Why does AI use water?

Every time you type a prompt into ChatGPT, Gemini, Claude, or any AI tool, a chain reaction of water consumption begins. It happens in three ways:

• Cooling the data centers. AI chips run incredibly hot. GPUs processing your prompts can reach over 80°C. Data centers pump millions of liters through cooling towers and evaporative systems to keep servers from overheating. A single large data center can use as much water as a small city — up to 5 million liters per day.
• Generating the electricity. AI is power-hungry. A single ChatGPT query uses roughly 10x the electricity of a Google search. Most of that power comes from thermal and nuclear plants that boil water into steam to spin turbines, then evaporate even more water to cool back down. This "hidden" water cost often exceeds the direct cooling cost.
• Manufacturing the chips. The GPUs and TPUs that power AI require ultra-pure water during fabrication — up to 40,000 liters per silicon wafer. Every rinse, etch, and chemical process demands water so clean it conducts almost no electricity. A single chip fabrication plant can use 40 million liters per day.

The hotter the climate where a data center is located, the worse it gets. A query processed in Phoenix, Arizona can use 3–5x more cooling water than the same query processed in cooler climates like Ireland or the Pacific Northwest.