What you need to know
For most of the 2010s, U.S. energy use remained fairly flat after growing for decades. A recent spike in energy consumption since 2017 (discounting a COVID-era brief decline) has led many experts to speculate that the cause of this increase is the proliferation of powerful but energy-intensive Artificial Intelligence (AI) computing technologies. Everything Policy researchers dug in to learn more about what’s happening and what to expect in the coming decade.
- Usage of Artificial Intelligence (AI) computing technologies is rapidly expanding.
- Is this trend driving increased power consumption by data centers?
- What are the implications of this trend for energy usage in the near future?
AI and Electricity Usage
As shown in the figure below, energy usage in the U.S. has begun to trend upwards after being largely flat for several years.
While the use of AI has been increasing in recent years, there have also been more electric vehicles on the road and a boom in cryptocurrency mining, both of which could be driving the spike in energy use. To parse out AI’s independent role in energy consumption, experts typically point to data center energy consumption as a direct proxy for AI, as these expansive commercial complexes are where AI technology is being implemented on a large scale.
AI uses energy in two ways: processing needs and cooling. On the front end, AI uses large amounts of processing power. That processing also creates large amounts of heat, which for most data centers today, requires electricity for cooling.
Researchers from the Berkeley Lab’s Energy Analysis and Environmental Impacts Division note that energy use at U.S. data centers was steady at about 60 TwH from 2014-2016 but grew to over 76 TwH by 2018, representing about 1.9% of total U.S. electricity consumption. By 2023, that figure increased to 176 TWh, about 4% of total U.S. electricity consumption. By 2028, they estimate that data centers will consume 7-12% of U.S. electricity.
Internationally, researchers at MIT estimate that data centers’ share of global energy consumption could grow from 1-2% currently to as high as 21% by 2030. Similarly, the International Energy Agency estimated back in 2022 that data centers’ global electricity use will double by 2026.
The Take Away
Electricity use by data centers has been growing in recent years, and experts expect that growth to skyrocket over the next decade.
The increased use of AI technologies by individuals, corporations, and governments appears to contribute to increased power consumption by data centers.
Enjoying this content? Support our mission through financial support.
Further reading
Penn State Institute of Energy and the Environment. 2025. Why AI uses so much energy—and what we can do about it. Available at https://tinyurl.com/sxwek9m5. Accessed 4/11/25.
Stackpole, B. 2025. AI has high data center energy costs: But there are solutions. MIT blog post, available at https://tinyurl.com/4ujanbma. Accessed 4/10/25.
Sources
International Energy Agency. 2024. Electricity 2024: Analysis and forecast to 2026. Available at https://tinyurl.com/y9tvmetu. Accessed 4/11/25.
Penn State Institute of Energy and the Environment. 2025. Why AI uses so much energy—and what we can do about it. Available at https://tinyurl.com/sxwek9m5. Accessed 4/11/25.
Shehabi, A., et al. 2024. United States data center energy usage report. Berkeley Lab Energy Analysis and Environment Impacts Division Report. Available at https://tinyurl.com/mtm545j2. Accessed 4/11/25.
Stackpole, B. 2025. AI has high data center energy costs: But there are solutions. MIT blog post, available at https://tinyurl.com/4ujanbma. Accessed 4/10/25.
U.S. Energy Information Administration. 2025. March 2025 monthly energy report. Available at https://tinyurl.com/tm754cu2. Accessed 4/10.25.
Contributors
John Arnold (Intern) Is a sophomore at Binghamton University majoring in Political Science and Economics
Robert Holahan (Content Lead) is Associate Professor of Political Science and Faculty-in-Residence of the Dickinson Research Team (DiRT) at Binghamton University (SUNY). He holds a PhD in Political Science from Indiana University where his advisor was Nobel Laureate Elinor Ostrom. His research focuses on natural resource policy, particularly domestic oil and gas production, and extends into international environmental policy. He was PI on a National Science Foundation grant that utilized a 3000-person mail-based survey, several internet-based surveys, and a series of laboratory economics experiments to analyze Americans’ perspectives on energy production issues like oil drilling and wind farm development.
William Bianco is Professor of Political Science at Indiana University and Founding Director of the Indiana Political Analytics Workshop. He received his PhD from the University of Rochester. His teaching focuses on first-year students and the Introduction to American Government class, emphasizing quantitative literacy. He is the co-author of American Politics Today, an introductory textbook published by W. W. Norton now in its 8th edition, and authored a second textbook, American Politics: Strategy and Choice. His research program is on American politics, including Trust: Representatives and Constituents and numerous articles. He was also the PI or Co-PI for seven National Science Foundation grants and a current grant from the Russell Sage Foundation on the sources of inequalities in federal COVID assistance programs. His op-eds have been published in the Washington Post, the Indianapolis Star, Newsday, and other venues.
