PJM's capacity shortfall is widening, not stabilising — and the demand data may already be wrong

PJM's capacity shortfall is widening, not stabilising — and the demand data may already be wrong A third straight capacity auction miss at the price ceiling exposes a supply gap that analysts betting on reform may be underestimating. PJM Interconnection's 2028/2029 Base Residual Auction, results released on Wednesday (2026-07-15), came up 6,831 megawatts short of the grid operator's own reliability requirement — the third consecutive year the largest US power grid has failed to secure sufficient forward supply commitments. The clearing price hit the FERC-approved ceiling of $325 per megawatt-day for the second auction running.⁵ The equity desk view, led by Jefferies analysts on Wednesday (2026-07-15), is that structural reform of the Base Residual Auction toward a continuing operating cost model will bring capacity prices materially lower over time. That read is plausible as a regulatory outcome. It is much less clear that it addresses the physical supply problem now in front of PJM, which serves 67 million customers across 13 states and Washington, DC.^6,5 The shortfall is not stable. The 2027/2028 auction was already short approximately 6,500 megawatts; the 2028/2029 gap came in at 6,831 megawatts. A market focused on the 2.5% decline in the price ceiling — from $333.44 to $325 per megawatt-day — is reading a softer signal into an auction where the physical deficit actually widened. PJM is planning to move its backstop reliability auction forward to September, rather than waiting until 2027, which the grid operator itself acknowledged reflects pressure it cannot defer.^5,3 The first overlooked problem is the quality of the demand data underpinning PJM's own forecasts. Pennsylvania Governor Josh Shapiro signed a budget on Sunday (2026-07-12) requiring data centers in the state to report exact water and power usage annually, and directing PJM to give state regulators additional visibility into its demand forecasting. The bill's impetus was direct: state representative Nick Miller said the process by which utilities and load-serving entities submit information to PJM is "opaque," with policymakers and regulators lacking confidence in the data's reliability. One utility, according to Miller, submitted forecasts PJM found questionable. If the demand inputs into PJM's auctions are understated, the 6,831-megawatt shortfall is a floor, not a ceiling.⁴ The IEA has attributed roughly half of all US incremental electricity demand growth to data centers. Globally, data center electricity consumption rose 17% in 2025, with AI-focused consumption growing at 50%. Pennsylvania's disclosure law is an early acknowledgement that grid operators are flying partially blind on the load side of the ledger — precisely when that load is accelerating fastest.¹ The second signal the market appears to be discounting concerns supply delivery, not just supply procurement. Of all capacity that submitted interconnection requests between 2000 and 2019, only 13% had reached commercial operations by the end of 2024; 77% had been withdrawn. The interconnection queue is routinely cited as evidence of potential future capacity, but its historical attrition rate suggests the pipeline is a poor predictor of what actually gets built. Against a backdrop where Grid Strategies projects US data center demand growth of 65 to 90 gigawatts by 2029, the gap between paper capacity and delivered megawatts has direct implications for whether backstop auctions actually close the physical shortfall.^1,2 PJM Western Hub real-time power was trading at $72.38 per megawatt-hour on Wednesday (2026-07-15). The spot market is not yet reflecting the 2028/2029 auction miss in any dramatic way. But the capacity market's repeated ceiling prints are a forward signal that real-time prices will face structural upward pressure as 2028 approaches if new supply does not materialise faster than historical interconnection rates would predict. The reform thesis — that structural changes to the auction mechanism will bring capacity prices down — could prove correct as policy. What it does not resolve is whether sufficient physical capacity will exist to serve peak demand in the 2028/2029 delivery year. Jefferies may be right that the price formula changes; the supply shortfall could persist regardless.⁶ The September backstop auction is the first hard data point to watch. If it clears at or near the price cap again, or if the volumes procured leave meaningful gaps, the reform-driven bearish thesis on capacity prices will require more evidence. Pennsylvania's new demand-reporting regime, once data begins flowing, may also revise upward the load forecasts that PJM currently uses to set its reliability requirement — the number that has now been missed three years running.^3,4⁵