BNEF Raises U.S. Data Center Electricity Demand Forecast by 83% to 2035

BloombergNEF (BNEF) raised its electricity demand projection for U.S. data centers in 2035 by 83%, from 106 gigawatts (GW) estimated in December 2025 to 194 GW, according to its report Power for AI: Easier Said Than Built, published on July 21. At this new rate of growth, data centers will consume one in every five kilowatt-hours generated in the country by the middle of the next decade.

The upward revision, the steepest since BNEF began tracking the segment, reflects the pace of artificial intelligence (AI) infrastructure deployment. Amazon, Google, Microsoft, and Meta are collectively expected to invest $700 billion this year in data centers, according to Latitude Media. The four firms control 42% of installed capacity in the United States, giving them disproportionate influence over electric grid planning. For Latin American markets seeking to attract that same investment (Brazil, Chile, and Mexico leading the way), the scale of U.S. demand intensifies competition for firm electricity, transmission equipment, and financing.

The BloombergNEF report projects that data centers will grow from representing 5.9% of current U.S. electricity consumption to nearly 20% by 2035. Even maintaining the historical peak interconnection rate of 7.1 GW per year, BNEF estimates a 19 GW gap between supply and demand at the close of the period. The forecast incorporates significant execution risks: 15 state legislatures have considered temporary moratoriums on data center development, and New York has already enacted the first statewide one. Mark Daly, a BNEF analyst, acknowledged that the methodology based on the historical construction rate may prove optimistic given local opposition, interconnection bottlenecks, and shortages of skilled labor.

The forecast recalibrates the accounting for transmission and generation projects in Texas (ERCOT), the country's second-largest data center market. ERCOT's quarterly interconnection pipeline update, expected in August, will provide the first measurement of AI demand's impact on the Texas grid, which is interconnected with northeastern Mexico.

This article was written with the assistance of artificial intelligence from verified sources and reviewed by a human editor before publication.