Every major U.S. grid operator now has to forecast demand from data centers and other very large customers. ISO New England has done so for the first time, and its answer stands out: almost nothing is coming.
The operator's 2026-2035 Forecast Report of Capacity, Energy, Loads, and Transmission, known as the CELT Report and published May 1, includes a large load forecast for the first time. Survey data from transmission owners suggest New England has just two proposed large load projects in the formal study phase, with combined demand of 285 MW at most, ISO New England said in a May 18 explainer. It expects them to add around 110 MW to summer and winter peak demand in the 2030s and about 130 MW in the 2040s.
The large load component is one of six building blocks in the forecast, alongside base demand in homes and businesses, electric vehicles, heating electrification, behind-the-meter solar and batteries paired with that solar. The operator now models demand hour by hour, a technique it developed during the 2025 cycle, rather than focusing only on annual energy and seasonal peaks.
How the forecast was built
To qualify for the forecast, a project must be larger than 20 MW in expected nameplate capacity and must have entered a formal study agreement with a transmission owner, according to the operator's March 27 presentation to its Load Forecast Committee. Two projects met that bar: a 200 MW data center in the Northeast Massachusetts zone with a timeline through 2027, and an 85 MW general electrification load in Connecticut with a timeline through 2040. Neither was under construction, so both were excluded from the summer 2026, winter 2026/2027 and summer 2027 periods.
The nameplate numbers are then cut to reflect what is likely to be built and used. ISO New England derates each project based on its maturity, how quickly it ramps up and how much of its requested capacity it will actually use. For medium-uncertainty projects, the effective result is typically 40% to 80% of the requested nameplate. The presentation notes that data centers will often be derated more heavily than other projects. A project that has just signed a study agreement is derated for the chance it will never be built.
The result is modest, even after a decade of growth. Large loads are expected to use around 800 GWh a year from 2030 to 2035, rising to about 1,000 GWh in the 2040s, and to have no effect on overall system demand before winter 2027/2028.
"There's a lot of work going on to get up to speed in this domain, and that's going to continue as the forecast evolves," said Victoria Rojo, ISO New England's load forecasting supervisor. The operator now sends a quarterly survey to transmission owners to track prospective large loads and judge how likely they are to proceed.
The contrast with New York
The gap with New England's western neighbor is stark. In its 2025-2034 Comprehensive Reliability Plan, the New York Independent System Operator said its interconnection queue held over 10,000 MW of large load projects seeking service before 2031 as of September 2025, more than double the roughly 4,000 MW at the end of 2024. New England's qualifying total is about 3% of that.
ISO New England's March presentation put it plainly: "To date, New England has not witnessed the scale of data center proposals seen in other regions of the country."
What drives New England's demand instead
Electrification, not computing, is the main growth story for the region. The CELT Report projects net annual energy use rising about 9% over the next decade, or 0.9% a year, reversing a downward trend since 2005 driven by efficient appliances, lighting and heating and by behind-the-meter solar. Electric vehicles are expected to add 7,074 GWh of annual use in 2035 and heating electrification 7,165 GWh. Against those figures, the large load forecast is a rounding error.
The forecast is slightly lower than last year's because of more conservative assumptions about electric vehicle and heat pump adoption in light of government policy changes. Behind-the-meter solar is projected to cut grid consumption by 7,056 GWh this year and 10,179 GWh in 2035. The 50/50 summer peak forecast rises from 25,228 MW in 2026 to 26,849 MW in 2035, and the operator expects the winter peak to nearly match the summer peak by 2035 as more homes switch to electric heating. Under hotter 90/10 conditions, the summer peak forecast rises from 26,473 MW in 2026 to 28,503 MW in 2035.
Why the boom skipped the region
The operator does not explain why developers have stayed away, and its forecast documents focus on method rather than market conditions. What they do show is that the screen is strict. Proposals that have not signed a study agreement, or fall below 20 MW, do not count at all, and those that do are derated for maturity and utilization. A region where developers are only exploring sites would look much like New England does now.
For the region's consumers, the lack of data center demand is a mixed result. It spares New England the cost allocation fights over large loads now playing out in PJM and Texas. It also means the region is not attracting the investment and tax base that data centers bring to places like Northern Virginia and Ohio.
What to watch
The quarterly transmission owner surveys are the early warning system. If more projects sign formal study agreements, they will enter the next CELT cycle and could change the forecast quickly, since a single hyperscale campus can be several times the size of the current pipeline. The 200 MW Massachusetts data center, with a timeline through 2027, will be the first test of whether the region's sole qualifying project gets built on schedule.
For now, New England's planners are preparing for a boom that has not arrived. Building the forecasting framework before the load shows up is a lesson other regions learned the hard way.