AI data centers face U.S. power bottleneck as electricity demand surges

  • OpenAI, Google and Meta are expanding AI infrastructure faster than U.S. power systems.
  • Data centers could consume nearly 12% of U.S. electricity by 2030.
  • NERC's 2025 assessment projects North American peak demand will add over 224 gigawatts.

OpenAI, Google and Meta are building AI infrastructure faster than the United States can add the generation and transmission capacity needed to power it, creating a timing mismatch between technology companies and utilities. Data centers are projected to consume nearly 12% of all U.S. electricity by 2030, almost six times their share in 2018, according to Lawrence Berkeley National Lab. NERC's 2025 reliability assessment projects North American summer peak electricity demand will rise by more than 224 gigawatts over the next 10 years, 69% above the increase projected a year earlier, driven by new AI data centers. Power access has become a leading constraint on data center development, with new projects seeking electricity faster than the grid can deliver it. A gigawatt can power 750,000 U.S. homes, but projects that began commercial operation in 2025 took a median of five years from requesting a grid connection to operating. Utilities may delay full service, offer interruptible provisional connections, or require developers to provide more financial protection for new generation and grid investments. Marsh executive Kathryn Burke predicts about 50% to 60% of data center projects will be delayed beyond companies' hoped-for one-to-two-year timelines.

当サイトの情報はAIを用いて生成されており、正確性を保証するものではありません。 参考情報としてご活用ください。