Bernstein raises U.S. power-demand growth forecast to 4% through 2030 on AI data-center buildout

Bernstein Societe Generale Group has increased its long-term U.S. electricity-demand growth forecast to a 4% CAGR through 2030, versus its previous 3% forecast. The firm attributes the acceleration primarily to AI data-center construction.

Bernstein estimates that U.S. electricity demand will add roughly 1,000 TWh between 2025 and 2030, with data centers expected to account for more than half of that incremental load. Under its baseline scenario, which assumes a PUE of 1.25, U.S. data-center IT load is projected to increase from 34 GW in 2025 to 149 GW by 2030—an addition of 115 GW. The model translates this expansion into more than 500 TWh of new annual electricity consumption from digital facilities.

To meet the projected demand while offsetting planned coal retirements, Bernstein models gas-fired and solar PV generation as the main sources of incremental supply. It forecasts gas-fired generation growing at a 2.9% CAGR through 2030 and solar generation at a 21% CAGR. The model also incorporates the potential restarts of Three Mile Island, Palisades and Duane Arnold; their inclusion in the scenario does not confirm that these facilities are operational.

For data-center operators, hyperscalers and power-market participants, the outlook points to grid interconnection, available generation and contracted energy supply becoming increasingly material to capacity deployment. If Bernstein’s projections are realized, the scale of expected IT-load growth could increase pressure on transmission and generation development, while supporting demand for new gas, solar and potentially returning nuclear capacity.

The projection remains a forward-looking analytical scenario rather than a measure of commissioned data-center capacity or actual electricity consumption.