The North American Electric Reliability Corporation filed its Cold Weather Data Annual Report with the Federal Energy Regulatory Commission on October 1, and the headline number is one the industry has worked toward since the February 2021 Texas freeze: 99 percent of total net winter capacity in the continental United States reported the ability to operate at calculated extreme cold weather temperatures during the winter of 2025–2026.
A large majority of capacity — 94 percent — operated at temperatures below 32 degrees, the threshold that triggers freeze-protection requirements under NERC’s cold weather standards. Released publicly this week alongside a companion review of best practices since Winter Storm Uri, it is the first full accounting of how the mandatory winterization regime is performing. The answer is that it is performing, measurably — and that the work is not finished.
The numbers behind the headline describe a system hardened in the aggregate, with specific weaknesses persisting at the edges. Generator owners reported total net winter capacity of 1,154,231 MW in the continental United States, a 7 percent increase over the winter 2024–2025 reporting period. Owners reported cold weather constraints for 633 units comprising 85,691 MW, with wind accounting for 71 percent of the units reporting a constraint. Units making up 4.3 percent of capacity are implementing corrective action plans to address reliability issues, and 60 percent of those plans will be completed before the winter of 2026–2027. A further 7.4 percent of capacity declared a constraint that precluded corrective action altogether. NERC’s judgment is that the risk from extended corrective action plans and cold weather constraints “is limited,” and forced outage rates in the last two winters were consistent with historical performance. That is a defensible reading of the data. It is not the same as saying the risk is gone.
The January test
It followed the joint FERC-NERC review of the January 2026 Arctic cold period, released September 10, which examined Winter Storms Fern and Gianna — the first severe, prolonged test of the winterized system. The review found generator outages comparable in magnitude to the January 2025 Arctic events, but overall performance improved relative to earlier winter storms, and there were no significant disruptions in either the electric or natural gas systems. Unplanned generator outages peaked at 68,169 MW on January 26, 2026, across the Eastern and Texas Interconnections, against 71,022 MW in January 2025. Behind the peak, 3,099 unique generators experienced 16,953 forced outages, derates, or failures to start. No manual load shed occurred in either January.
The comparison that matters is against the storms that produced the standards. During Winter Storm Elliott in December 2022, 90,500 MW of generation was unavailable in the Eastern Interconnection. In ERCOT, outages peaked at about 19,420 MW in January 2026, against about 35,300 MW during Winter Storm Uri in February 2021, when about 4.5 million Texans lost power. The cold period was not mild: the system carried 138.8 terawatt-hours of cumulative energy demand over eleven days, January 22 to February 1, and U.S. electric use peaked at 666 GW. More than a million customers still lost power during Fern, mostly from ice, falling trees, and damage to local distribution lines — a reminder that bulk-system reliability and keeping one house lit are different problems. But the bulk system held without rotating outages — the outcome the winterization standards were built to produce.
What made the difference
The joint review credits four things: lessons from past cold weather events; advanced commitments by generators ahead of the cold; strengthened preparedness and situational awareness; and implementation of NERC’s cold weather reliability standards. Two more deserve emphasis, because they show up in the data. The first is interregional transfers: the Midcontinent Independent System Operator’s imports peaked at about 9 GW, mainly from PJM, but also from Canada and neighboring systems — transfers that only work when the neighboring system has surplus to send. The second is the federal emergency orders: the Department of Energy issued 15 orders under section 202(c) of the Federal Power Act, the first of their kind, giving grid operators authority to direct backup generation at large loads to operate as a last resort. No one called on backup generation at data centers and other large loads to operate during the 2026 winter storms. The orders bought headroom that turned out not to be needed — the best possible outcome for an emergency authority, and also the reason to ask how much of January’s success depended on tools that are not part of normal operations.
The natural gas system also performed well by the review’s account: only moderate production declines and some limited, short-duration force majeure events — even as U.S. gas demand peaked at 172 billion cubic feet per day on January 24, one of the highest levels ever recorded, and the EIA reported a record net storage withdrawal of 360 billion cubic feet in the week ending January 30. The gas system met that demand. But the review’s own recommendations give the game away: more gas infrastructure, better gas-electric coordination, and better load forecasting. A system that needs those things is one whose January performance was closer to its limits than the absence of load shed suggests.
The caveat NERC itself states
NERC President and CEO Jim Robb said it plainly: “Both the electric and gas systems were stretched over multiple days, and multiple reliability challenges brought the system close to the edge. We cannot let this be the new norm of operating the grid.” That is the sentence to hold against the 99 percent figure. Capacity that can operate at extreme cold is not the same as capacity that did operate, without strain, through eleven days of Arctic cold. The 68,169 MW peak is smaller than Elliott’s and smaller than January 2025’s — but it is still 68,169 MW unavailable at the moment the system needed it most. The improvement is real, and so is the scale of what remains.
The companion ERO Enterprise report, Themes and Best Practices: Extreme Cold Weather, identifies significant and measurable improvements since Uri, driven by the standards, improved coordination, advance fuel procurement, and better preparedness. The work it catalogs is unglamorous: heat tracing that gets inspected, freeze-protection equipment maintained rather than installed and forgotten, fuel procured before the storm rather than during it. As this publication’s explainer on extreme weather planning set out, winterization moved from voluntary guidance to mandatory standards because the payoff arrives only in rare events, and markets underprovide investments whose return comes once a decade. The standards are doing what mandates are supposed to do. The question is whether they are doing enough of it, fast enough.
