Data centers will demand roughly four times more electricity by 2035 than they do today, consuming approximately one-fifth of all electricity generated in the United States. This staggering projection comes from BloombergNEF, which predicts data center capacity will surge to nearly 200 gigawatts over the next decade, driven primarily by artificial intelligence (AI) compute demands.

What makes this forecast particularly striking is how quickly estimates are being revised upward. BloombergNEF's 2035 electricity demand projection is 83 percent higher than what the same consultancy predicted just eight months earlier in December 2025. Other major forecasting organizations are following suit. The Electric Power Research Institute (EPRI), a nonprofit focused on electrical industry research, has more than doubled its 2024 estimate, while S&P Global's forecast rose by more than one-third between October 2025 and April 2026.

Why Are Data Center Power Needs Exploding?

The primary driver behind this explosive growth is the race to build AI infrastructure. Nearly half of the projected 200-gigawatt capacity will be devoted to training and inference, the two most computationally intensive AI tasks. Most of this capacity will remain concentrated in the United States, which is expected to host 64 percent of AI chips by power demand by 2033.

The speed of data center development has caught many regional power grids off guard. The PJM Interconnection, which spans from Virginia to Illinois and already hosts a large concentration of the country's data centers, will see 34 percent of its electricity go to data centers by 2035. Meanwhile, ERCOT, which covers most of Texas, will need to devote 22 percent of its generating capacity to data centers.

What Challenges Are Regional Power Grids Facing?

The supply-demand imbalance is creating real strain on electricity infrastructure. PJM, already struggling to manage connection requests from both large power generators and large energy consumers, paused applications for new grid connections for four years. Though the interconnection reopened its queue to new generating sources in April 2026, the situation has become so dire that American Electric Power, a major utility, has threatened to withdraw from the interconnection entirely.

The financial impact is already visible. Electricity prices in the PJM region have surged 76 percent over the past year, reflecting the intense competition for limited grid capacity. Despite this congestion and cost increase, data centers continue to prioritize connections to PJM. In the grid manager's most recent capacity auction, data centers represented 38 percent of all charges, underscoring their outsized demand.

How to Understand Data Center Power Demand Growth

• Geographic Concentration: The United States will host the majority of AI compute infrastructure, with 64 percent of AI chips by power demand concentrated domestically by 2033, creating regional bottlenecks in already-strained grids.
• Rapid Forecast Revisions: Electricity demand projections have been revised upward by 83 percent in less than a year, suggesting the pace of data center development is accelerating faster than experts anticipated.
• Grid Capacity Constraints: Major interconnections like PJM and ERCOT will dedicate one-third to one-fifth of their total generating capacity to data centers, leaving limited room for other electricity consumers.

The global implications are equally significant. If AI adoption continues along an aggressive trajectory, data centers will create approximately 1,935 terawatt-hours of new electricity demand worldwide by 2033, nearly equivalent to the total annual electricity consumption of India.

These projections underscore a fundamental tension in the AI boom. While artificial intelligence promises transformative benefits across industries, the infrastructure required to power it is straining electricity grids that were designed for a different era. Policymakers, utilities, and technology companies are now racing to expand generation capacity, upgrade transmission infrastructure, and develop new power sources to meet this unprecedented demand.