Christine Lemmer-Webber
boosted
#Report by the United Nations University quantifies, for the first time, the carbon, #water and land footprint of #AI. The figures are staggering : 945 TWh of electricity projected for 2030, 9.3 trillion litres of water, 14,500 km² of land.
Spoiler : It will take the water and electricity consumption of 1.3 billion people to meet your prompts by 2030
« ENVIRONMENTAL COST OF AI'S ENERGY USE - Carbon, Water & Land Footprints »
› https://collections.unu.edu/eserv/UNU:10647/UNU-INWEH-Report-The_Env_Cost_of_AI-2026.pdf (.pdf)
Nearly half of the world’s data centres are located in the United States. More than 150 countries have no data centres dedicated to AI.© UNU-INWEH 2026 / Cloudscene
From text classification (0.002 Wh) to video generation (415 Wh), the energy consumed per query varies by a factor of 200,000.© UNU-INWEH 2026
The environmental footprint of a single data centre varies considerably depending on the country in which it is located. At the top, carbon intensity per kWh. In the middle, water footprint. At the bottom, land footprint. © UNU-INWEH 2026