Where the constraints sit
The constraint data is the most useful part of the annual report: it shows where the aggregate numbers stop describing reality. Wind accounts for 71 percent of the units reporting a cold weather constraint — a declared limitation at a calculated temperature that operators plan around. But it does mean the constrained fleet is concentrated in a single fuel type, and that the capacity most likely to be limited in extreme cold is the capacity the system is adding fastest. The picture is uneven by region: about 13 percent of net winter capacity is constrained in ReliabilityFirst territory, about 12 percent in the Midwest Reliability Organization region, and about 4 percent in the Northeast Power Coordinating Council territory. A national 99 percent is an average over regions that differ by a factor of three.
Then there is the 7.4 percent of capacity that declared a constraint precluding corrective action — units that cannot fix the problem the standard identified. And of the 4.3 percent of capacity on corrective action plans, 40 percent of those plans will not be complete before this winter. NERC calls the resulting risk limited, and on the numbers that is fair: the constrained and uncorrected capacity is a small share of a fleet that grew 7 percent in a year. But winter risk is not distributed evenly. It concentrates where the system is already stretched, and a few gigawatts unavailable in the wrong place at the wrong hour is what turns a tight day into an emergency. The annual report measures the fleet. The storm will measure the margin.
PJM’s winter, in miniature
PJM’s winter outlook, published this month, is the regional version of the same story. The operator expects 182,426 MW of operational capacity against a forecast winter peak of about 147,500 MW, plus about 3,200 MW of average exports to neighboring systems. If reached, that peak would set a new winter record, beating the 143,700 MW set on January 22, 2025. PJM has also studied a higher-stress scenario — demand up to 152,200 MW with less generation available — and says it could still keep the lights on by calling pre-emergency and emergency demand response. Forecast winter peak demand has risen by about 7,500 MW, or 5 percent, since 2022.
The arithmetic looks comfortable — roughly 35 GW of headroom — until it is read alongside the January experience. In early 2026, PJM held reliability through the most prolonged extreme cold since the 1990s, with hourly winter peaks above 130,000 MW for eight straight days, January 26 to February 2. Eight days at that level strains fuel supplies, crew endurance, and equipment thermal limits in ways a one-day spike does not — the duration problem this publication’s explainer flagged as the hardest part of the new planning reality.
PJM is also explicit that reserve margins continue to tighten as demand outpaces new supply, and that reliable winter operations “will depend on continued preparation and strong generator performance, particularly during periods of extreme cold” — with transmission constraints limiting delivery in some Maryland, Virginia, and District of Columbia zones.
The demand underneath the readiness
The reason the margins keep tightening is not a mystery. The EIA expects U.S. electricity sales to total 4,135 billion kilowatt-hours in 2026, up almost 2 percent from 2025 and rising to 4,211 billion in 2027, driven by data center development and increased manufacturing activity. NERC’s Long-Term Reliability Assessment, released in January 2026, found winter peak demand growing at the fastest rate since the organization began tracking reliability data — 246 GW of projected growth over the next decade, with demand up 20 GW since last winter alone. The winterization story and the load-growth story are the same story from opposite ends: the fleet is better prepared for cold, and the peak it must serve is larger every year.
Load forecasting is one of the joint review’s three named areas for further work, and January supplied a vivid illustration of why. School and business closures during the Arctic cold period cut demand in PJM by 2,000 to 3,000 MW — 1.4 to 2.2 percent of the region’s 139,410 MW peak — and several operators reported that closures and other behavioral responses tended to lead to over-forecasting. ERCOT, ISO New England, and NYISO all showed load-model and weather errors moving in the same direction, amplifying total forecast error. The lesson is not that forecasting failed — the system held — but that the largest source of forecast error in January was human behavior, the hardest variable to model as electric heating and large flexible loads grow as a share of the winter peak.
What to watch
The 2026–2027 NERC Winter Reliability Assessment is not yet published — it is scheduled for release around November 18, with a public webinar on November 19 — so its findings are not part of this story. When it lands, the number that matters is not the headline adequacy finding — resources are almost certainly adequate for a normal winter peak — but the elevated-risk finding for prolonged, wide-area cold snaps, where the last assessment flagged several regions at elevated risk.
Three other things will tell us whether this winter validates the preparation or exposes it. The first is the corrective action plans: whether the 60 percent due before winter actually complete, and what happens to the 40 percent that do not. The second is the gas system in a second severe cold period — January showed the gas system could meet a record demand week, but record storage withdrawals draw on a finite buffer.
The third is the first real cold test of the season: whether generator performance matches the reported capability, whether the wind fleet’s declared constraints behave as modeled, and whether last winter’s emergency orders become routine rather than a one-off backstop.
The January data confirms the bottom line this publication’s explainer reached a month ago: preparation works, and preparation is not the same as safety. The standards have made the fleet measurably better at surviving cold it used to fail in. The demand the fleet must serve keeps growing, the gas system it depends on is asked to do more with the same pipes, and the storms keep lasting longer than the models assumed. The grid entering this winter is the best-prepared the country has ever had. It is about to be tested by the hardest winter load it has ever faced. Both of those things are true, and the second is why the first is not enough.