PJM's Data Center Demand Story Has a Demand-Drop Problem
A mystery 3,000-megawatt load disappearance, a FERC ruling on virtual power plants, and a price cap distorting investment signals complicate the clean bull case on PJM capacity.
Federal regulators on August 13 (2026-08-13) ordered PJM Interconnection to allow virtual power plants into its capacity market, opening the door for batteries, demand response, and distributed solar to compete directly against conventional generators in the country's largest power market. Most capacity traders appear to have set the ruling aside. PJM Western Hub spot power stood at $73.72/MWh at its last recorded session on August 30 (2026-08-30), and the prevailing posture heading into the autumn forward window remains broadly bullish — driven by record auction results, a widening capacity shortfall, and surging data center load.6
The auction record gives that bullishness solid footing. PJM's 2028/2029 Base Residual Auction cleared at $554.72/MW-day across most of the footprint, with the COMED load delivery area at $776.69/MW-day. Generator payouts for the delivery year starting June 2028 matched PJM's all-time high of $16.4 billion, according to PJM. The auction also came up 6.8 GW short of the reliability requirement — a gap large enough to push PJM to accelerate a backstop reliability auction to September (2026-09) rather than waiting until 2027.3,1
Yet Monitoring Analytics, PJM's independent market monitor, found that the $554.72/MW-day clearing sat roughly 70% above the $325/MW-day level at which price cap mechanisms had previously held the market. Without those caps, the monitor concluded, clearing prices would have been 70% higher. Developers calibrating new capacity investment to published clearing prices are working against a constrained signal, not a free-market one, and the gap between capped and uncapped levels is an input that standard pro-forma models may not fully capture.3
Capacity rates across PJM have risen more than 1,000% over the past two years, according to Utility Dive, and first-quarter 2026 power prices jumped 76%, with data center expansion widely cited as the driver per Monitoring Analytics data. The structural demand thesis is real. Those costs are already showing up in municipal utility and cooperative budget cycles across the 13-state footprint.2,3
But on the morning of Wednesday, July 22 (2026-07-22), PJM's measured demand fell roughly 3,000 MW in the space of minutes just before 08:00 local time, producing a visible frequency spike across the grid. Wattclarity documented the event using real-time demand trend data and noted that PJM had published no reporting or explanation for the drop. At a grid that dispatches over 100 GW on peak summer days, 3,000 MW is modest in absolute terms. The speed of the disappearance is not. Demand that can fall by that amount in minutes carries embedded flexibility — curtailment agreements, automated load shedding, or behind-the-meter response — that sits awkwardly alongside the assumption of inelastic, ever-rising data center-led load.4
The prevailing assumption is that data centers do not curtail. That may be true for the servers themselves. But they co-locate with other load classes, and industrial and commercial demand running alongside them responds to price signals in ways that can aggregate into material swings in real-time draw. An unexplained 3,000 MW event does not disprove the structural demand thesis, but it raises questions about how much flexibility is embedded in a load profile the market is treating as purely inelastic.4
The FERC ruling on virtual power plants adds another layer. Qualifying resources under the order include newly built or upgraded power plants, solar, wind, battery projects, and demand-response aggregations, according to Canary Media. If VPP developers begin aggregating latent flexibility — the kind visible in the July 22 (2026-07-22) real-time data — effective peak demand, the load that determines capacity need at critical hours, could compress below what the 2028/2029 clearing level implies. The qualifying volume and timeline remain uncertain.6,5
PJM is also directing states to manage peak shortfalls locally rather than solving the gap centrally, a structure Canary Media reported on August 5 (2026-08-05). That decentralisation slows coordinated new-build pipelines but creates room for state-level demand management programmes that could aggregate into real reductions during critical peak hours.5
The September (2026-09) backstop auction PJM pulled forward by more than a year is the nearest hard data point to test that math. Clearing volumes there — particularly the share from VPP and demand-response resources newly eligible under the FERC order — will show whether the gap between the regulatory cap and true scarcity pricing is as wide as the Monitoring Analytics analysis suggests, or whether new flexibility resources are already eroding the shortfall that drove this year's record generator payouts.1,36