| Google (Gemini) | Yes | Yes | Yes | Yes | Yes |
Water Usage Reports total water consumption annually (10.9B gallons in 2025, up 34% from 2024), with withdrawal, discharge and consumption broken out for data centers and offices. View source →Energy Usage Reports total energy consumption (44.0M MWh in 2025, of which 42.4M MWh data centers) and a 2025 fleet-wide average PUE of 1.09. Reports ~65% global hourly carbon-free energy in 2025. View source →Per-Facility Data Publishes water withdrawal, discharge and consumption for each data center location (e.g., Council Bluffs, IA: 1.35B gallons consumed in 2025) and PUE per facility. View source →AI-Specific Data Published per-prompt estimates for the median Gemini Apps text prompt (0.24 Wh energy, 0.03 gCO2e, 0.26 ml water) with a technical methodology paper (2025); repeated in the 2026 Environmental Report. View source →Water Pledge Aims to replenish 120% of the freshwater it consumes, on average across offices and data centers, by 2030. Replenished ~7.7B gallons (~78% of freshwater consumption) in 2025. View source → |
| Microsoft (OpenAI) | Yes | Yes | No | No | Yes |
Water Usage Reports total water withdrawal (13,266 ML) and consumption (8,170 ML, up 22% year over year) for FY25 in its Environmental Data Fact Sheet. Reports a global average datacenter WUE of 0.27 L/kWh for FY25. View source →Energy Usage Reports total energy consumption (37.5M MWh in FY25) and a global average datacenter PUE of 1.17 for FY25. View source →Per-Facility Data Publishes datacenter PUE and WUE for three broad regions (Americas, Asia Pacific, EMEA) but not water or energy figures for individual facilities. View source →AI-Specific Data Does not publish measured water or energy figures for its AI workloads. The FY25 report cites Microsoft research that optimized large-scale inference can use under one watt-hour per query, but gives no figures for Copilot or OpenAI workloads. View source →Water Pledge Pledged to be water positive by 2030 by reducing water use and replenishing more than it withdraws. Reports replenishing over 14.2M cubic meters in FY25, exceeding its global withdrawals. View source → |
| Amazon (AWS) | Yes | Partial | Partial | No | Yes |
Water Usage Reports total water withdrawal for AWS data centers (9.4B liters in 2025) and a global WUE of 0.12 L/kWh. Does not report a separate water consumption figure. View source →Energy Usage Reports a global AWS data center PUE of 1.14 for 2025 and that electricity consumed is matched with renewable energy. Does not report a total energy consumption figure for AWS. View source →Per-Facility Data Publishes AWS fact sheets by location with water withdrawal, PUE and WUE (e.g., Virginia: 263M gallons withdrawn in 2025), but not figures for individual data centers. View source →AI-Specific Data No published water or energy data specific to AI/ML workloads. Checked 2026-10-30 Water Pledge Pledged that AWS will be water positive across its data centers by 2030. Reports being 75% of the way toward that goal at the end of 2025. View source → |
| Meta (LLaMA) | Yes | Yes | Yes | Partial | Yes |
Water Usage Reports total water withdrawal (5,637 ML) and consumption (3,123 ML, 2,974 ML from data centers) for 2024 in its Environmental Data Index. Reports an annual data center WUE of 0.19 L/kWh. View source →Energy Usage Reports total electricity consumption (18.4M MWh in 2024) and a data center PUE of 1.08 in its Environmental Data Index. View source →Per-Facility Data Publishes water withdrawal and electricity consumption for each data center (e.g., Prineville, OR: 328 ML withdrawn and 1.73M MWh consumed in 2024). View source →AI-Specific Data Llama model cards report training GPU hours, per-GPU power and estimated training emissions (e.g., Llama 4: 7.38M GPU hours). No water data and no inference energy data published. View source →Water Pledge Set a goal to become water positive in 2030, restoring 200% of water consumed in high water stress regions and 100% in medium water stress regions. Reports restoring over 1.6B gallons in 2024. View source → |
| Anthropic (Claude) | No | No | No | No | No |
Water Usage No sustainability report or water usage data published as of October 2026. Checked 2026-10-30 Energy Usage No energy consumption data published. Relies on AWS and Google Cloud infrastructure. View source →Per-Facility Data Not applicable — Anthropic does not operate its own data centers. Checked 2026-10-30 AI-Specific Data No per-model or per-query energy or water data published. Checked 2026-10-30 Water Pledge No water-related sustainability commitments published. Checked 2026-10-30 |
