AI Infrastructure May Consume 6.6 Billion Cubic Metres of Water by 2027, Near Half of UK Withdrawal
Updated
Updated · spacedaily.com · Jun 3
AI Infrastructure May Consume 6.6 Billion Cubic Metres of Water by 2027, Near Half of UK Withdrawal
2 articles · Updated · spacedaily.com · Jun 3
Summary
Global AI demand is projected to require 4.2 billion to 6.6 billion cubic metres of water withdrawal a year by 2027, with the high end approaching half of the UK's annual total.
Evaporative cooling drives the surge: high-end AI chips shed 300 to 700 watts each, about 80% of cooling water is lost to evaporation, and indirect water used to generate electricity can be roughly 12 times direct cooling use.
Corporate data already show the trend accelerating. Google used about 8.1 billion gallons in 2024—95% at data centres—nearly doubling from 2021, while Microsoft reported 1.7 billion gallons in 2022 and one GPT-4 training cluster used 24.9 million gallons in two months.
Water stress is increasingly local as well as global: Microsoft said 42% of its 2023 water use came from water-stressed regions, Google put 15% of its freshwater withdrawals in high-scarcity areas, and projects in Chile, Mexico, Uruguay, Arizona and Spain have drawn opposition or delays.
Researchers say company disclosures understate the footprint because they often omit indirect power-generation water, blur withdrawal versus consumption, and avoid facility-level data even as AI infrastructure investment is projected to reach $5.2 trillion by 2030.
With trillions invested in AI, who pays the environmental bill for its massive water use?
Can AI's thirst be quenched by new tech before it drains critical water reserves?
AI’s Thirst: How Data Centers and Chip Manufacturing Are Driving a Water Crisis in 2026 and Beyond
Overview
The rapid expansion of AI infrastructure is driving an unprecedented and urgent demand on global water resources, mainly due to the intensive cooling required for data centers and the water-heavy manufacturing of specialized AI chips. As data centers increase in number, size, and complexity to meet AI's growing needs, their water consumption is intensifying, putting significant pressure on already stressed water systems. This surge in demand is especially evident in regions like Arizona, where numerous new data centers have been built or announced, highlighting the escalating competition for limited water resources and the urgent need for sustainable solutions.