Short answer
AI is unlikely to be the sole cause of water shortages on its own, but large-scale data centers can contribute to local water stress when they rely on water-intensive cooling or operate in regions where freshwater is already limited.
Why AI data centers may use water
AI workloads run on dense computing infrastructure that produces heat. Some data centers use evaporative or water-based cooling to keep servers and accelerators within safe operating temperatures. The amount of water used depends heavily on the cooling design and local climate.
Local context matters more than global averages
The water impact of AI depends strongly on where the data center is located. A facility in a water-stressed region can create more concern than a similar facility in an area with abundant renewable water resources or cooler weather.
AI can also use water indirectly
Some water use is indirect. Electricity generation can require water, especially for thermal power plants. When AI increases electricity demand, part of the associated water footprint may occur upstream in the power system rather than inside the data center itself.
When can AI contribute to water stress?
AI infrastructure can contribute to water stress when large data centers are concentrated in dry regions, when cooling systems consume freshwater, when electricity comes from water-intensive generation, or when reporting is not transparent enough for communities to evaluate local impact.
How the risk can be reduced
The risk can be reduced through better site selection, more efficient cooling, reclaimed water, air cooling where appropriate, low-water electricity sources, transparent reporting and public planning around data center expansion. For measurement assumptions, see the Methodology.