technology & sustainability•Sun, September 20, 2026•5 min read

The Hidden Water Crisis Behind the AI Data Center Boom

While AI innovation captures headlines, a new report highlights a major, overlooked environmental cost: massive water consumption tied to electricity generation. Experts warn that the surge in data center demand could place significant strain on water-stressed regions.

A New Lens on AI Sustainability

When we talk about the environmental impact of AI, the conversation often centers on carbon emissions and energy efficiency within the data center itself. However, a new analysis from Ceres, titled 'Water Behind the Watts: The Hidden Risk of Powering Data Centers,' argues that we are missing the bigger picture. The most significant water consumption associated with these facilities isn't happening on-site—it is happening at the power plants that keep them running.

The Indirect Cost of Computing

The report focuses on seven states—Virginia, Texas, California, Illinois, Georgia, Ohio, and Arizona—which collectively host roughly half of all U.S. data centers. These regions frequently face water stress or drought, making the massive energy requirements of AI infrastructure particularly problematic. Thermoelectric power plants, which supply much of the grid, require significant amounts of water for cooling. As AI demand forces these power plants to work harder, the water 'footprint' of every query, model training run, and cloud storage operation grows.

Projected 400% increase in water use tied to electricity consumption for data centers.
Cooling operations for data centers are expected to see an 870% increase in water demand.
Data centers can increase localized water stress in affected basins by up to 17% annually.
Two-thirds of data centers built since 2022 are located in water-stressed regions.

Why Efficiency is a Moving Target

Data center operators have increasingly turned to air-cooling technology to reduce direct water use. However, this shift creates a paradoxical dilemma: air cooling is typically more energy-intensive than traditional water-based cooling. This increased power demand forces the grid to draw more from power plants, which in turn consumes more water—essentially trading on-site water usage for indirect, grid-level water usage. This 'hidden' demand makes coordinated water-energy planning essential for any future growth in the sector.

Communities and regional water officials must engage in discussions about data centers early on, and with a coordinated, holistic understanding of existing resources and potential impacts on the energy grid and the watershed.
— Mary Ann Dickinson, Lincoln Institute

Key Takeaways

The majority of water usage for data centers occurs off-site at power plants, not through direct cooling inside the facility.
Data center growth is currently concentrated in water-stressed regions like Arizona and Texas.
Switching to air-cooling can reduce direct water use but may inadvertently increase indirect water consumption by raising energy demand.
Total water use linked to data center electricity is projected to surge by 400% in the coming years.
Integrated planning between water utilities and energy providers is necessary to manage the cumulative impact of these massive infrastructure projects.

FAQ

Do data centers use a lot of water?

Yes. While direct usage is a concern, indirect usage—the water required by power plants to generate the electricity used by data centers—is significantly higher.

Why is AI specifically linked to this water usage?

AI requires high-density computing power, which leads to massive increases in electricity demand. Because power plants rely on water for cooling, higher energy consumption translates directly to higher water consumption.

Can data centers use less water?

Yes, through technologies like air cooling. However, these systems often require more electricity, potentially offsetting water savings depending on how the local grid generates power.

Are these water impacts felt everywhere?

The impacts are highly localized. Water stress is most severe in regions where data center clusters are placed in areas already experiencing drought or limited water resources.