Field report NW-94 / M3 · sheet /macro
Boardroom MacroFiled from the territory
US Hyperscale Data Centers Emitted 25.7 Million Tons of CO2 in a Year
A Harvard study in Nature Sustainability finds 400+ US hyperscale data centers emitted 25.7 million metric tons of CO2 in a year, consuming up to 99 TWh — rivaling Denmark's national emissions.
- Grid
- NW-94 / M3
- Surveyed
- November 20, 2026
- Reading
- 3 min read
- Correspondent
- Olivia Hart

Field notes
- Just over 400 US hyperscale data centers emitted about 25.7 million metric tons of CO2 between May 2024 and April 2025.
- The facilities drew 68–99 terawatt-hours of power, equal to about 1.8% of total US electricity consumption.
- About 54% of the electricity came from fossil fuels: 38% natural gas and 15% coal.
- Virginia, Oregon, Ohio and Iowa together represented a majority of the electricity demand studied.
- The study was published Monday in Nature Sustainability by researchers from Harvard's T.H. Chan School and Data Science Initiative.
Just over 400 hyperscale data centers in the United States produced roughly 25.7 million metric tons of carbon dioxide between May 2024 and April 2025, a Harvard-led study published Monday in Nature Sustainability finds. That total rivals the annual emissions of some U.S. states and entire countries — nearly matching Bolivia or Denmark, according to The New York Times.
The figure runs two to three times higher than earlier estimates for 2018. Researchers from the Harvard T.H. Chan School of Public Health and the Harvard Data Science Initiative say the gap reflects how quickly AI has transformed the sector's footprint. Hyperscale facilities, the largest class of data center, support cloud computing, large language model training and the deployment of AI tools.
How much power does AI infrastructure actually draw?
Depending on operating scenarios, the facilities consumed between 68 and 99 terawatt-hours of electricity over the study period. The upper estimate approaches what Indiana, Michigan or Tennessee each use in a year. Across the national grid, the demand amounted to about 1.8% of total U.S. electricity consumption.
The generation mix matters as much as the volume. About 54% of the electricity powering these facilities came from fossil fuels:
- Natural gas supplied 38%
- Coal supplied 15%
- Roughly a quarter of the electricity came from grid regions dirtier than the national average
Geography concentrated the impact. Virginia, Oregon, Ohio and Iowa together accounted for a majority of the electricity demand the study covered — meaning a handful of state-level permitting and grid decisions shape most of the sector's emissions profile.
How did the researchers measure it?
The team built its facility list from satellite imagery, then used Environmental Protection Agency data to assign emissions to each site according to which power plants feed its regional grid. That bottom-up method avoids relying on operator self-reporting, which has historically understated data center energy use.
Lead author Gianluca Guidi, a postdoctoral fellow in the Department of Biostatistics, framed the findings as a moving target. "Data centers are growing incredibly fast, driven largely by AI, and decisions about where to build them and how to power them are being made right now," he said in a statement.
What has changed since the study window closed?
Guidi cautioned that the numbers already understate the present. "The data center boom hasn't slowed down since our study period ended in April 2025," he said. "That means the electricity use and carbon footprint we're reporting could already be substantially larger now in 2026."
The study carries three explicit limits:
- Available data did not cover every U.S. hyperscale facility
- Researchers could not separate general-purpose computing from AI-specialized operations at each site
- The study did not directly measure health effects
That last gap defines the next phase of the work. Corresponding author Francesca Dominici, director of the Harvard Data Science Initiative, said health impacts are the follow-up step.
What does this mean for siting and power decisions?
Dominici pushed the findings toward policy. "Decisions about where data centers go and how they're powered are critical environmental and public health decisions, and they should be made with good data," she said in a statement.
For operators and grid planners, the study's structure offers a practical lesson: emissions depend heavily on where a facility plugs in, not just on how much it computes. A quarter of the electricity in the study came from dirtier-than-average grid regions, and the four states dominating demand show how concentrated the buildout has become. As Guidi notes, those siting and power choices are being locked in now — with footprints that the 2026 numbers will likely exceed.
via qz.com (Source)
More from Olivia Hart
Show full bio
News editor covering regulation, policy and their impact on corporate strategy and chief-executive leadership at CEO Frontier.
474 articles
Adjacent sheets
- Harvard Study: Hyperscale Data Centers Now Outdraw Entire States
- Data Center Gas Plants Could Add 10% to US Power Emissions
- Data Centers Need Community Stakes and Regulatory Accountability
- Midterm Voters Rejected Data Centers. A Clean Energy Standard Is the Answer, Climate Advisers Says
- AI Capex Nears $700 Billion, and the Cash Burn Worries Investors