A recent BloombergNEF report forecasts that data centers will consume one-fifth of all electricity generated in the United States by 2035, representing a fourfold increase from current levels.
This significant growth is attributed to a surge in AI computing, which is projected to expand data center capacity to almost 200 gigawatts within the next ten years. The report indicates that nearly half of this capacity will be dedicated to AI training and inference, with the majority of this infrastructure remaining concentrated within the U.S. Specifically, by 2033, the United States is anticipated to host 64% of the world's AI chips based on power demand.
It's worth noting that, when compared to prior projections, these figures from BloombergNEF could potentially be conservative.
Indeed, BloombergNEF's updated estimate for 2035 electricity demand marks an 83% increase over its December prediction. This trend is mirrored by other industry bodies; EPRI, an electrical industry nonprofit, has more than doubled its 2024 estimate, while S&P's forecast saw an increase of over one-third between October and April. These significant revisions underscore the intense and rapid pace of data center expansion throughout the U.S.
Looking ahead to the next decade, BloombergNEF anticipates that the bulk of new data centers will connect to electrical grids already under considerable strain. For instance, the PJM Interconnection, serving a region from Virginia to Illinois, is projected to allocate 34% of its electricity to data centers, while ERCOT, which manages most of Texas's grid, will need to dedicate 22% of its generating capacity to meet this demand.
PJM, a region already home to a substantial proportion of the nation's data centers, has faced considerable challenges in managing connection requests from both large-scale generators and major load consumers. This pressure led the interconnection to halt applications for new energy sources to connect to the grid for four years, creating a precarious situation as electricity demand continued its upward trajectory.
Although PJM reopened its queue for new generating sources in April, the severity of the situation has prompted one utility, American Electric Power, to threaten withdrawal from the interconnection. This significant supply-demand imbalance has contributed to a 76% increase in electricity prices over the past year.
Despite the existing congestion, data centers continue to seek connections within the PJM system, evidenced by their accounting for 38% of the charges in the grid manager's most recent capacity auction.
While the U.S. is poised to dominate AI computing, data center expansion is not confined to its borders. Should AI adoption proceed along an aggressive path, global data centers are projected to generate an additional 1,935 terawatt-hours of electricity demand worldwide by 2033, a figure nearly equivalent to India's current annual electricity consumption.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.