đź“° Full Story
A United Nations University report published 3 June 2026 warns that the rapid expansion of artificial intelligence could place severe strains on electricity, water and land.
The UNU‑INWEH study projects that data centres powering AI could consume about 945 terawatt‑hours of electricity by 2030, require roughly 9.3 trillion litres of water (equivalent to the basic annual needs of 1.3 billion people in sub‑Saharan Africa) and occupy about 14,500 square kilometres of land.
Inference—the day‑to‑day queries users send to deployed models—now accounts for an estimated 80–90% of AI energy use.
The report cites examples such as ChatGPT processing some 2.5 billion prompts daily, using about 383 GWh a year.
It warns that carbon‑focused metrics alone hide trade‑offs: moving to bioenergy can lower CO2 but greatly increase water and land footprints.
Projected e‑waste could reach up to 2.5 million tonnes annually by 2030.
UNU‑INWEH urges standardized environmental disclosure by AI firms, regulatory reporting, siting rules to protect water‑stressed regions, efficiency improvements, and behavioural changes by users (for example, shorter prompts or avoiding energy‑heavy image/video generation).







đź’¬ Commentary