AI models have a huge water footprint that most people do not even consider when they use generative AI. Behind every chatbot response or image generated lies the hidden cost of fresh water. Training and running large AI models like GPT-3 can consume millions of liters of freshwater, from cooling servers to producing electricity. This ‘thirst’ is often overlooked on multiple fronts, by consumers as well as industries because companies rarely report their full water use across data centers, electricity generation, and hardware manufacturing. Without greater transparency and sustainable practices, the rapid growth of generative AI challenges global water scarcity at a time access to clean water is already under threat.
Read More