AI Data Centers Put Local Water and Power Under Strain

U.S. data centers could use as much as 110 billion gallons of water a year by 2030, equal to 7% of annual consumptive use in public water systems, according to a March analysis by scientists from the University of California, Riverside and other institutions. AI growth is driving up electricity and water demand. Residents and public officials are asking who benefits and who pays the local costs.

Resource Costs Close to Home

A separate report from nonprofit advocacy organization Ceres found that data centers across seven states used around 3.4 trillion gallons of freshwater a year to generate electricity. That’s roughly 12 times the combined annual water use of Los Angeles, Phoenix and Washington, D.C. Around two-thirds of U.S. data centers built or under construction since 2022 sit in water-stressed areas, according to a 2025 Bloomberg analysis.

Location, electricity sources and technology all affect the impact. Precise estimates are difficult because AI providers share limited information. One oft-cited estimate says generating a 100-word email with an AI chatbot takes about one bottle of water, just over two cups; OpenAI CEO Sam Altman has said an average ChatGPT query uses about one fifteenth of a teaspoon. Technology companies say they’re developing cooling systems that use less water, while some analysts argue that concerns have been overstated by comparing data-center demand with the water used to produce a burger or cotton shirt.

AI data-center electricity consumption surged 50% last year, according to the International Energy Agency. In 2025, AI data centers used around 0.5% of global electricity. Our World in Data projected that share could reach 1.5% by 2030. Nearly half of global data-center electricity consumption happens in the United States. Lawrence Berkeley National Laboratory said data centers could account for up to 12% of U.S. electricity consumption by 2028.

Power Demand and Public Financing

The United States is seeking to restart three shuttered nuclear plants to meet rising electricity demand, including demand from large-scale data centers. The Department of Energy is lending as much as $4.4 billion for the Palisades Power Plant in Michigan, the Crane Clean Energy Center at Pennsylvania's Three Mile Island and the Duane Arnold Energy Center in Iowa. Together, the plants would restore 2.25 gigawatts of electricity, enough to power roughly 2 million homes and businesses.

Two plants have contracts with technology companies willing to pay a premium for large amounts of carbon-free electricity for data centers. NextEra Energy and Google are working together to restart Duane Arnold by early 2029. NextEra CEO John Ketchum said, “This is a perfect example of data centers paying their own way,” and said a long-term contract with a hyperscaler means residential customers in Iowa aren’t paying the project's costs.

Constellation Energy announced a $1.6 billion plan to restart the functioning reactor at Three Mile Island under a 20-year agreement with Microsoft. The Energy Department provided a $1 billion loan, and Constellation hopes to bring the reactor online in 2027. The Palisades project also received federal support: the Biden administration approved a $1.5 billion loan in 2024, and the Trump administration continued payments. Holtec began loading nuclear fuel in August but paused after a fuel assembly tilted from its upright position.

Local Effects and Oversight

Data centers can bring noise from cooling systems, gas turbines, diesel generators and other equipment. In Vineland, New Jersey, residents near a large data center have filed noise-pollution lawsuits. Arizona State University scientists found that data-center heat pollution warmed downwind Phoenix neighborhoods by as much as 4 degrees Fahrenheit. Lead author David Sailor said an added heat-island magnitude of 1 or 2 degrees “can still have a very significant impact on our lives.”

Before building a new data center, officials should establish “who benefits,” said Kaveh Madani, director of the United Nations University Institute for Water, Environment and Health. They should consider whether it creates jobs or improves the local economy, and whether it takes water from users who need it more, such as farmers. “There are trade-offs like any other technology,” Madani said. “It has has downsides that need to be regulated, monitored, and mitigated.”