how does ai use water

AI uses water mainly because the computers that run it need huge amounts of power and cooling, not because algorithms themselves “drink” water.

Core idea

When you use AI, your request is processed in a data center full of servers that get very hot and use a lot of electricity. Water comes in at two main stages: cooling those servers and generating the electricity that powers them.

Where the water goes

• Direct cooling on site
  Many data centers run cooling towers or evaporative cooling systems that use water to carry heat away from servers, and that water partly evaporates into the air. On average, a large data center can consume hundreds of thousands to a couple of million liters of water per day just for cooling.

• Indirect use via power plants
  A lot of electricity still comes from thermoelectric plants (like coal, gas, or nuclear) that use water to produce and cool steam. Each kilowatt-hour of electricity used for AI can represent several additional liters of water used upstream at the power plant.

AI’s growing water footprint

• Training big AI models (the “learning” phase) is especially intense because it runs thousands of powerful chips for weeks, which means more power and more cooling. Inference (everyday use, like chat or image generation) adds up too, because it happens millions or billions of times.

• Analyses in recent years estimate that global water use from AI data centers is already in the hundreds of billions of liters annually and could roughly double by 2030 if growth continues. New AI data centers are often being built in regions that already face high water stress, which amplifies local tensions over drinking water and agriculture.

Efforts to cut water use

• Some operators are shifting to air-based or liquid-loop cooling that uses far less or no evaporated freshwater on site, sometimes using non‑potable or recycled water instead of drinking water. Others are experimenting with advanced liquid coolants and tighter heat management to reduce both electricity and water needs.

• Switching data centers to renewable energy (solar, wind) can shrink the indirect water footprint because these sources typically use much less water than fossil-fuel power plants. There is also growing pressure for companies to report how much water their AI services use and to site new facilities where water resources are less strained.

TL;DR: AI uses water because data centers need cooling and electricity, and both often depend heavily on water, especially at large scale.