The forecast that Texas transmission utilities handed ERCOT this year would more than double the grid's size in six years. In materials for its April board meeting, ERCOT showed a utility-provided long-term load forecast that climbs to 218 GW by 2031. Most of the growth is new data center load, and most of that is supported not by signed contracts but by letters from utility officers attesting that the load is coming.
ERCOT is not adopting the number as its planning base. It has built an adjusted forecast that applies discounts drawn from its own experience of how much requested load actually shows up, and when. The result is still a very large increase, but the gap between the two forecasts is the clearest measure yet of how uncertain data center demand has become.
Where the 218 GW comes from
ERCOT's board presentation explains that its forecast process changed because of state law. House Bill 5066 did not apply to ERCOT directly, but two rule changes that followed, NPRR1180 and PGRR107, did. Those rules allow load forecasts for transmission planning to be supported by executed contracts, credible third-party forecasts, or a letter from a transmission or distribution utility officer attesting to the load. The Public Utility Commission of Texas approved the two revisions on January 21, 2025. ERCOT says the most important difference from its previous process is that it must now accept officer-attested letters as reasonable.
The effect on this year's request for information was dramatic. Seventeen transmission utilities responded, up from seven in 2024. The number of contract loads rose from 64 to 132, and the number of officer letter loads from 92 to 281. The 2025 survey included 55 GW of additional data center load supported by officer letters for the forecast year 2030. ERCOT's chart of the utility-provided forecast steps up from 94 GW to 109 GW, 138 GW, 173 GW, 197 GW, 208 GW and finally 218 GW at the end of the period.
Data center growth for 2030 alone rose from 29,614 MW in the 2024 forecast to 77,965 MW in the 2025 forecast. ERCOT's summary was blunt: new data centers continue to be the major area of new growth.
The discount
ERCOT's adjusted method applies three corrections, each based on what has actually happened on its grid.
First, it delays the in-service date of all new contract and officer letter loads by 180 days. ERCOT says large loads with in-service dates in 2022 to 2024 came online about 220 days late on average.
Second, it reduces all new data center demand to 49.8% of the requested amount. That is the share of requested load that data centers with 2022 to 2024 in-service dates actually used.
Third, it reduces all officer letter loads to 55.4% of the request, which is the share of projects with 2024 in-service dates whose load was in service by February 2025.
ERCOT notes that these factors can be updated as more contract and officer letter loads are energized. For forecast year 2030, the adjusted approach sits within the 130 GW to 148 GW range used in its 2024 Regional Transmission Plan. Even after the haircuts, ERCOT says data center load remains the largest growth category.
The long-term forecast report ERCOT published on April 8 puts numbers on the adjusted path. Summer peak demand rises from 85,759 MW in 2025 to 94,650 MW in 2026, 121,543 MW in 2028, 138,944 MW in 2030 and 144,522 MW in 2031. Annual energy climbs from 486 TWh to 1,038 TWh over the same period. The report notes that historical energy use grew at an average annual rate of 3.1% from 2014 to 2024, while the forecast rate for 2025 to 2031 is 13.6%. In other words, even the discounted forecast more than doubles the energy the grid must deliver in six years.
Where the adjusted forecast will be used
ERCOT will use the adjusted forecast for its Capacity, Demand and Reserves report starting in May, with the utility-provided forecast shown as a comparison scenario. The Regional Transmission Plan will also use the adjusted forecast, and ERCOT says it has begun talks with market participants on how to fold the utility forecast into longer-term transmission planning. Regional Planning Group project reviews will start from the adjusted forecast but will accept the utility forecast. ERCOT also plans to update the calculation of how much generation it allows to be out for maintenance on any day.
The distinction matters because demand forecasts drive transmission spending, reserve margins and price signals. Planning to the higher number could mean building lines and substations for load that never arrives. Planning to the lower number could leave the grid short if the data centers do come.
How fast can supply follow
ERCOT's presentation included a reality check from its own history. In the early 2000s, more than 27,000 MW of new gas generation was added over a five-year period. In the last three years, almost 25,000 MW of nameplate wind and solar was added, along with about 9,000 MWh of energy storage. ERCOT offers these as rough guides to how much new generation can be reliably connected in a short period.
Set against a forecast that adds more than 120 GW in six years, those historical rates look small. Even the adjusted forecast, which reaches about 145 GW in 2031, implies supply additions faster than anything the grid has seen. That is the core tension: data center developers can sign letters and buy land quickly, but generation and transmission take years.
What the officer letters reveal
The officer letter rule was meant to keep ERCOT's planning from lagging behind real demand. It has done that, and it has also exposed how speculative much of the data center pipeline is. Utilities pass along what customers ask for, and ERCOT's 49.8% factor for data centers is an empirical estimate of that overstatement.
The adjusted forecast is ERCOT's answer for now. The longer-term answer will need rules that make large-load requests more binding, through study fees, financial commitments and disclosure of duplicate requests. Until then, the grid will plan around a number it has deliberately cut in half.