UN report warns AI data centers could use water equivalent to 1.3 billion people by 2030

A United Nations Institute for Water, Environment and Health study projects that the AI industry’s data‑center infrastructure will require about 9.3 trillion litres of water annually by 2030 – a volume comparable to the basic water needs of roughly 1.3 billion people. Electricity demand is forecast to rise to 945 TWh per year, roughly three times the combined consumption of Pakistan, Bangladesh and Nigeria, and would represent more than 3 % of global electricity use. To accommodate the hardware, the industry would need about 14,500 km² of land – ten times the size of Mexico City. The same report estimates a carbon footprint of 399 million tonnes of CO₂‑equivalent and predicts up to 2.5 billion tonnes of electronic waste by 2030.

AI workloads already account for 20 % of data‑center power use, and could reach 40 % by the end of the decade. Most of the energy (80‑90 %) is consumed during everyday inference rather than model training, and a single AI‑generated video can consume up to 4.1 litres of water. Offsetting these impacts would require planting millions of trees over the next ten years. The authors call for a “net‑positive energy AI” approach to ensure that the technology’s benefits outweigh its environmental costs.