Winter came early.
How electrification is bending New England's load and cost curves, on seven years of real ISO New England data. The grid context and four findings, for planners.
The ground it stands on
New England runs one grid, operated as a single market by ISO New England across six states. Massachusetts is the largest load on it, about 46 percent of the region's electricity, split into three zones, with Greater Boston alone larger than any other single zone. The system is gas-dominated, roughly 56 percent of grid-scale generation is natural gas on top of a 25 percent nuclear base, and that is the whole story, because the same pipelines that fuel the power plants also heat the homes. On the coldest days gas goes to heating first, starving the generators exactly when electricity demand is highest. New England is usually called summer-peaking, with the winter transition placed in the 2030s. On the data, the transition is further along than that, and it is already visible in load, in price, and in the value of flexibility.
→ colder load lands on a gas-constrained supply stack → winter price convex and rising
Method, briefly: daily ISO-NE demand and Internal (Mass) Hub settlement prices, 2019 to mid-2026, against a population-weighted New England temperature from ERA5 and Henry Hub gas. About 2,700 daily observations. Slopes are compared across an early window (2019-21) and a recent window (2023/24-26), with bootstraps and a gas control.
Finding 1: the cold side of the load curve is steepening
The demand response below 50F rose from 3.0 to 3.9 GWh per degree between 2019/21 and 2024/26, a 29 percent increase, positive in every one of 2,000 bootstrap resamples. The summer air-conditioning side barely moved. Cold weather now pulls materially more electricity than it did four years ago, and only on the cold side. This is the electrification signature isolated in the record.
Finding 2: winter has already overtaken summer in energy
In daily-energy terms the seasonal crossover is not a projection. Winter mean daily demand moved above summer around 2025, as summer softens under efficiency and behind-the-meter solar while winter climbs. The peak-megawatt crossover is a distinct measure and still ahead, but the direction of travel is in the record now, not in a forecast.
Finding 3: the winter price is convex, and its cold sensitivity has tripled
Because the winter grid is gas-constrained, cold load clears at the steep end of the supply stack. On-peak day-ahead price averages about 37 dollars per MWh in mild weather but 187 below 10F, where 86 percent of days spike above 200 dollars in real time, and the coldest tenth of days carry more than twice the price of the rest of the year. More consequential, that cold sensitivity is itself rising: controlling for gas, the price move per degree of cold went from 1.23 dollars in 2019-21 to 4.00 in 2023-26, more than three times steeper on essentially the same gas price. The load-side steepening in Finding 1 shows up in price, and gas does not explain it. The customer cost of a cold snap is now growing faster than the weather alone would imply, because the same cold draws more load onto a constrained system.
Finding 4: so the value of flexibility is rising every winter
Two consequences follow for planning. A load forecast fit on history under-forecasts the winter peak, in a like-for-like backtest a black-box model missed it by about 5 percent while a mechanism-based forecast tracked it, so the drift is a planning-horizon risk, not an operations one. And because winter cost concentrates in a few cold hours and that concentration is growing, the value of shaving those hours rises too. In a simple valuation that holds weather and gas fixed and moves only the measured cold-price slope, a call on winter on-peak power that was worthless under the 2019-21 slope is worth about 22 dollars per MWh under the 2023-26 slope. The same megawatt of managed charging, thermal pre-heating, networked geothermal, or storage is worth more each winter than the last.
The winter is not arriving. It has arrived in the data. The open question for a planner is no longer whether, but how fast the value of flexibility is compounding, and whether the forecast that sizes it is keeping up.
What it means for planning
None of this dictates a decision, but it moves three questions from stable to time-varying: the winter peak forecast, which a stationary fit understates; the value of demand flexibility, which is rising with the cost sensitivity; and the customer-cost case for winter reliability, which is now as much a load story as a gas story. The load lever, unlike the pipeline, is one a utility and its programs can actually move. These are independent, public-data observations to test against a planner's own richer data, not conclusions.
Caveats
Per-degree slopes and bootstraps address the largest confounders but not all of economic growth, population, and weather-year composition. Behind-the-meter solar accelerates the energy crossover independently of heating, so it should not be read as a peak-megawatt crossover. The day-ahead operational forecast is well-calibrated; the drift is a planning-horizon concern, not a claim that operators are wrong tomorrow. The gas control is Henry Hub, not the Algonquin basis that drives cold-snap scarcity, so the cold coefficient bundles a load channel and a gas-basis channel, though the tripling on similar gas points at load. In-sample throughout.