Water footprint of large AI models: risks to global water security and sustainable optimization strategies
Water Policy · 02.11.2026
DOI: 10.2166/wp.2026.233 ↗Kurz gefasst
Fachartikel in der Zeitschrift Water Policy, online erschienen am 02.11.2026. Verfasst von Tahdia Tahmid und 1 weiteren Autor:innen. Die Arbeit nennt 154 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Diagram linking AI model training and inference to lifecycle water use, estimated at 4.2–6.6 billion m³ annually by 2027, and showing water reuse, efficient cooling, smarter workloads, and standardized reporting as strategies to protect global water security. The rapid proliferation of large-scale artificial intelligence (AI) models has introduced a significant yet underexamined environmental challenge: freshwater consumption. This narrative review synthesizes scientific literature, industry sustainability reports, and case studies to quantify the water footprint of AI systems across their full lifecycle, from semiconductor fabrication and data center operations to electricity generation and supply chains. Direct water use occurs primarily through evaporative and liquid cooling in data centers, where a single large facility can consume 11,000–19,000m3/day. Indirect water use is embedded in electricity generation and chip manufacturing, with advanced fabrication plants drawing 10,000–37,000m3/day. Major technology companies, including Google, Microsoft, and Meta, collectively consume billions of cubic meters in 2022–2024, with consumption rising alongside accelerating AI deployment. Critically, over 52% of the world's top 100 data centers are projected to be located in high water stress regions by 2030, where AI infrastructure increasingly competes with agricultural, municipal, and ecological needs. Sustainable alternatives, including immersion cooling, nonpotable water reuse, and AI-driven workload optimization, offer measurable reductions but remain inconsistently adopted. The direct and indirect water footprints of AI models during their entire lifecycle are reviewed. Two of the main direct water usage sources are semiconductor manufacturing and data center cooling. Clip manufacturing and electricity generation significantly increase the indirect water footprint of AI. In areas with scarce freshwater supplies, the growth of AI technology may worsen water stress. Standardized reporting frameworks and lifecycle-based water accounting are urgently needed to safeguard global water security.
Zitieren
APA
Tahmid, T., & Upaul, S. (2026). Water footprint of large AI models: risks to global water security and sustainable optimization strategies. Water Policy. https://doi.org/10.2166/wp.2026.233
Vancouver
Tahmid T, Upaul S. Water footprint of large AI models: risks to global water security and sustainable optimization strategies. Water Policy. 2026. doi:10.2166/wp.2026.233
BibTeX
@article{Tahmid2026,
author = {Tahmid, Tahdia and Upaul, Sharfan},
title = {Water footprint of large AI models: risks to global water security and sustainable optimization strategies},
journal = {Water Policy},
year = {2026},
doi = {10.2166/wp.2026.233}
}
EUVIMED