AI uses a significant amount of water mainly due to the cooling needs of the data centers that power AI systems, as well as water used indirectly in electricity generation and hardware manufacturing. Estimates suggest that by 2027, AI could consume between 4.2 billion and 6.6 billion cubic meters (4.2 to 6.6 trillion liters) of water annually globally. For each interaction with AI, such as with ChatGPT, the water consumption can range around 0.3 to 500 milliliters depending on the efficiency and model, with roughly 500 milliliters (about a standard water bottle) being a common cited figure for a single prompt including processing and cooling requirements.
How AI Uses Water
- Cooling data centers: AI data centers generate a lot of heat, requiring water-intensive cooling systems that can consume up to 9 liters of water per kWh of energy used by cooling.
- Indirect power use water: Most electricity powering AI data centers comes from sources (coal, nuclear, hydroelectric) that require large amounts of water for generation.
- Hardware manufacturing: Producing semiconductor chips used in AI servers requires thousands of liters of ultra-pure water per wafer.
Scale of Water Use
- ChatGPT alone uses roughly 150 million liters per day worldwide based on user interaction estimates.
- Hyperscale data centers can use up to 550,000 gallons (~2 million liters) of water per day.
- The overall AI water footprint includes direct cooling plus indirect water embedded in electricity and hardware supply chains.
Environmental Concerns
- AI's water demand competes with local communities and ecosystems, particularly in water-scarce regions.
- Without sustainable water management, growing AI infrastructure may exacerbate water shortages and environmental impact.
Thus, AI's water use is substantial and growing, highlighting the importance of innovations in more efficient cooling technologies and renewable energy integration to reduce both water and energy footprints.
