A study led by the University of California, Riverside in collaboration with the California Institute of Technology (Caltech) highlights that growing water demand from US data centers poses significant infrastructure challenges. Cooling the millions of servers that power AI queries and other computations is water-intensive, and the researchers say this factor is often overlooked in planning and regulation.
Data and methods behind the findings
The authors analyzed a large set of publicly available sources, including government registries and utility water databases. From these data they estimate that modern, large data centers using evaporative (water) cooling commonly exceed peak daily water use of 3.78 million liters, and that some planned facilities could demand more than 25 million liters per day at peak.
National projection: how much additional capacity might be needed?
The study projects that if current water-use intensities persist, US data centers will require an additional 2.6 to 5.5 billion liters per day of peak water capacity by 2030. The researchers note that this magnitude is comparable to the daily water consumption of New York City, and even optimistic reductions would leave the aggregated new demand near roughly half of New York City’s daily needs for much of the year.
Why peak demand matters separately from annual totals
Municipal water systems are generally designed to meet maximum instantaneous demand, so data centers’ peak water use is a critical factor for infrastructure design, resilience and operational reliability. However, the authors point out that most operators disclose only annual water consumption, while omitting data about peak-period demands.
Risks from insufficient water supply
Insufficient water for cooling can directly increase the risk of system failures, outages and even fires in data centers. When water is unavailable, facilities often must switch to dry (air-based) cooling, which is less efficient and increases electrical power demand — placing further stress on the grid, especially during summer peaks.
Proposed responses and partnerships
The researchers outline several approaches to address rising water demand:
- Increase transparency: data centers should report not only annual water use but also peak-period demands.
- Finance and infrastructure partnerships: develop corporate and community partnerships to fund and build water infrastructure so that costs are not borne solely by local residents.
- Flexible cooling strategies: coordinate with utilities to switch cooling modes — using water-based cooling when the electric grid is constrained and switching to dry cooling when community water systems face high pressure. Such switching can employ fans or vapor-compression technologies similar to building air conditioners, though these options typically require substantially more electricity.
Scientific status and transparency
The results are available as a preprint on arXiv and have not yet undergone formal peer review. Nevertheless, the paper contributes to public debate and may prompt reconsideration of planning practices across technological and utility sectors.
Implications for the technology sector
The study warns that rising peak water needs can directly affect the feasibility of data center projects: increasing costs, causing delays or leading to cancellations. Shifting to dry cooling increases electricity demand and grid stress, creating additional challenges for planners and operators during the energy and climate transition.
In sum, the research argues that water is not merely a supporting resource for data centers but a strategic one that needs transparent reporting and coordinated planning to ensure reliable, sustainable operation.
