Data centers in the United States are projected to consume four times more electricity by 2035 than they do today, accounting for roughly one-fifth of all generated power, according to a new report from BloombergNEF. The surge is primarily driven by the rapid build-out of artificial intelligence compute infrastructure, with total U.S. data center capacity expected to reach nearly 200 gigawatts over the next decade.
The report, released in early 2025, estimates that almost half of the new capacity will be dedicated to training and running AI models. The consultancy now forecasts that U.S. data centers will demand 106 GW by 2035, a 36% jump from its April 2025 projection of 78 GW and a staggering 83% higher than its December 2024 estimate. Other groups are also racing to revise numbers upward: EPRI more than doubled its 2024 outlook, and S&P Global raised its forecast by over a third between October 2024 and April 2025.
The growth is placing unprecedented strain on already congested grids. The PJM Interconnection, covering states from Virginia to Illinois, could see 34% of its electricity consumed by data centers. In Texas, ERCOT will need to dedicate 22% of its generating capacity to these loads. PJM recently reopened its connection queue after a four-year pause, but electricity prices in the region have already surged 76% over the past year. Data centers still represented 38% of charges in the latest capacity auction, demonstrating undeterred demand.
The boom is not confined to the U.S. Globally, aggressive AI adoption could add 1,935 terawatt-hours of new electricity demand by 2033—nearly the annual consumption of India. While the U.S. is expected to host 64% of the world’s AI chips, the report raises concerns about energy security and economic competitiveness.
For the crypto industry, the trend has a direct implication: Bitcoin miners are already repurposing infrastructure to service AI workloads, converting their data centers from proof-of-work to high-performance computing. As grid constraints tighten and electricity costs rise, more mining operations may pivot to AI hosting or face diminishing margins. The report underscores an urgent need for grid investment and policy planning to balance AI’s economic promise with reliable, affordable power.
