Data Centers Expected to Consume Four Times More Electricity by 2035

Data centers are forecast to consume one-fifth of the electricity generated in the United States by 2035, quadruple today's usage, according to a new report from BloombergNEF. The surge in AI compute will push data center capacity to nearly 200 gigawatts over the next decade, with about half of that capacity devoted to training and inference. Most of this will remain concentrated in the U.S.

By 2033, the U.S. will host 64% of AI chips measured by power demand. BloombergNEF's latest estimate is 83% higher than what it predicted in December. Other organizations have also revised their projections upward: EPRI has more than doubled its 2024 estimate, and S&P raised its forecast by over a third between October and April. These revisions reflect the rapid pace of data center construction across the country.

In the coming decade, BloombergNEF expects most new data centers to connect to already strained electrical grids. The PJM Interconnection, spanning from Virginia to Illinois, will see 34% of its electricity go to data centers, while ERCOT, covering most of Texas, will need to devote 22% of its generating capacity. PJM, which already hosts many data centers, has struggled with connection requests and paused applications for four years, only reopening the queue to new generation sources in April. The situation has become so critical that one utility, American Electric Power, has threatened to exit the interconnection.

The supply-demand imbalance has driven electricity prices up 76% over the past year. Despite congestion, data centers still want to connect to PJM—they represented 38% of charges in the grid manager's most recent capacity auction. While the U.S. claims the majority of AI compute, data centers will continue to grow elsewhere. By 2033, if AI adoption follows an aggressive trajectory, data centers will create 1,935 terawatt-hours of new electricity demand worldwide, nearly as much as India consumes annually.