Powering AI: U.S. Electrification in Five Charts

U.S. electricity demand is entering a long-term growth cycle led by AI data centers, creating major investment needs across generation, grids, equipment, and onsite power.

• Total U.S. electricity demand could rise from 4,024 TWh in 2025 to more than 6,200 TWh by 2040, an increase of up to 55%; data centers could represent 20% of consumption by 2035, equivalent to 194 GW versus 47 GW in 2025.
• BloombergNEF raised its forecast for U.S. data-center capacity in 2030 by 52% from its December 2025 estimate, to 118 GW; capacity could reach 194 GW by 2035, while the PJM region could add 72.5 GW through 2035.
• U.S. utilities plan to spend at least $1.4 trillion between 2026 and 2030 on generation and transmission and distribution infrastructure, compared with $870 billion in the preceding five years; nearly half of already allocated capital is expected to fund grid infrastructure.
• Grid bottlenecks are intensifying: transformer lead times have reached at least two years, about 55% of U.S. distribution transformers are more than 33 years old, and the average interconnection timeline has risen to five years from 3.5 years in 2020.
• Onsite power could supply roughly one-third of large-load demand by 2030; 61% of data-center developers surveyed would choose onsite generation during a grid bottleneck, while Bloom Energy agreed with Oracle to provide up to 2.8 GW of fuel-cell capacity.

Takeaway: The AI-driven power buildout should support utilities, grid-equipment suppliers, and onsite-generation providers, although equipment shortages, aging infrastructure, and lengthy interconnection queues could constrain the pace of deployment.