Artificial intelligence data centers use significant amounts of freshwater to cool servers, with projections suggesting a massive increase in consumption by 2030.
This environmental cost is often invisible to users, but it poses a growing challenge to local water supplies as the AI industry scales globally.
Data centers rely on water-based cooling systems to dissipate heat generated by server hardware. These systems evaporate water to keep equipment at safe operating temperatures, creating a measurable water footprint for every request made to an AI model [1], [5].
According to reports, each individual AI prompt consumes approximately two tablespoons of water [1]. While this amount is small on a per-request basis, the cumulative effect across millions of users is substantial.
A United Nations report said that by 2030, AI could consume as much water as 1.3 billion people [2], [3]. Other estimates suggest the figure could be as high as the water used by 13 billion people [1].
These facilities draw freshwater from local supplies, which can strain resources in regions already facing water scarcity [1], [3]. The process involves a constant cycle of drawing water and evaporating it to manage the thermal output of high-performance computing clusters [1], [2].
The reports said that the hidden cost of every prompt includes not only water, but also carbon emissions and land use [3]. As the demand for generative AI grows, the pressure on these natural resources is expected to intensify.
“Each individual AI prompt consumes approximately two tablespoons of water.”
The shift toward large-scale AI integration creates a tension between technological advancement and environmental sustainability. Because data centers are often located in areas with existing water stress, the industry's reliance on evaporative cooling may force a transition toward more expensive air-cooling or closed-loop liquid systems to avoid depleting local freshwater reserves.
