Global AI Could Withdraw 6.6 Billion Cubic Meters of Water by 2027, Equal to Half the UK
Updated
Updated · ScienceBlog.com · Jul 10
Global AI Could Withdraw 6.6 Billion Cubic Meters of Water by 2027, Equal to Half the UK
1 articles · Updated · ScienceBlog.com · Jul 10
Summary
4.2 billion to 6.6 billion cubic meters of water withdrawal could be tied to global AI demand in 2027, according to model-based projections that frame the issue as infrastructure strain rather than per-query trivia.
Two channels drive that footprint: direct water consumed in data-center cooling and indirect water used to generate electricity, with totals varying by model, location, weather, cooling design and accounting method.
0.26 milliliters per Gemini text prompt in a 2025 Google paper and bottle-scale ChatGPT estimates are not directly comparable because they measure different systems, workloads and boundaries.
517 of 809 planned U.S. data centers were in places hit by drought in the prior year, underscoring why water location matters more than a single global average.
Hundreds of millions of gallons a day of new U.S. water capacity may be needed through 2030 if 2024 intensity persists, shifting the debate toward disclosure, siting and public water-system limits.
With data centers moving into dry areas, who truly pays the price for AI's massive water consumption?
Tech giants pledge to replenish water, but can innovation outpace AI's demand in drought-prone regions?
AI’s Thirst: How Artificial Intelligence Could Consume Trillions of Liters of Water Annually by 2027—and What We Must Do Now
Overview
The rapid expansion of AI infrastructure is driving a sharp increase in global water demand, especially as data centers consume more water for cooling. If current water-use intensity continues, U.S. data centers alone could require hundreds of millions of gallons per day by 2030, turning AI’s water footprint into a trillion-liter-a-year issue. This rising demand puts significant pressure on public water systems, as communities must reserve large amounts of water for data centers, often during peak periods when resources are already strained. As AI grows, managing its water use becomes a critical challenge for both industry and society.