PJM's capacity backstop exposes a supply gap that keeps widening as data center load outruns every forecast
PJM's one-time backstop auction addresses a 6.8-GW mid-2028 shortfall, but the supply-response machinery has not accelerated to match a demand curve that has beaten forecasts for three consecutive years.
PJM's board released a proposal on Monday (2026-07-27) to hold a one-time backstop capacity auction beginning in September, covering a 6.8-GW shortfall in the delivery year starting mid-2028. Embedded in the same proposal: curtailment rights for new data centers, an instrument PJM has not historically needed. Both items together signal that the grid operator does not expect the supply pipeline to close the gap on its own.7
The bearish consensus on PJM capacity rests on project attrition — many data center pipeline entries will slip or be cancelled, keeping realized demand well below headline project totals. That view is not without merit. But the direction of prior forecast error points the other way. BloombergNEF analysts said on Wednesday (2026-07-22) that almost all U.S. regions ended 2025 with more data center capacity than BNEF had anticipated, with Texas posting the largest gap between forecast and actual build.6
BNEF Senior Associate Nathalie Limandibhratha put the scale of that acceleration bluntly on Wednesday (2026-07-22): "We track data centers across the U.S. and in the last year, we've added about 100 GW of project capacity. It's not only the number of data centers and new developers that are flooding the market, but it's also the size of these data centers." BofA analysts separately estimated data center load additions at roughly 125 GW over the same period. BNEF described its own base case as sitting in the middle of third-party estimates, not at the conservative end.6
The two BNEF scenarios for U.S. data center electricity demand by 2030 now diverge by 42 GW — more than four times New York City's peak load — with the higher scenario reaching 207 GW by 2033. Markets have concentrated on that uncertainty. What the attrition argument requires is a reversal in a trend that has run consistently in one direction for several years.6
How much pressure that demand is already placing on the grid became visible on Thursday (2026-07-02), when a heat dome pushed heat indices past 110 degrees across the mid-Atlantic and PJM electricity demand surged to roughly 163 GW. That fell just short of the grid's 2006 all-time peak of 165,563 MW, even though PJM had forecast that day's load could top 166,000 MW. Data centers were identified as a contributing factor in the demand event.3
The supply side has not kept pace. Just under 4 GW of newly built and uprated capacity has been included in PJM capacity auctions since 2024 — roughly a fifth of the 20 GW added across the five prior auctions.5 The backstop auction starting in September addresses the mid-2028 delivery year. It does not close the near-term distance between demand growth and available supply.7
Interconnection data makes the longer-term constraint concrete. Across U.S. grids, only 13% of capacity that submitted interconnection requests between 2000 and 2019 had reached commercial operations by the end of 2024, with 77% of requests withdrawn, according to IEA analysis. Permitting, supply chains, and queue backlogs have not accelerated in proportion to data center development. The IEA assessed that data centers now account for approximately half of U.S. incremental electricity demand growth.1
Virginia illustrates how durable the trajectory is. Commercial electricity sales in the state climbed by nearly 30 million MWh between 2019 and 2025 — faster than any state except the much larger Texas — according to EIA data. That growth predates the current AI-driven build wave.4
Tesla, Sunrun, and Renew Home announced a 16 GW virtual power plant partnership that could in theory cushion peak demand events.2 Still, PJM's decision to embed curtailment rights for data centers in the backstop proposal suggests the grid operator is not relying on demand-response to close the supply gap.7
The September backstop auction is the first hard test of whether supply constraints are tighter than the bearish consensus allows. If PJM fails to fill the 6.8-GW shortfall, or if clearing prices come in above what forward markets currently assume for mid-2028 delivery, that would indicate the gap between what gets built and what gets connected is still widening rather than narrowing.7