Winter capacity math is flattering PJM and MISO. The demand side is moving faster.
NERC sees 20 GW more winter capacity. Data centers and cold-weather limits may erase the cushion before January.
PJM is still sorting through what happened in mid-August, when about 3,800 MW of computational load tripped offline in one event, and the grid operator said Tuesday (2026-08-11) it is now weighing new interconnection reliability requirements for data centers and crypto-mining facilities. The proposed "ride-through" standards are a direct response to that August trip.6
That August event matters less for what it did to prices that day than for what it signals about the coming winter. PJM's own seasonal outlook, published in early June, leans heavily on unannounced operational tests of generators that have not run for several weeks heading into cold weather, a recognition that units fail when called upon after long idle stretches. The tests are meant to ensure resources are ready.2
The headline number from NERC's Winter Reliability Assessment is a 20.2 GW, or 2.5%, increase in forecast bulk power system capacity over last winter. Battery additions account for a large share, along with some new natural gas-fired generation. But dig into the table and the picture gets thinner: generation accounts for just 1,335 MW of the total 9,445 MW increase in peak demand capacity, with demand response programs contributing the larger share.1
Wind capacity values have also been revised down by as much as 6% for peak demand hours in some areas, which partially offsets the headline gain. Cold winter peak hours tend to be still, and NERC has baked that into its planning numbers. The market consensus reading PJM real-time, MISO real-time and ERCOT real-time as bullish for winter may be leaning on a nameplate figure that overstates what those batteries and demand response programs will actually deliver in a three-day cold snap.1
The five winters since Winter Storm Uri have rewritten the playbook on cold-weather preparation, from EOP-012-3 to dual-fuel conversions and cold weather critical component inventories. Yet Winter Storm Fern, the most recent test of that regime, showed the system still ran "very close to the edge," according to reporting published Monday (2026-06-08). Five years of weatherization work has not eliminated the tail risk; it has moved the baseline.4
Battery storage is the biggest single source of new winter capacity, and it is also the resource most likely to be depleted before the third day of a prolonged cold event. MISO's own seasonal readiness data shows how quickly uplift costs spike as temperatures fall: a stretch with peak demand at 108 GW and temperatures around 6.5 degrees Fahrenheit produced 17 GW of incremental outages and a 3-day uplift total of $4 million, while a milder window at 13 degrees saw outages fall to 9 GW and uplift drop to $1.5 million.3
The difference between those two MISO cases is roughly 6.5 degrees Fahrenheit. That is the sensitivity range that matters for winter pricing. A forecast that calls for 108 GW of peak demand is built on assumptions about wind output, battery state of charge and generator availability that all degrade non-linearly as temperatures drop. The gap between a $4 million uplift event and a $1.5 million event is not a planning margin; it is the difference between a normal cold day and a scramble.3
On the demand side, PJM's computational load issue is not going away. The August trip of 3.8 GW was not a weather event. It was a facilities issue at data centers and crypto-mining operations that drew power and then lost it in a way the system operator did not anticipate. Ride-through standards, once implemented, will impose costs on those loads. But between now and January, those facilities remain connected under current rules, adding demand that NERC's winter forecast may not have captured at the same granularity as traditional load.6
DOE emergency orders have also kept fossil-fueled plants online that would otherwise have retired, and those units are producing significantly less electricity than they did before the orders took effect. Keeping a plant in service and keeping it available at peak are different things. If those units are the ones PJM's unannounced winter tests find wanting, the capacity they contribute on paper will not translate into energy when the cold arrives.5
The bearish case for PJM real-time this winter is straightforward: if the market is pricing cold scarcity, it is ignoring that the incremental capacity cushion is heavy on batteries and demand response, light on dispatchable generation, and now facing a load class in data centers that has already shown it can trip 3.8 GW in one event. The bullish case rests on the 20.2 GW headline and the weatherization progress of the past five winters. Both can be true until the third cold day, when batteries are drained and the demand response programs have already been called.1
What would confirm the bearish view is a cold-weather event in December or January that lasts more than 48 hours, combined with real-time prices at MISO Indiana Hub and PJM Western Hub that stay elevated after the first day, rather than spiking and collapsing as emergency resources come online. What would falsify it is a PJM test season that finds the re-energized DOE units and new gas plants perform as advertised, and a data center fleet that holds its load through a cold snap without tripping. The difference between those outcomes is likely to show up first in the uplift line items, not the capacity headlines.3