What does the current auction actually say?
PJM procured capacity is below its own reliability requirement. Installed ICAP, accredited UCAP, and queue nameplate are not interchangeable.
Read this decision layer →The decision
Approve large loads only with the capability that makes them supportable.
The current evidence does not support a clean yes-or-no answer to “Can PJM absorb the surge?” It supports a contract and sequencing answer: match energization to accredited supply, demonstrate deliverability, and require tested data-center flexibility until both are in service.
The simple model already changes the decision. PJM's 2028/29 auction procured 149,182 MW UCAP—6,831 MW below its 156,013 MW requirement. PJM explicitly says that does not necessarily mean load cannot be served; it means slimmer reserves and greater risk.
Capacity ledger
Four quantities that cannot be swapped.
QuantityCurrent valueWhat it means
Eligible capacity202.3 GW ICAPNameplate-like installed capacity eligible for the auction or FRR. Not the accredited reliability contribution.
Procured capacity149.2 GW UCAPAccredited RPM + FRR capacity procured for 2028/29.
Reliability requirement156.0 GW UCAPPJM's target capacity requirement for the delivery year.
New + uprated cleared0.5 GW UCAPCleared contribution from new resources and uprates—not all planned or queued projects.
Question chain
One core screen, five decision layers.
Will the megawatts perform and arrive?
An Elliott-window counterfactual and a dated Dominion import-capability screen expose two risks that a system total hides.
Read this decision layer →What can this model honestly price?
It can show a planning-level capital requirement and market signal. It cannot turn those into a retail-rate forecast.
Read this decision layer →When does self-supply help the grid?
Only capability that is grid-connected, accredited, deliverable, and available when needed earns a reliability credit in this screen.
Read this decision layer →How much does curtailability change?
Flexibility is applied only to incremental data-center load and only in hours that would otherwise cross the ceiling.
Read this decision layer →ADM fidelity ladder
Add detail only where it changes the condition.
Authoritative accounting
The auction shortfall removes the premise that current accredited capacity offers unclaimed headroom.
Hourly weather replay
Nine complete demand years show when a transparent 170 GW scenario ceiling is crossed. This produces shortage-screen hours, not LOLE.
Sensitivity surface
Exact combinations of bridge capacity, new firm capability, and DC-only flexibility show which conditions move the answer.
Selective stress checks
Elliott adds performance risk; Dominion adds locational deliverability. A fabricated cascade model adds neither and was retired.
Stop at the trust boundary
A project approval still requires PJM accreditation, transmission, and interconnection studies. This public screen does not impersonate them.
What changed in the repair
The impressive-sounding claims were the ones that had to go.
No fake LOLE
The former nine-hour proxy is gone. Shortage hours and events are labeled as screen outputs.
No cascade theater
The hand-built logistic cascade and population-impact claims are removed.
DC-only flexibility
Demand response can now reduce only the incremental DC load, never an arbitrary percentage of total PJM demand.
Traceable publication
The page, interactive JSON, tests, source hash, script hash, assumptions, and correction log now travel together.
Primary sources
Dated facts, linked directly.
- Hourly Electric Grid Monitor — Form EIA-930U.S. Energy Information Administration · 2015-2024 snapshot
- Inquiry into Bulk-Power System Operations During Winter Storm ElliottFERC and NERC · 2023-10
- 2026 Long-Term Load Forecast overviewPJM · 2026-01-14
- 2027/2028 RPM Base Residual Auction Planning Period ParametersPJM · 2025
- 2028/2029 Base Residual Auction ReportPJM · 2026-07-14