UNU Report Warns AI Data Centers to Use Water Equal to Needs of 1.3 Billion People by 2030

A United Nations University (UNU) study projects that by 2030 artificial‑intelligence data centres will consume roughly 9.3 trillion litres of water – the amount required for the basic annual household needs of 1.3 billion people – and will require about 945 terawatt‑hours of electricity each year, almost three times the combined consumption of Pakistan, Bangladesh and Nigeria today. The same infrastructure is expected to occupy over 14 500 km² of land, twice the size of the Jakarta metropolitan area, and to generate up to 2.5 million tonnes of electronic waste annually.

The report notes that most AI energy demand now comes from inference, which accounts for 80‑90 % of total AI power use. For example, ChatGPT processes around 2.5 billion queries daily, consuming about 383 GWh of electricity each year. More than 90 % of specialised AI compute capacity is located in the United States and China, while more than 150 countries lack their own AI infrastructure but bear the environmental costs of resource extraction and e‑waste. The authors call for a broader accounting of AI’s environmental footprint and for responsible development that balances benefits with planetary limits.