US hyperscale data centers emit as much as Denmark: Report
Reported by GeekWire, Independent Journal Review, The Washington Post and 19 more outlets22 sources ↓
A Harvard-led study estimates that 403 U.S. hyperscale data centers emitted 25.7 million metric tons of CO₂ from May 2024 to April 2025, more than Denmark. Their estimated electricity use was 68–99 terawatt-hours, about 1.8% of U.S. consumption.
Key points
- The study estimates annual electricity use at 68–99 terawatt-hours, with a central estimate of 82 terawatt-hours
- About 54% of the electricity was generated from fossil fuels, mainly natural gas and coal
- Virginia had the most facilities in the study, with 142; Oregon had 56
- The estimates cover May 2024 to April 2025 and exclude facilities the researchers could not verify
Sources
- GeekWireRead ↗Study: 403 U.S. hyperscale data centers rival Denmark’s emissions; Oregon ranked No. 2 in power use
- Independent Journal ReviewRead ↗Harvard study estimates US data centers emitted 25.7 million tonnes of CO2
- The Washington PostRead ↗Largest data centers take up about 2 percent of U.S. electricity use, study says
- letsdatascienceRead ↗Harvard study estimates 403 US hyperscale data centers approach Michigan's electricity use
- Hacker NewsRead ↗Just 400 US Data Centers Use More Electricity Than Michigan
Details
Researchers identified 403 facilities from 675 candidate sites using satellite imagery and building-footprint information. Because operators do not generally disclose facility-level electricity use, the team estimated consumption from site capacity and assumed utilization, then applied regional power-grid emissions data.
The Decoder reports that natural gas accounted for 38% of the electricity attributed to the facilities and coal for 15%. The study’s estimates varied with assumed utilization; the central estimate was about 82 terawatt-hours. Tom’s Hardware reports that Virginia accounted for 21.4 terawatt-hours, while Oregon accounted for 11.1 terawatt-hours.
The estimates do not capture all data centers or necessarily reflect current conditions: the observation period ended in April 2025, and the study was limited to facilities researchers could verify. Tom’s Hardware also notes that the method used regional grid averages and did not account for clean-power contracts or credits; on-site generation may also be missing from grid data.
Written from coverage as of
Images
- The Verge
- CNET
- Tom's Hardware
- The Decoder