AI's thirst for water may hit 600 billion gallons by 2030.
Data centers that run AI models rely on direct‑cooling systems for GPUs. Those systems use only a small slice of the total water required, but future GPUs and denser rack designs will need far more power. As the electricity demand climbs, the associated water for cooling rises proportionally.
The surge matters because water scarcity is already a concern in many regions hosting large farms. Higher water usage could force operators to relocate, invest in expensive cooling tech, or cap AI workloads altogether.
In short, today's push for faster AI may end up draining more than just electricity—it could strain a resource many tech firms haven’t planned for.