Majority of US AI Data Centers Planned for Drought-Hit Areas
New analysis reveals 64% of upcoming infrastructure projects face severe water scarcity risks
A record-shattering drought has placed the majority of the United States' planned AI data center expansion in direct competition with local water needs. An analysis of 809 upcoming infrastructure projects found that 517 are slated for regions that have experienced drought conditions within the past year.
Key details
The surge in artificial intelligence compute demand has triggered a massive build-out of physical infrastructure, but geography is increasingly at odds with resource availability. According to research published this week, 64% of the 809 planned AI data centers in the U.S. are located in drought-stricken zones as classified by the National Oceanic and Atmospheric Administration (NOAA).
While direct cooling of server racks accounts for approximately 4% of AI's total water footprint, the broader infrastructure required to support these facilities—including semiconductor fabrication and water-intensive thermal power generation—consumes the vast majority of the resource. Some individual high-density data center projects have already come under scrutiny for consuming as much as 29 million gallons of water over a 15-month period during construction alone.
Why this matters
The concentration of AI infrastructure in water-stressed regions creates a significant operational and reputational risk for the industry. As data centers compete with agriculture and residential users for declining water tables, the potential for regulatory intervention and project cancellations increases. This "water bottleneck" may soon rival energy grid constraints as the primary limiting factor for AI scaling.
Context
This report follows a series of local moratoriums and protests in states like Maine, Nevada, and Washington, where residents have voiced concerns over rising utility bills and resource depletion. The industry is currently exploring direct-to-chip liquid cooling as a potential efficiency fix, with some systems claiming up to 300 times the water efficiency of traditional air cooling, though these figures often exclude the water used in the upstream power generation required to run them.
Risks and open questions
The primary risk is the social and political backlash from rural communities where many of these facilities are planned. If federal or state regulators begin mandating water reduction for industrial users during peak drought periods, AI operators may face forced curtailments or expensive transitions to closed-loop cooling systems. It remains unclear if the industry's efficiency gains can keep pace with the sheer volume of new 100+ MW facilities being commissioned.
What happens next
Expect increased pressure on major tech firms to disclose granular water and energy usage data. Investors are already beginning to demand transparency on how AI expansion will impact long-term resource availability. In the coming months, we will likely see more data center projects shifting toward "behind-the-meter" water restoration projects as companies attempt to offset their local footprints.
Source: The Guardian Published on AI Usage Global, author: AUG Bot
