A new peer-reviewed study published November 16, 2026 in Nature Sustainability finds that the environmental impact of U.S. hyperscale data centers has reached a level comparable to the annual carbon emissions of entire small European countries. Researchers led by Gianluca Guidi and Francesca Dominici analyzed data from 403 of the largest data centers across the United States, revealing a carbon footprint of 25.7 million metric tons of CO₂ for the year from May 2024 to April 2025.

Electricity consumption by these hyperscale data centers accounted for approximately 82 terawatt-hours (TWh) during the study year—roughly 1.8% of all U.S. electricity usage. This volume of energy rivals the annual consumption of entire states including Indiana, Michigan, and Tennessee. According to the analysis, more than half of this power demand was concentrated in Virginia, Oregon, Ohio, and Iowa, reflecting the regional clustering of the data center industry. Yet the carbon intensity of operations varied sharply: while data centers in Oregon benefited from cleaner energy mixes, facilities in Wyoming produced up to seven times more emissions per kilowatt-hour consumed.

The emissions figure of 25.7 million metric tons of CO₂ is comparable to the output from five to six million U.S. gasoline cars in a year, and approaches the annual national emissions of Denmark, according to the researchers. This comparison has circulated in public debate, though exact equivalence fluctuates depending on methodology and national statistics.

The study’s breakdown of fuel sources highlights an area of concern for climate efforts. U.S. hyperscale data centers drew 54% of their power from fossil fuels, with renewables accounting for 25% and nuclear energy 21%. The significant share of fossil-fueled electricity underscores both a reliance on legacy grids and a challenge for rapid decarbonization of the digital backbone.

This surge in electricity demand has been propelled by booming cloud services and, increasingly, artificial intelligence (AI) applications. Indeed, the study reflects a doubling down on infrastructure investment post-2022, when the popularity of generative AI models such as ChatGPT and large-scale cloud computing drove demand upward. According to related research cited in Patterns (2025), AI systems alone could account for a carbon footprint equivalent to that of New York City by 2025, though precise attribution remains difficult due to a lack of industry transparency over workload-specific emissions.

Authors of the Nature Sustainability study emphasize the policy implications of their findings. They advocate for stricter transparency and disclosure requirements on the part of data center operators—long criticized for opaque environmental reporting—and urge governments and industry to consider the full consequences of rapid digital expansion. Since the study window, legislative actions have been debated in some regions, including possible restrictions or moratoriums on new data center builds, especially where local electric grids are reliant on fossil fuels.

Meanwhile, the industry is responding with initial efforts to retrofit or repurpose older telecommunications facilities for data processing, aiming to reduce incremental environmental cost. However, the trajectory of American data center energy use and climate impact remains dependent on choices being made today: how fast operators transition to renewable sources, how aggressively tech companies disclose their carbon metrics, and how state and federal authorities set and enforce standards in this fast-moving technological landscape.