Amazon says its data centers used ~2.5B gallons of water in 2025, or 0.12 liters of water per kWh, and water use at sites it owns and operates fell 2% from 2024
The company said water use at sites it owns and operates directly fell 2% from 2024 levels, even while it expanded its data-center footprint
Context & Ripple Effects
Data-center water use has become a persistent point of scrutiny: reporting previously put consumption in Virginia’s large data-center cluster above 1.85 billion gallons in 2023, while earlier coverage examined Google’s use in water-stressed areas. Microsoft has also made a major water-reduction pledge, making operator-level efficiency claims increasingly comparable.
Amazon’s disclosure arrives amid an AI-led buildout by the largest cloud platforms. The same expansion is associated in related coverage with rising emissions from Amazon’s data-center construction and a sharp increase in debt among major U.S. data-center builders.
First-order effects
- Amazon can point to lower water use at directly owned and operated sites despite expanding its data-center footprint, supported by a disclosed intensity measure of roughly 0.12 liters per kWh.
- The disclosure gives communities, customers, and investors a concrete—though company-reported—metric with which to assess Amazon’s cooling and facility-efficiency efforts.
Second-order effects
- Google, Microsoft, and other large operators face greater pressure to provide similarly specific water-use and intensity disclosures, rather than relying primarily on targets or broader sustainability commitments.
- In regions hosting dense data-center capacity, water efficiency becomes a more central factor in site permitting and in operators’ choice of cooling systems and locations.
Third-order effects
- If AI infrastructure continues to grow, the sector’s environmental debate will shift from whether individual facilities can improve efficiency to whether aggregate water demand can remain acceptable as capacity expands.
- Water accounting is likely to become a standard competitive and policy metric alongside power use and emissions, though lower intensity alone will not settle concerns where total local demand rises.
The trend: AI-driven data-center expansion is making water efficiency, local resource impacts, and standardized disclosure core constraints on cloud infrastructure growth.