AI Data Centers Threaten US Water Supply, Study Warns

A UC Riverside study finds AI data centers could demand NYC-scale water volumes daily by 2030, costing up to $58B

13 Mar 2026

Rapid expansion of artificial intelligence data centres could place significant strain on US water systems, according to new research from the University of California, Riverside and the California Institute of Technology.

The study, released in early March, estimates that by 2030 data centres may require between 697m and 1.45bn gallons of additional peak water capacity each day, a volume comparable to the daily supply used by New York City.

Researchers say the risk has received limited attention because operators typically report only annual water consumption. That measure can mask sharp spikes in demand during periods of high temperatures, when cooling systems rely heavily on evaporative processes.

On peak summer days, such systems can require six to ten times a facility’s average daily water use. A single large data centre can consume more than 1m gallons of water on its hottest day, the report says. Some projects currently under construction have secured water allocation agreements of up to 8m gallons a day, a level comparable to the needs of several small towns.

Meeting that demand could require substantial investment in new water infrastructure, including treatment plants, reservoirs, transmission pipelines and pump stations. The study estimates the cost at between $10bn and $58bn, depending on how quickly the sector expands.

Geography may compound the challenge. Roughly 40 per cent of US data centres are located in regions already experiencing water stress, concentrating new demand in areas where supplies are limited and infrastructure is often under strain.

US water utilities already face mounting pressure from ageing networks and constrained budgets. The Environmental Protection Agency has previously warned of a multi-trillion-dollar funding gap across the country’s drinking water and wastewater systems.

The study’s authors urge greater transparency from data centre operators on peak water usage, as well as closer partnerships with local utilities to help finance infrastructure upgrades. They also recommend more flexible cooling strategies that reduce demand during periods of peak system stress.

Several states, including Pennsylvania, Georgia and Oklahoma, have begun considering legislative and budget measures aimed at managing the growing water requirements of the data centre industry. Policymakers and utilities are now weighing how to accommodate the sector’s expansion while safeguarding local water supplies.