Verify any claim · lenz.io
Claim analyzed
Science“A study published in the journal Patterns estimated that the AI industry consumed between 312 billion and 764 billion liters of water in 2025.”
Submitted by Wise Zebra fe7f
The conclusion
Open in workbench →Reliable primary and secondary sources indicate that a Patterns study estimated AI's 2025 water footprint at roughly 312.5 to 764.6 billion liters. The attribution and range are substantially correct. The main caveat is that this is a modeled estimate with uncertainty, not a directly measured industry-wide total, and the study's scope is somewhat narrower than the phrase “AI industry” suggests.
Caveats
- The figures are modeled estimates, not audited or directly metered water consumption across the entire AI sector.
- The wording "consumed" is slightly stronger than the study's more conditional framing of what AI water use could have been in 2025.
- "AI industry" may overstate the study's scope, which is closer to AI systems and related data-center impacts.
Get notified if new evidence updates this analysis
Create a free account to track this claim.
Sources
Sources used in the analysis
The carbon footprint of AI systems alone could be between 32.6 and 79.7 million tons of CO2 emissions in 2025, while the water footprint could reach 312.5-764.6 billion L.
Based on my calculations, AI systems alone use between 312.5 and 764.6 billion litres of water. That is of the same order of magnitude as all bottled water consumed worldwide in a single year. Recent research by Alex de Vries-Gao shows that, by 2025, AI systems could have a carbon footprint comparable to that of a global city such as New York and consume as much water as all bottled water drunk worldwide in a year. The study has been published in the scientific journal Patterns.
De Vries-Gao estimated that the carbon footprint of AI systems could reach as high as 80 million tonnes in 2025, with water usage potentially hitting 765 billion liters. These findings are published in the journal Patterns. The study released on Wednesday assessed the global environmental repercussions of this swiftly evolving technology, revealing that AI-related water consumption now surpasses the total demand for bottled water worldwide.
AI water footprint estimated at 312.5-764.6 billion liters in 2025—comparable to global bottled water consumption ... While energy consumption captures headlines, water usage represents an equally critical challenge. Research published in Cell Patterns by Alex de Vries estimates AI systems consumed 312.5-764.6 billion liters of water in 2025—comparable to the entire global bottled water industry.
My latest research on the carbon and water footprints of AI was just published in the journal Patterns. This study highlights how “AI created as much carbon pollution this year as New York City and guzzled up as much H20 as people consume globally in water bottle.” It also highlights how big tech disclosure typically falls short of delivering what’s needed to get more accurate data, “often excluding key details, like their indirect water consumption from electricity demand and how much is used for AI specifically.”
Embedded supply-chain water, in semiconductor fabrication, hardware production, and construction, is the least visible component; one major operator reports that supply-chain water exceeds 99% of its corporate footprint [10], and global AI water footprints are estimated at 312.5–764.6 billion litres for 2025 [26].
New research published in the journal Patterns suggests that the rapid expansion of artificial intelligence could carry a major environmental cost. According to data scientist Alex de Vries Gao, AI data centers may emit between 33 and 80 million metric tons of CO₂ in 2025, potentially surpassing the annual emissions of New York City. ... Beyond carbon output, the research also estimates AI-related water consumption could reach 312 to 767 billion liters in 2025, possibly exceeding the global bottled water industry’s yearly use. De Vries-Gao cautioned that these figures carry significant uncertainty, largely due to limited transparency from major AI companies about their energy and water usage.
Annual water consumption for 2025 is estimated to range between 312.5 billion and 764.6 billion liters. A. de Vries-Gao noted that this is equivalent to the total annual volume of bottled water consumed by the entire human population. The figure is also more than a third higher than earlier 2024 estimates.
The paper, published in peer-review journal Patterns, shows that the water consumption from AI in 2025 may have exceeded the world’s consumption of bottled water, and may have released more CO2 in the atmosphere over the year that America’s most populated city, New York. However, AI use has exploded, and the conservative estimates from Alex de Vries-Gao’s more recent study places AI’s water consumption between 312.5 and 764.6 billion liters of water. The likely consumption over the lifetime of the hardware manufacturing and disposal or recycling are much higher.
In 2025, a peer-reviewed study published in the journal Patterns revealed that the environmental cost of artificial intelligence has surged to industrial scales. Researchers estimated that AI systems globally consumed between 312.5 and 764.6 billion liters of water this year. At the high end, this footprint exceeds the approximately 446 billion liters used annually by the global bottled water industry.
