Massive Data Centers, Vanishing Wells

Network switches with blue Ethernet cables — Photo: asharkyu / Shutterstock

AI data centers are moving into drought zones and drawing water on a scale that can rival whole towns, putting local families and farms at risk.

Story Highlights

Large AI sites can use up to 5 million gallons of water per day for cooling.
Two-thirds of new U.S. data centers since 2022 sit in high water-stress areas.
U.S. data center water use rose from 21.2 to 66 billion liters between 2014 and 2023.
Investigations place dozens of Big Tech facilities in drought-hit regions.

AI Buildout Targets Thirsty Regions With Limited Water

Reports show two-thirds of new American data centers since 2022 are landing in high water-stress zones, where supply is already tight and drought risk runs high. Investigators also identified 38 operational sites owned by Amazon, Microsoft, and Google in stressed regions, with 24 more being built. These locations were chosen for land, tax breaks, and grid links. Local leaders often chase short-term revenue, while families nearby worry about wells, crops, and basic water security as demand spikes.

Daily on-site water needs can be staggering. Some large artificial intelligence facilities can consume up to five million gallons in a single day to cool racks of servers. That rivals what tens of thousands of people use. National totals are climbing fast, too. Direct data center consumption jumped from 21.2 billion liters in 2014 to 66 billion liters in 2023 as artificial intelligence loads soared. These withdrawals may come from city systems or groundwater, placing stress on the same sources that families and farmers use.

Numbers Point To Mounting Strain, But Local Proof Is Sparse

Evidence confirms heavy and rising water use, but there is a gap in direct links to named private wells failing. The research does not provide site-level hydrology, well logs, or permits that prove a specific artificial intelligence site caused a neighbor’s well to run dry. That missing link gives Big Tech room to deflect. Yet a peer-reviewed preprint found utility burdens from artificial intelligence sites ranging from 0.2 percent to 134 percent of host capacity across ten American locations, showing community-scale risk is real.

Texas offers a warning on scale. Analysts project data centers in the state could use 49 billion gallons in 2025 and up to 399 billion gallons by 2030 if growth continues. National projections show annual water demand from American data centers hitting 73 billion gallons by 2028, up from 17 billion gallons in 2023, as artificial intelligence drives new facilities and larger cooling loads. These numbers do not prove a single dried well. They do show a trend that can overpower local systems if growth outpaces planning.

Disclosure Gaps Leave Communities In The Dark

No federal law requires private data centers to report water use, which means vital details remain hidden from the public and many local boards. Some states, such as Minnesota, Virginia, and California, have added reporting and aquifer checks, but coverage is uneven and slow. Without clear facility-level numbers, families cannot verify how much water a site draws from their shared source. That information gap fuels mistrust, invites poor siting choices, and makes smart planning much harder for county leaders.

Industry groups say solutions exist, including performance standards and credits for reclaimed or nonpotable water. Those ideas could help if made mandatory and enforced. Companies also promote the point that most water tied to data centers is “indirect,” through electricity generation, not on-site cooling. That claim may be true at a broad level, but it misses the core local issue. What matters to a family on a well is the water pulled from their basin today, not distant totals spread across a grid.

What Responsible Growth Should Require Right Now

County boards should demand site-specific groundwater studies before permits. Leaders should require facility-level disclosure of on-site withdrawals and cooling methods, reviewed by independent hydrologists. State agencies should post permits and monitoring data online for public review. Congress should set a simple national reporting floor so communities can compare sites across regions. These steps defend property rights, keep families first, and support growth that does not jeopardize water for homes and farms.

Big Tech’s push into dry regions did not start the water crisis, but it can make a bad situation worse if secrecy and weak rules persist. America can welcome jobs and innovation without risking taps and wells. The path forward is clear: choose locations with sustainable supply, disclose real numbers, and use reclaimed water wherever possible. Local families should not lose out so distant servers can chase the next algorithm. Sunlight, standards, and strong local control must come first.