Particle.news

Harvard Study Links AI Data‑Centre Build‑Out to Large Power Use and Rising Emissions

The peer‑reviewed findings show concentrated U.S. hyperscale facilities used vast amounts of electricity, risking grid strain and higher carbon output unless power systems and sourcing change.

Overview

  • A Nature Sustainability paper released this week found 403 U.S. hyperscale data centres used about 82 terawatt‑hours of electricity between May 2024 and April 2025, roughly 2 percent of U.S. power supply.
  • Researchers estimated those facilities produced about 25.7 million tonnes of CO2 during the year measured because roughly 54 percent of their electricity came from fossil fuels.
  • Lead author Gianluca Guidi told CBS News that U.S. hyperscale capacity has more than doubled since the study period and that roughly 2,000 additional facilities are planned or under development.
  • The International Energy Agency projects global data‑centre electricity use could double to about 1,000 TWh by 2030 and that related emissions could rise roughly 70 percent if incremental demand is met largely with fossil fuels.
  • The expansion is driving large private investment and creating grid connection and supply‑chain bottlenecks, prompting state policy moves such as moratorium talks in Oklahoma, local complaints about higher bills and water use, and a COP31 voluntary pledge that major U.S. hyperscalers have not yet signed.