U.S. Hyperscale Data Centers Consume Electricity Equivalent to Entire States, Study Finds
Hyperscale data centers (HDCs) across the United States consume as much electricity as entire U.S. states and are predominantly powered by fossil fuels, according to a new study published on November 16 in Nature Sustainability. The research, led by Harvard T.H. Chan School of Public Health and the Harvard Data Science Initiative, provides one of the first facility-level estimates of HDCs' electricity use and carbon dioxide (CO₂) emissions since the significant rise of artificial intelligence (AI).

HDCs, defined as the largest and most powerful class of data centers, support critical services including cloud computing, large language model training, and the global deployment of AI tools. Gianluca Guidi, the lead author and a postdoctoral fellow in the Department of Biostatistics, highlighted the rapid growth of data centers, largely driven by AI, and the immediate decisions being made about their construction and power sources. Corresponding author Francesca Dominici, a professor of biostatistics and director of the Harvard Data Science Initiative, emphasized the study's goal to provide concrete, up-to-date environmental data for public, policymaker, investor, and developer decisions.
To conduct the study, researchers utilized satellite imagery to identify 403 HDCs across the U.S., estimating their annual electricity consumption between May 2024 and April 2025 under various operating scenarios. Environmental Protection Agency data was then used to attribute carbon emissions based on the regional grid's power plant mix. The findings revealed that these HDCs collectively used between 68 and 99 terawatt-hours (TWh) of electricity. The upper estimate of 99 TWh is comparable to the annual electricity consumption of states such as Indiana, Michigan, and Tennessee.
During the study period, HDC electricity demand constituted approximately 1.8% of total U.S. electricity consumption and was linked to about 25.7 million metric tons of CO₂. This CO₂ figure is two to three times higher than an earlier estimate for 2018. Researchers noted that current electricity demand and carbon emissions are likely substantially higher in 2026, as the data center boom has continued unabated since the study period concluded in April 2025. Roughly 54% of the electricity powering the HDCs was generated by fossil fuels, with natural gas accounting for 38% and coal for 15%. Approximately a quarter of the electricity came from grid regions dirtier than the national average, with many facilities concentrated in areas heavily reliant on coal-fired power plants. Four states—Virginia, Oregon, Ohio, and Iowa—accounted for over half of the studied electricity use.
The study acknowledged limitations, including not covering all U.S. HDCs and an inability to differentiate between general-purpose and AI-specialized power use. Future research plans include quantifying the health impacts associated with HDCs, as fossil fuel power plants release pollutants directly linked to heart and lung disease and premature death, in addition to CO₂'s role in climate change.
What to watch: The continued growth of AI and its impact on data center energy demands and emissions.
Editor's note: Accurately reflects study findings and quotes; concise but includes all major numbers.
AI-generated and fact-checked against the original report; claims the gate cannot verify are held back.