,100 megawatts of the December 2025 shortfall alone^[6]. On its surface, this looks like grid discipline: make the load that is driving the emergency pay for its own power. But the proposal's real function is to launder a decade of bad cost allocation into a customer choice problem.

Here is what the record does not say plainly: PJM's member utilities have already locked in billions in capital investment justified by load forecasts that included these same data centers. Those forecasts were models, not contracts. Utilities submitted them to state regulators as evidence for rate increases to fund generation and transmission upgrades that ratepayers financed upfront. Now, when the load arrives but refuses to pay for dedicated firm capacity, the utility keeps the asset (owned, regulated, earning its allowed return) while the ratepayer absorbs the risk that it sits half-used. PJM's "non-capacity-backed" label does not retroactively unbuild those plants or transmission lines. It redistributes the shortfall onto whoever shows up next.

The clearer scandal lies in what the PJM proposal does not require: transparency about which utilities have already signed or are negotiating special contracts with data-center operators, and on what terms. These contracts are routinely filed with state commissions under confidentiality seals, their rates and minimum-take clauses redacted. A utility can lock in a below-market price to a hyperscaler, pass the cost of new capacity to the general ratepayer base via the interconnection process, and then claim reliability justifies the investment. When the data center later chooses to site elsewhere or delay, the stranded cost stays on the system bill^[4]. PJM's registry of new large loads, while a step toward visibility, does not solve this: it documents what is coming but not what was already promised in sealed dockets.

The proposal's three transmission service options and behind-the-meter generation rules represent a FERC attempt to make data centers co-locate with power plants instead of becoming grid load, which could reduce interconnection queue delays stretching beyond eight years^[6]. That is sensible infrastructure planning. But it leaves intact the core mechanism: utilities keep the cost-allocation power. A data center that brings its own 500 MW of gas generation and connects it at a utility substation still triggers network upgrades, which the utility assigns to its general customer base unless a state tariff explicitly isolates the cost. Virginia, Ohio, and Oregon have filed large-load tariffs with teeth: multi-decade minimum-take commitments (85 percent of contracted transmission demand, 60 percent of generation demand) and collateral requirements that shift risk from ratepayers to the customer^[4]. Most states have not. PJM's filing does not require them to.

The window to act is now, while FERC reviews the proposal and states prepare to implement rules. Any utility seeking approval for new generation or transmission upgrades justified by data-center load should be required to file a detailed interconnection agreement showing the contracted demand, the minimum-take ratchet, the term matched to asset life, and the cost-allocation method. If the tariff is weak (ratchet below 70 percent, term shorter than 12 years, cost socialized), intervention dockets exist in every state's public utilities commission. The alternative is already clear: bring-your-own-generation plus a flexible, curtailable grid service that lets data centers energize faster while sparing ratepayers the risk of unused capacity. PJM has acknowledged this is technically feasible. The question is whether states will demand it.