In late July 2026, PJM's independent market monitor — Monitoring Analytics, Joseph Bowring — reported that data centers drove roughly $6.3 billion of the most recent PJM capacity auction, about 38% of a $16.4 billion total, and $29.4 billion, or about 46%, across the last four auctions combined. Bowring called the load surge a paradigm shift PJM was treating as business as usual, and pressed for a separate capacity auction for large loads.
That is the cost-shift, priced. It is not a queue screenshot and not a vacancy print. It is why large-load tariff and moratorium climate is 20% of the HD Market Scorecard and why PJM is a D.
What followed
The monitor's report landed three weeks before the August 17, 2026 deadline for the six RTOs and ISOs to answer FERC's June 18 Section 206 show-cause orders on large-load interconnection. New York had already enacted a statewide hyperscale moratorium. New Jersey's Assembly had passed a ≥50 MW take-or-pay tariff. PJM, separately, began putting some cost-allocation questions onto the states.
Reliability is the other half. After 3.8 GW of data-center and crypto load tripped offline, PJM and NERC opened a conversation on computational-load performance. Curtailment risk and capacity cost are now the same political object: who pays, and who sheds, when the load is the size of a city.
How HD scores it
PJM's grade is D, trend down. Elevated capacity prices, evolving co-location guidance, and reliability-driven curtailment are the rationale. Northern Virginia vacancy at record lows does not help the grade; it is a real-estate fact sitting on top of a grid that has not isolated new large load from other customers.
The scarce asset in PJM is powered land with a contracted, ring-fenced path that does not depend on socialized capacity cost staying cheap. Queue position that assumes the last four auctions were an anomaly is the trade the monitor just marked.