Just as news settles in that at least two new data centers have been proposed in Kern County, a report released Thursday raises concerns about the strain that installations powering artificial intelligence can impose on water supplies available to nearby and regional communities.
The study out of Santa Clara University, though hampered by what its authors allege are calculated efforts to deny the public information about how much water data centers consume, concludes the facilities increasingly threaten areas like the Central Valley deemed at-risk because of economic need and water scarcity.
“As data centers continue to expand across California, we’re seeing development move from cities into towns where water resources are already under strain and communities are more vulnerable,” stated founder F. Noel Perry of Next 10, a nonpartisan think tank that helped support the study.
“Without stronger safeguards, this growth risks compounding existing inequities — but with the right approach, it also presents an opportunity to build a more sustainable and inclusive model for digital infrastructure.”
Titled “The Intersection of Data Center Development, Water Availability, and Environmental Justice in California,” the report reinforces the most common complaint against costly installations that generate substantial tax revenue but use large amounts of electricity while producing relatively few jobs.
Water worries have already been expressed by opponents of a 238,000-square-foot data center a Trona-based company proposes to build along Highway 395 in Inyokern. Nearby residents raised the concern at a Ridgecrest City Council meeting last week, despite the developer’s insistence the development will come with a “highly efficient hybrid cooling system,” the details of which have not been disclosed.
Separately, local oil producer California Resources Corp. has reported plans for a data center on property it owns in Elk Hills, reflecting what some see as a strong likelihood the county’s ample land and business-friendly attitude will attract more such proposals.
Focusing on three existing facilities and two planned, the new report notes generative AI and the hyperscale data centers undergirding it consume and pollute substantial amounts of water “at nearly every stage of their supply chain.”
The researchers found that, as the installations are increasingly sited outside major metropolitan areas, they are ending up in outlying areas where they stress the water supplies of not only nearby communities but entire regions.
“When large, water-intensive facilities are added into these areas — particularly those relying on imported water or shared basins — the impacts extend beyond a single site,” lead author Iris Stewart-Frey, a hydrologist and professor of environmental science at Santa Clara University, said in a news release.
While previous studies have presented estimates data centers consume an average of 500,000 acre-feet per day, or as much as 10 times that volume, the authors do not say exactly how much water the data centers they studied consume. The authors blame that missing detail on a lack of transparency they said limits a full assessment of how the installations will affect California’s AI ambitions.
Stewart-Frey added, “Efficiency claims by corporations mean little without transparency and accountability. We can’t manage what we cannot measure and right now, data center water use is largely invisible.”
Director Katy Larson of the California Energy Research Center at Cal State Bakersfield was not familiar with the study. But she said by email Friday environmental reviews of data center projects should consider, among many factors, power and water management, alongside potential community benefits like job creation and community programs.
She noted that most, but not all, data centers use potable water to keep computer servers cool.
“While water usage continues to be a difficult conversation in the valley, there is room for cross-sector collaboration as much of the region’s produced oil and gas also produces fresh water that can potentially be recycled for data centers,” Larson wrote.