AI and water
AI consumes water to cool servers, to generate electricity and to manufacture chips. Depending on the scope counted, from cooling alone to the full life cycle, one query ranges from 0.26 mL to 45 mL.
By Projet Celsius
Focus areaData centre cooling0.26 to 45mLof water for one query to an AI, from cooling alone to the full life cycleGoogle; Mistral AI, 2025See its charts →
Focus areaTypes of withdrawal560billion litresof water per year for the world's data centres, on site and for their electricityIEA, April 2025See its charts →
Focus areaLocal impacts50%of Microsoft's water withdrawals in areas of water stress, financial year 2025Microsoft, 2026See its charts →Charts and maps of the topic
Charts and maps →Focus area
ChartOne question · logWater for one question, from 0.26 to 50 mL depending on scopeGoogle, Gemini0.26OpenAI, ChatGPT0.32UC Riverside10 to 50Mistral AI, Le Chat45 mL× 170from Google to Mistral AIGoogle; OpenAI; UC Riverside; Mistral AIChartCompany · 2020 to 2025Microsoft's water use, 2020-202513.3million m³withdrawn in 2025Microsoft, 2026 data sheetKey figureOne question · 2025Water for one question to Gemini0.26mLabout 5 drops of water evaporatedOn-site cooling onlyGoogle, August 2025Key figureWorld · per yearWater consumed by the world's data centres560billion litresDirect and indirect waterIEA, April 2025MapWorld · 2023Water stress, country by countryFrance 21%Spain 43%India 66%withdrawals / resources17.6%world averageFAO, SDG 6.4.2, via Our World in Data
Analyses and articles
Analyses
Where does this figure come from?
Water20 October 2026"An email written by an AI is a bottle of water"The comparison comes from a paywalled Washington Post article (September 2024), produced with researchers: we were unable to check its figure, so we do not use it. Their method is published, and providers have since measured the water used by one question.
Analysis
Climate12 November 2026Why do estimates of AI's impact differ?Calculated with French electricity, one question to an AI emits 0.012 gram of CO2e; Mistral AI publishes 1.14 gram, more than 90 times as much. Its water use ranges from 0.26 to 45 millilitres depending on the published measurements. These figures are accurate; they do not cover the same scope.On the Projet Celsius blog ↗
Sources for this topic
- Elsworth et al. (Google), Measuring the environmental impact of delivering AI at Google scale, August 2025 ↗
- Sam Altman (OpenAI), The Gentle Singularity, June 2025 ↗
- Li, Yang, Islam and Ren, Making AI Less "Thirsty", 2023 ↗
- Mistral AI, with Carbone 4 and ADEME, Our contribution to a global environmental standard for AI, July 2025 ↗
- Microsoft, 2026 Environmental Sustainability Report, Data Fact Sheet, July 2026 ↗
Across the hub
AI Impact DatabaseAll the figures, their methodology and their sources.
AI Ranking10 assistants rated out of 100: performance, ethics and environment.
AnalysesThe signed pieces: analyses, viewpoints, "Where does this figure come from?".
Charts and mapsThe reference visuals, with their data and sources.
ToolsOpen calculations, your assumptions on the left, the result on the right.
A question about a figure?
An error, a source to add, a methodology question, a press request: write to us. Guillaume Pakula, Clément Reynaud, Sébastien Pierfederici and Alexandre Veilly will reply.
Write to us →
Projet Celsius, which runs this hub, also helps companies measure and reduce the footprint of their AI use.
Accounting for AI in your Bilan Carbone® ↗
