PJM's winter capacity math understates data-center volatility and demand response risk.
NERC's 20.2 GW capacity gain looks solid until you examine what it's made of — and what tripped offline in August.
PJM is considering mandatory "ride-through" standards for data centers and crypto-mining facilities after roughly 3,800 MW of computational load tripped offline in a single event, the grid operator said on Tuesday (2026-08-11). That figure deserves more attention than it's getting from traders pricing winter real-time power.6
The market's constructive view on PJM and MISO real-time this winter leans heavily on NERC's June reliability assessment, which put total bulk power system resource capacity up 20.2 GW, or 2.5%, versus last winter's forecast. Battery additions account for 19,659 MW of nameplate capacity and 11,121 MW of peak demand capacity in that figure. PJM Western Hub spot settled at $73.72/MWh on Monday (2026-08-24), with MISO Indiana Hub at $70.25/MWh — still range-bound, with no weather premium yet.1
But the composition of that 20.2 GW is less reassuring than the headline suggests. Generator capacity accounts for just 1,335 MW of the total 9,445 MW increase in available capacity; the larger share comes from demand response programs. Demand response is only as reliable as the customers enrolled in it. In a February cold snap, when residential heating demand peaks and commercial load is at its highest, enrolled customers face their own pressure to keep operating.1
Wind capacity values have also been revised down by 6% at peak demand hours in several assessment areas, per the NERC tables. The increase in total capacity is partially offset by those lower on-peak wind valuations — meaning the resource mix NERC is counting on for a cold snap leans harder on intermittent sources precisely when the system needs firm megawatts. PJM will perform unannounced operational tests of generators that have not run for several weeks heading into winter, an acknowledgment that thermal availability cannot simply be assumed from capacity market enrollment.1,2
The 3.8 GW data-center trip is the sharpest challenge to the bullish demand thesis. PJM's winter planning treats computational load as a stable, growing demand base. The August event shows it can shed 3,800 MW in a single incident — and come back on quickly, creating two-sided volatility in real-time prices that seasonal assessments are not designed to capture. If PJM's ride-through discussions produce firm curtailment protocols or service requirements for these facilities, peak demand estimates will need revision downward, freeing capacity the market currently prices as scarce.6
MISO's own readiness data adds texture to the battery question. Last January's cold event, with temperatures dropping to 6.5°F and renewable output at 19 GW, produced 17 GW of incremental outages and a three-day uplift total of $4 million. On warmer days during the same stretch — temperatures at 13°F — outages fell to 9 GW and three-day uplift totaled $1.5 million. Batteries help with short-duration ramps, but a multi-day polar event drains stored energy quickly, and last winter's data shows unit commitment efficiency collapsing under sustained cold.3
Plants operating under DOE emergency orders add a further complication. Power Dive's reporting from June found those plants are producing significantly less electricity than before the orders were issued. Capacity counted in NERC's assessment may overstate actual output from units that are nominally available but underperforming — a gap between nameplate and realized generation that would surface in real-time markets during an extended cold event.5
Winter Storm Fern, the most recent system-wide test, showed the grid ran "very close to the edge," according to Power Magazine's June analysis. Five years of weatherization standards, cold-weather critical component inventories, and dual-fuel conversions since Uri have narrowed the tail risk but not eliminated it.4
The contrarian view — bearish PJM real-time with 65% confidence, per the consensus tracker — hinges on demand. What would confirm it: a PJM load forecast update that revises down computational load assumptions, or formal curtailment rules for data centers that visibly reduce winter peak demand projections. What would cut against it: a cold December in which the battery fleet sustains multi-day discharge through an event comparable to what MISO saw last January. That test hasn't come since the buildout accelerated, and the August data-center trip is a reminder that the load side of this equation is less predictable than the capacity additions suggest.6,3