These figures exclude the substantial “embodied” water required to manufacture AI chips and hardware, prompting UC Riverside researcher Shaolei Ren to argue that the real total is even higher and that earlier projections were too conservative.
References (APA style)
De Vries-Gao, A. (2025). Environmental impacts of artificial intelligence: Energy use, emissions, and water consumption. Patterns.
Wolverton, T. (2025, December 19). Study: Artificial intelligence models might be more thirsty than thought. San Francisco Examiner. Voir moins
