Protests are erupting across the American Southwest as residents push back against the rapid buildout of AI data centers in their communities. Signs reading “Water for people not AI” have appeared from Texas to Arizona, where locals are angry about the strain these facilities place on water, energy, and local environments. The processors inside data centers run hot—up to 176 degrees Fahrenheit—and many facilities rely on evaporative cooling, which works like sweating to keep equipment from overheating.
But the actual water picture is murkier than the protest signs suggest. Researchers estimate data center cooling consumed 66 billion liters of water in 2023, less than 1 percent of total U.S. consumption. That’s a rounding error compared to agriculture or manufacturing. Yet public claims range from the absurd—chatbots supposedly drinking 500 ml per query—to outright denial that AI has any water problem at all. The tech companies themselves don’t help much, as their consumption data tends to be incomplete and inconsistent.
The real issue isn’t national totals; it’s geography. A data center in a water-rich region like the Pacific Northwest barely registers. But the same facility in drought-strapped New Mexico or Arizona adds real stress to an already strained supply. That’s where the anger comes from, and it’s not irrational—even if the online discourse around it frequently is.
Fengqi You, an energy systems expert at Cornell, struck a pragmatic note: “There are many good ideas out there.” Engineers are already deploying alternative cooling methods, and combined with efforts to replenish local water resources, You sees a path toward net-zero water use for on-site cooling. That’s a meaningful shift from the doom-and-gloom framing, though it requires companies to actually invest in those solutions rather than defaulting to the cheapest option. Amazon, Google, and Microsoft have all made water-positivity pledges in recent years—the question is whether those commitments survive the AI infrastructure gold rush.
