Data Centers Account for 40% of Costs in PJM Capacity Auction

PJM Interconnection's third consecutive capacity auction closed at $16.4 billion for the 2028/29 delivery year, with data centers representing 40% of total costs, according to Monitoring Analytics.

PJM operates the largest electric grid in the United States, serving 67 million customers across 13 Mid-Atlantic and Midwest states. The auction secured 138.3 GW of generation but fell 6.8 GW short of its reliability target, making it the clearest signal yet of how data center demand for artificial intelligence is reshaping North American electricity markets. Only 525 MW of new capacity entered the auction, compared to the 20 GW added across the five previous rounds, Canary Media reported. In total, projected data center demand has added more than $29 billion in additional capacity costs since 2024, according to Monitoring Analytics.

Facing mounting pressure, PJM is preparing two reforms. The first is the Reliability Backstop Procurement (RBP), an extraordinary auction approved in June 2026 that will allow 15-year contracts between developers and data centers, with a price cap of $555 per MW-day and costs assigned to large users. PJM plans to file the RBP with the Federal Energy Regulatory Commission (FERC) this month, with a first auction in September. The second is the Connect and Manage framework, which seeks to require data centers to assume reliability risk when interconnecting without new generation available. However, none of the 11 Connect and Manage proposals reached the required majority in the market participant vote.

PJM's capacity market redesign is the reference case for other grid operators across North America facing similar data center load pressures. FERC's initial ruling on the RBP and the September auction will determine whether the mechanism successfully shifts costs to those who originate them without fragmenting the market.

This article was written with the assistance of artificial intelligence from verified sources and reviewed by a human editor before publication.