- Utah approved the Stratos AI data center in Box Elder County on May 13, 2026, despite the project spanning 40,000 acres and requiring roughly 9 gigawatts of power in a region already under severe drought stress.
- A May 2026 report from Next10 and Santa Clara University found that many generative AI data centers do not publicly disclose water usage, while Google’s data centers consumed 6.1 billion gallons in 2023 and Microsoft projects 18 billion liters annually by 2030.
- California has bills in progress to mandate public reporting of data center water use, and direct-to-chip cooling systems like ZutaCore’s HyperCool are emerging as zero-evaporation alternatives to conventional cooling.
Utah just approved a data center that would require more power than the entire state currently uses, in one of the driest corners of the American West. The Stratos AI project is the sharpest example of a collision playing out across the country: AI infrastructure expanding fast into water-scarce regions, with communities and regulators scrambling to catch up.
Utah’s Data Center Flashpoint
The May 13, 2026 approval of the Stratos AI data center in Box Elder County drew immediate backlash. The project spans over 40,000 acres across three sites and is projected to require roughly 9 gigawatts of power, more than the state’s entire current consumption, in a region already under severe drought stress. The Sierra Club’s Utah chapter publicly condemned the approval, warning that Stratos could damage the Great Salt Lake ecosystem, which is already struggling. The facility’s precise water demand has not been publicly disclosed, and critics say that opacity is itself part of the problem.
The AI Water Bill: Billions of Gallons Consumed
Google‘s data centers consumed 6.1 billion gallons of water in 2023, according to the company. Microsoft projects its data centers will consume 18 billion liters annually by 2030. Those figures cover operational cooling only. The manufacturing side adds more: chip fabrication plants require large volumes of ultrapure water daily, though facility-level figures are rarely disclosed publicly.
Local Communities Push Back on “Thirsty” Tech
A May 19, 2026 report from Next10 and Santa Clara University found that many generative AI data centers do not publicly disclose their water usage, complicating planning for local water authorities. The problem is sharper in drier, rural areas, where AI companies have increasingly chosen to build and where water systems have less capacity to absorb new demand. Opposition to projects like Stratos goes beyond water: local concerns also include rising utility bills and noise pollution.
Regulatory Scrutiny and Calls for Transparency
California has bills in progress that would mandate public reporting of data center water use, and several other states are pursuing similar measures. Internationally, water governance is expected to feature in discussions at the 2026 United Nations Water Conference.
Cooling Innovation: The Push Toward Zero Evaporation
Direct-to-chip liquid cooling is emerging as the most credible technical response. Systems like ZutaCore‘s HyperCool pull heat away at the processor level, enabling high-density racks with no evaporative water loss. On World Water Day, March 22, 2026, Google announced it had replenished over 7 billion gallons of water across 97 watersheds in 2025, part of its stated goal to replenish more freshwater than it consumes by 2030. The company also reports using reclaimed or non-potable water at more than a quarter of its data center campuses. Whether zero-evaporation designs become the norm depends less on the technology, which works, and more on whether regulation forces adoption at scale. For more coverage of AI chips and infrastructure, visit our AI Hardware section.
