Generative AI models like ChatGPT and Gemini consume massive amounts of water and electricity. Training GPT-3 evaporated 700,000 liters of fresh water, and Google data centers used 2.9 trillion liters in 2023. By 2027, AI could need 4.2 to 6.6 trillion liters annually, straining regions already facing water stress. The problem isn't just carbon emissions. Solar power can cut electricity-related emissions, but cooling data centers still requires enormous water volumes. India, already facing severe water shortages, is becoming a hub for data centers in cities like Mumbai and Chennai, putting pressure on local drinking water supplies. Reducing AI's environmental impact requires more efficient models and better cooling technologies. Users also need to be mindful of how often they generate content, since each query has a real water and energy cost.