The study also finds that AI-related water consumption could reach between 312.5 and 764.6 billion liters. Artificial intelligence has already generated as much carbon dioxide as New York City emits in a year and used as much water as the world consumes in bottled water in 2025, according to new research. The study, published on December 17 in the academic journal Patterns by Alex de Vries-Gao, a PhD candidate at the VU Amsterdam Institute for Environmental Studies, highlights the growing environmental footprint of AI.
A recent study from 2025, published in the journal Patterns by Alex de Vries-Gao, specifically examined the environmental impact of AI systems. The study shows that AI systems are estimated to have consumed around 765 billion liters of water in 2025, an amount greater than the total global consumption of bottled water in that same year.
What do you think of the claim?
Your challenge will appear immediately.
Challenge submitted!
For developers
This same pipeline is available via API.
Verify your AI's output programmatically.
/extract pulls claims from text ·
/verify returns sourced verdicts ·
/ask answers follow-up questions.
Continue your research
Verify a related claim next.
Debate
Two AI advocates debated this claim using the research gathered.
Argument for
The PubMed-indexed paper dated 2025-12-17 explicitly reports that “AI systems alone” could have a 2025 water footprint of 312.5–764.6 billion liters, matching the motion's 312–764 billion liter range (Source 1, PubMed: “The carbon and water footprints of data centers and what this could ...”). Independent reporting and the author's institution confirm these figures were published in the journal Patterns—VU Amsterdam states the study was published in Patterns and gives the same 312.5–764.6 billion liters estimate, and The Guardian reiterates the Patterns publication and the ~765 billion liters upper bound (Sources 2 and 3).
The Proponent's reliance on Sources 1-3 rests on repeated conditional phrasing such as 'could reach' and 'could have,' which denotes projections rather than documented actual consumption by the AI industry. This approach ignores the explicit caveats in Source 7 regarding major uncertainty from incomplete corporate disclosures, leaving the motion's claim of verified 2025 totals unsupported.
Argument against
Source 1 and Source 2 show the Patterns study only projected hypothetical water footprints that AI systems 'could reach' or 'use' in 2025, not verified actual consumption by the full AI industry. Source 7 further notes these figures carry major uncertainty from incomplete corporate disclosures, undermining any claim of measured industry totals.
The Opponent equivocates between “measured, verified actuals” and the motion's narrower wording—“estimated”—even though the Patterns-linked PubMed record explicitly quantifies a 2025 “water footprint” range of 312.5–764.6 billion L for “AI systems alone,” which is exactly what an estimate is (Source 1, PubMed: “The carbon and water footprints of data centers and what this could ...”). The Opponent's appeal to uncertainty is non-responsive because Source 7's caveat about limited disclosure does not negate the existence or validity of the published estimate, and multiple independent summaries still attribute the same 312.5–764.6 billion L 2025 estimate to the Patterns study (Sources 2 and 3).
Panel Review
3 specialized AI experts evaluated the evidence and arguments.
Reviewer 1 — The Logic Examiner
The logical chain from the evidence to the claim is direct and sound, as multiple independent sources (Sources 1, 2, 3, 4, 5, and 11) confirm that a study by Alex de Vries-Gao published in the journal Patterns estimated the 2025 AI water footprint to be between 312.5 and 764.6 billion liters. The Opponent's objection conflates 'estimated' figures with 'verified actuals,' which is a straw man argument since the atomic claim itself explicitly uses the word 'estimated.'
Reviewer 2 — The Source Auditor
High-authority sources—especially the PubMed-indexed record of the Patterns paper (Source 1) and the author's university news release (Source 2)—attribute to the Patterns study an estimate that AI systems' 2025 water footprint could be about 312.5–764.6 billion liters, with major media coverage (Source 3, The Guardian) repeating the same range and attribution. Because the claim is explicitly about an estimated range published in Patterns (not verified metered totals), the most reliable sources support it, with the main caveat being that the underlying estimate is uncertain due to limited disclosure rather than absent.
Reviewer 3 — The Precision Analyst
The claim's 312-764 billion liter range and Patterns journal attribution match the study's reported 312.5-764.6 billion liter estimate for AI systems (Sources 1-3), with only minor rounding; however, the wording 'consumed' overstates the evidence's consistent use of conditional projections ('could reach', 'could have') rather than verified actuals. The claim is therefore accurate at its stated strength with only minor imprecision.