NERC sees 20.2 GW winter capacity gain as batteries outpace wind de-rating
Winter peak capacity grows 2.5%, but wind output valuations fall 6% in a fleet shift that tightens the cold-weather reserve calculus.
NERC’s winter outlook for 2026-27, released in early June, puts total bulk power system capacity at 20.2 GW above last winter’s planning forecast, a 2.5% gain that appears to ease the supply picture heading into peak season.2
The number flatters the underlying mix. Generation accounts for just 1,335 MW of the 9,445 MW net increase in on-peak available capacity, with demand response programs contributing the larger share.2 Batteries add 19,659 MW of nameplate capacity and 11,121 MW of peak-demand capability, while new natural gas-fired generators contribute a smaller increment, NERC’s assessment shows.2
Wind is the drag. Revised valuations of wind resource capability at peak demand hours in several assessment areas cut the effective on-peak rating by roughly 6%, NERC said, which offsets part of the battery and gas build.2
That matters for the three markets traders are watching hardest this winter: PJM, ERCOT and MISO. All three cleared the last test, Winter Storm Fern, but by a margin that grid operators described as “very close to the edge,” according to a review published in June.5
Five years after Winter Storm Uri, the bulk power system has rewired its cold-weather playbook through EOP-012-3, cold-weather critical component inventories, and dual-fuel conversions.5 The system is more prepared than it was, yet the operator commentary after Fern suggests the reserve cushion remains thinner than the headline capacity numbers imply.
PJM is acting pre-emptively. The RTO said in its seasonal outlook that it will run unannounced operational tests on generators that have not run for several weeks leading into winter, checking they can deliver in cold before extreme weather arrives.3 The tests respond to the known limitations of certain units in extreme cold, PJM said.3
MISO’s seasonal readiness material shows what those limitations cost in practice. In the mid-January event captured by the RTO, a 17 GW incremental outage coincided with a 6.5-degree-F temperature and 108 GW peak demand; the three-day uplift bill ran $4 million.4 A few days later, at 13 degrees F, outages fell to 9 GW and the uplift to $1.5 million.4
The pattern is the point for real-time traders: winter uplift in MISO is a function of outage size, not temperature alone. The spread between the two events gives a rough sense of how much capacity actually shows up when the mercury drops.
ERCOT is facing a different side of the same problem. The Texas grid operator is set to vote on rules that would require data centers to install controls preventing sudden trips offline, which grid managers warn could cascade into blackouts.1 A large load that can disconnect in an instant is a winter risk as much as a summer one, and ERCOT has no capacity market to price that contingency.
The consensus view among market participants is bullish on PJM, ERCOT and MISO real-time power for the winter, with five separate signals pointing in that direction and none bearish. [consensus] The driver is straightforward: capacity is growing slower than the planning forecasts suggest, and the de-rated wind values tighten the margin when peak demand actually shows up.
The counterweight is demand. Data center load growth is the reason grid operators are writing new rules in Texas, and it is the reason PJM’s unannounced tests exist.1,3 If winter peak materializes on a cold, still night, the battery fleet will be the marginal resource, and its four-hour duration limit will be tested against an 11 GW peak capability that NERC has already flagged as the core of the added capacity.2
For traders, the real-time signal to watch is MISO’s day-ahead commitment efficiency against the incremental outage pattern in the seasonal readiness data.4 The gap between the 17 GW outage event and the 9 GW one is the difference between a $4 million uplift and a $1.5 million one. That spread is winter’s actual risk premium, and it has nothing to do with the nameplate numbers.