The Hidden Water Cost of Powering Data Centers

Published September 7, 2026

Data centers have become one of the fastest growing consumers of electricity in the United States, and that growth is now surfacing a resource question that has received far less attention than emissions or grid capacity: water. Most conversations about data center sustainability focus on the water used on-site to cool servers. A new report from the sustainability nonprofit Ceres shows that this framing misses most of the picture. The water consumed to generate the electricity that powers data centers is often several times larger than the water used for direct cooling, and it is concentrated in regions that are already under water stress. As artificial intelligence workloads accelerate electricity demand, water is emerging as a material risk for utilities, regulators, and the companies that operate or rely on data center infrastructure.

Data Centers’ Water Footprint Extends Far Beyond Cooling

Ceres’ report, “Water Behind the Watts: The Hidden Risk of Powering Data Centers,” released in August 2026, examined data center activity across seven states with heavy concentrations of facilities: Virginia, Texas, California, Illinois, Georgia, Ohio, and Arizona. The report found that data centers in these states indirectly consume approximately 3.4 trillion gallons of freshwater annually through the electricity they draw from the grid. That figure is projected to rise to between 4.1 and 7.6 trillion gallons by 2030 as new facilities come online to support cloud computing and AI training and inference. By that year, electricity-related water use could account for as much as 72 percent of a data center’s total water footprint, well above the water used for on-site cooling towers and chillers. The scale becomes clearer at the state level. Virginia’s data centers, the report found, use roughly 21 times the water that Washington, D.C. consumes in a year, while California’s data centers use about 10 times the annual water consumption of Los Angeles. These comparisons illustrate why water, historically treated as a facilities-level operating cost, is increasingly a regional infrastructure concern tied directly to how electricity is produced.

Electricity Generation Is the Larger, Hidden Water Draw

The reason electricity carries such a large water footprint comes down to how power plants generate steam and manage heat. According to the Ceres report, 78 percent of the electricity supplying data centers in the states studied comes from thermoelectric sources such as natural gas, coal, and nuclear plants, all of which require water for cooling and steam condensation. Only 22 percent comes from water-efficient renewables like wind and solar, which need little to no water to operate. Research from Lawrence Berkeley National Laboratory, cited in a July 2026 report from the Information Technology and Innovation Foundation, puts a sharper point on the gap between direct and indirect use. In 2023, U.S. data centers consumed an estimated 17.4 billion gallons of water directly for cooling, while the electricity used to power them required roughly 211 billion gallons, about 12 times as much. That gap varies by region depending on the generation mix: water consumption runs about 2.1 gallons per kilowatt-hour in the hydropower-heavy Pacific Northwest, compared with roughly 0.13 gallons per kilowatt-hour in solar-heavy California. This matters because data center electricity demand itself is rising quickly. The International Energy Agency estimates that U.S. data centers consumed about 183 terawatt-hours of electricity in 2024, more than 4 percent of national electricity consumption, and projects that figure could reach 426 terawatt-hours by 2030. In Virginia, data centers already draw roughly 26 percent of the state’s total electricity supply. As that demand grows, so does the water embedded in meeting it, particularly in states still dependent on thermoelectric generation.

Regional Water Stress Raises the Stakes for Companies and Communities

The Ceres report found that 66 percent of the electricity powering data centers in the seven states studied comes from regions already classified as water-stressed. That overlap between electricity demand growth and water scarcity is the central risk the report highlights. Water stress affects agriculture, municipal supply, and industrial users in the same watersheds that host data centers and the power plants serving them, meaning competition for water can intensify during droughts or peak demand periods. Kirsten James, Ceres’ senior program director, noted that better planning requires understanding the full scope of a data center’s water dependency, including the electricity side of the equation, rather than looking only at what happens inside the building. For companies with data center operations, colocation contracts, or significant reliance on cloud infrastructure, this creates a practical planning problem. Facility-level water metering alone does not capture exposure to electricity-driven water risk, and site selection decisions made without visibility into a region’s generation mix and water stress profile could carry costs that only become apparent later, through rate increases, supply constraints, or permitting delays. Some studies, including analysis referenced by Carnegie Mellon University researchers, estimate data center growth could raise average U.S. electricity bills by around 8 percent by 2030, with increases exceeding 25 percent in high-demand markets such as northern Virginia. Researchers at North Carolina State University reported in September 2026 that planned natural gas capacity dedicated to U.S. data centers had already surpassed 189 gigawatts, as tech companies plan to spend more than 700 billion dollars on AI infrastructure this year alone. Water and electricity costs are increasingly linked outcomes of the same underlying growth trend.

Conclusion

The Ceres findings point to a broader lesson for any organization managing energy and sustainability data: water and electricity are not separate line items to track independently, they are connected resources whose combined footprint needs to be measured together to get an accurate picture of environmental impact and operating risk. A data center’s water use on paper may look modest, but its indirect water footprint through electricity consumption can be several times larger, especially in regions still reliant on thermoelectric generation. As AI-driven electricity demand continues to grow, and as more of that demand lands in water-stressed regions, companies that track energy consumption without accounting for its water intensity are working with an incomplete risk picture. Building water metrics into energy and emissions reporting, alongside regional water stress data, gives organizations a clearer view of where future costs, regulatory attention, and community pushback are most likely to emerge. For any company evaluating data center partners, cloud providers, or facility locations, water management is becoming as relevant to due diligence as carbon accounting has already become.