UN Report Highlights AI Video Generation's High Water Consumption

A United Nations Institute of Water, Environment and Health study found that creating a single artificial‑intelligence‑generated video can consume up to 4.1 liters of water. The report attributes the water use to cooling requirements in data‑center facilities that power AI models. It warns that as AI tools become more prevalent, global water, electricity and land demand could rise sharply. By 2030, AI‑related data centers might require around 945 terawatt‑hours of electricity annually and up to 9.3 trillion liters of water, occupying roughly 14,500 square kilometres of land. Most of the energy is used during everyday operation rather than model training, with 80‑90 % of power demand occurring while the systems serve millions of users.