The artificial intelligence industry is on track to consume up to 600 billion gallons of water by 2030, according to Tom's Hardware — and the primary culprit isn't the servers themselves, but the energy needed to run them.
Data centers housing AI workloads draw enormous amounts of power, and that energy demand is the main driver of the water footprint. According to Tom's Hardware, directly cooling the GPUs inside those facilities accounts for only a fraction of the water consumed — the larger share comes from energy consumption itself.
The trajectory is likely to get steeper. Tom's Hardware reports that future GPUs and rack systems are being engineered to handle even greater power loads, meaning data centers will grow more resource-intensive as AI capabilities scale, not less.
This water story is part of a wider reckoning with AI's environmental costs. Renewable Matter is among the publications now working to map the full range of AI's real environmental impacts — a sign that scrutiny is expanding beyond carbon emissions to resource use more broadly.
The stakes are real: water is a finite resource under growing pressure in many of the regions where large data centers operate, and 600 billion gallons by 2030 is not a distant projection but a near-term deadline arriving alongside the next wave of AI hardware.