Fossil Fuels Supply 54% of Power to US Hyperscale Data Centers
A Harvard-led study in Nature Sustainability estimates 403 US hyperscale data centers used 82 TWh of power, with fossil fuels supplying 54% and CO2 emissions near 25.7 million metric tons.
By Grace Kim
3 min read

What's News
- Fossil fuels supplied roughly 54% of electricity attributed to US hyperscale data centers — 38% natural gas, 15% coal.
- 403 facilities consumed about 82 TWh in the central scenario, ~1.8% of US electricity use.
- Attributable CO2 emissions were estimated at 25.7 million metric tons (range 21–31 million).
- Virginia, Oregon, Ohio, and Iowa accounted for more than half of the electricity examined.
- The study, published November 16, 2026, could not separate AI-specific power use from general computing.
Fossil fuels supplied roughly 54% of the electricity attributed to US hyperscale data centers, according to a Harvard-led study published November 16, 2026 in Nature Sustainability. Natural gas accounted for 38% and coal for 15% of that share.
The study estimates that 403 US hyperscale data centers consumed approximately 82 terawatt-hours (TWh) of electricity under its central scenario — about 1.8% of total US electricity consumption. Attributable carbon dioxide emissions came in at roughly 25.7 million metric tons, within a scenario range of 21–31 million metric tons.
Across facility-load scenarios, estimated consumption spanned approximately 68–99 TWh. Harvard noted the upper estimate approaches the annual electricity use of states such as Indiana, Michigan, and Tennessee.
How did researchers build the numbers?
The estimates cover May 2024 through April 2025. The team used satellite imagery to compile the facility list, then modeled annual electricity consumption for each site under a range of operating scenarios. EPA data allowed them to attribute emissions based on the mix of power plants in each facility's regional grid.
The paper calls its method a location-based, total-output balancing-authority attribution approach. In practice, it allocates generation and emissions from plants in the relevant regional grid to the facilities. That produces a facility-resolved estimate of electricity sources and emissions — not direct meter readings from every site.
In short: this is modeled attribution, not measured consumption at the meter.
Why does the regional grid matter?
Carbon intensity varied sharply by region. On the paper's electricity-weighted, total-output measure, data centers ran at approximately 314 grams of CO2 per kilowatt-hour, below the US total-output average of roughly 348 grams.
The aggregate hides wide spread. About a quarter of the electricity was consumed in grid regions with carbon intensity above the national average, and many facilities clustered where coal-fired generation contributes heavily to supply.
Virginia, Oregon, Ohio, and Iowa together accounted for more than half of the electricity use examined. The takeaway in plain words: a national average cannot represent every campus, because emissions attributed to a facility depend on the generation mix of its regional grid.
Can the totals be read as AI's footprint?
No — and the researchers are explicit about it. Available data did not cover every US hyperscale data center, and the team could not separate electricity used for AI-specialized operations from general-purpose computing at individual facilities. The sites support cloud computing, large language model training, and global deployment of AI tools, so the totals cannot be interpreted as AI-only consumption.
Lead author Gianluca Guidi said in Harvard's announcement that the reported electricity use and carbon footprint "could already be substantially larger now in 2026." That projection reflects facilities built and campuses expanded after the study period, plus continued growth in AI computing demand — not a new measurement.
What's next for the research?
The study did not directly measure health effects, which the researchers plan to investigate in future work. Corresponding author Francesca Dominici laid out two pathways: carbon dioxide drives climate change, affecting health through heat waves, wildfire smoke, flooding, and other channels; and fossil-fuel plants release fine particulate matter, nitrogen oxides, and sulfur dioxide linked to heart and lung disease and premature death.
That second pathway has a siting implication. People living near and downwind of those plants may bear health costs even when they live far from the data center itself.
Dominici said the goal is to give the public, policymakers, investors, and developers a more concrete basis for decisions about where data centers go and how they are powered. The framework pushes that accounting past a single national total, down to regional grids and their power plants — the level where hyperscale siting, electricity supply, and attributable emissions actually meet.
Source: unite.ai
More from Grace Kim
Show full bio
Market editor covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.
2,178 articles