Data centres drove a tenfold spike in PJM capacity prices — and passed $9.3 billion to consumers

#energy #markets #finance

📌 Capacity prices in PJM rose from $28.92 to $329.17 per MW-day in two years. Data centres were responsible for 63% of the increase in the most recent auction — a redistribution of energy cost that is only beginning.

Power Markets · April 7, 2026

Capacity markets exist to ensure there is enough generation available to meet peak demand — and to price the cost of that adequacy into the electricity system. What they are not designed to absorb is a single new demand category growing fast enough to reprice the entire market. That is what AI data centres have done to PJM Interconnection, the largest grid operator in the United States.

Capacity prices in PJM rose from $28.92 per megawatt-day in the 2024/25 delivery year to $329.17/MW-day in 2026/27 — roughly a tenfold increase in two years. The independent market monitor for PJM estimated that data centres were responsible for 63% of the price increase in the 2025/2026 auction, translating to $9.3 billion in costs to be recovered from customers across the PJM region through higher electric rates. In the December 2026 capacity auction, data centre load accounted for $6.5 billion, or 40%, of the total $16.4 billion in auction costs.

The transmission to retail bills is direct. Washington D.C. Pepco residential customers saw average monthly bills rise by $21 starting in June 2025, with roughly half of that increase attributable to the capacity price spike. PJM projects peak demand will grow by 32 gigawatts from 2024 to 2030 — with all but 2 gigawatts of that growth coming from data centres. The grid infrastructure that serves households and small businesses is being repriced by AI infrastructure investment at a rate that was not anticipated in any utility planning model.

The market structure question this raises is acute. Capacity markets socialise the cost of grid adequacy across all ratepayers. When the incremental demand is almost entirely from large industrial users — data centres with dedicated substations and contracted capacity — the mechanism by which costs are distributed may need to change. Virginia has already created a new data centre electricity rate class in response. The policy debate about whether AI infrastructure should bear a larger share of capacity costs directly is accelerating.

For energy market participants, the tenfold capacity price move in two years is a structural signal, not a cyclical one. PJM's demand growth forecast through 2030 is essentially a data centre forecast. The investors, utilities, and power generators that have positioned for sustained elevated capacity prices — and for the new generation assets needed to serve incremental AI load — are sitting on a structurally different set of forward exposures than those who treat this as a temporary distortion.

📊 Model View

Capacity price = f(demand growth, generation adequacy margin, transmission constraints). When demand growth is concentrated in a single category growing at 30%+ annually, the capacity price function is highly convex. The PJM spike is an empirical demonstration of that convexity.

⬛ Bottom Line

AI data centres have already repriced the U.S. electricity market — and the policy, infrastructure, and investment consequences are still propagating.