| Perplexity | No | No | No | No | No |
Water Usage No sustainability report or water usage data published as of October 2026. Checked 2026-10-30 Energy Usage No energy consumption data published. Relies on third-party cloud infrastructure (AWS, Azure, CoreWeave). View source →Per-Facility Data Not applicable — Perplexity does not operate its own data centers. Checked 2026-10-30 AI-Specific Data No per-query energy or water data published. Checked 2026-10-30 Water Pledge No water-related sustainability commitments published. Checked 2026-10-30 |
| OpenAI (ChatGPT) | Partial | Partial | No | Partial | Partial |
Water Usage CEO disclosed ~0.3 ml direct water per average ChatGPT query (mid-2025), but no total water consumption figures or methodology published. Excludes indirect water from electricity generation. View source →Energy Usage CEO disclosed ~0.34 Wh per average ChatGPT query (mid-2025). No total energy consumption figures published. Runs on Microsoft Azure infrastructure. View source →Per-Facility Data Does not operate its own data centers (runs on Microsoft Azure). No facility-level data published. Stargate project campuses are under construction. View source →AI-Specific Data Published per-query energy and water estimates for ChatGPT via CEO statement, but figures are unverified, exclude training costs, and definition of 'average query' is unclear. View source →Water Pledge Stargate Community plan commits to prioritizing closed-loop or low-water cooling at Stargate campuses. No company-wide water positive or replenishment pledge published by OpenAI. View source → |
| xAI (Grok) | No | No | No | No | Partial |
Water Usage No water usage data published. Memphis Colossus facility draws ~1.3M gallons/day from the aquifer per reporting by environmental groups, but xAI has not disclosed figures itself. View source →Energy Usage No energy consumption data published. Operates its own data centers with gas turbines and battery storage, but has not disclosed total energy usage. View source →Per-Facility Data Operates the Colossus facility in Memphis, TN, but publishes no environmental data for it. Environmental impact information comes from third-party investigations and permit filings. View source →AI-Specific Data No per-model or per-query energy or water data published for Grok models. Checked 2026-10-30 Water Pledge Committed to building an $80M wastewater recycling plant to supply its Memphis facility instead of aquifer water. Construction paused in April 2026; the company has said it will resume no later than Q1 2027. No company-wide water pledge. View source → |
| Mistral (Le Chat) | Yes | Yes | No | Yes | No |
Water Usage Published total water consumption for Mistral Large 2: 281,000 m3 over 18 months (training + inference). Conducted ISO 14040/44-compliant lifecycle analysis with Carbone 4 and ADEME. View source →Energy Usage Published energy data and greenhouse gas emissions (20.4 kt CO2e for Mistral Large 2 training + 18 months of use) via comprehensive lifecycle analysis. View source →Per-Facility Data Operates a 40 MW facility in Essonne, France and is building data centers in Sweden, but does not publish per-facility environmental breakdowns. View source →AI-Specific Data Published per-query environmental impact: 45 ml water, 1.14g CO2e, and 0.16 mg antimony-equivalent resource depletion per 400-token Le Chat response. View source →Water Pledge No water positive or replenishment pledge. Committed to ongoing environmental reporting and contributing to international AI sustainability standards. Checked 2026-10-30 |
| Cohere | No | No | No | No | No |
Water Usage No sustainability report or water usage data published as of October 2026. Checked 2026-10-30 Energy Usage No energy consumption data published. Published a policy primer on AI energy efficiency but disclosed no data about its own operations. Relies on third-party cloud infrastructure. View source →Per-Facility Data Not applicable — Cohere does not operate its own data centers. Checked 2026-10-30 AI-Specific Data No per-model or per-query energy or water data published. Checked 2026-10-30 Water Pledge No water-related sustainability commitments published. Checked 2026-10-30 |
| Alibaba (Qwen) | Yes | Yes | No | No | No |
Water Usage Reports total water consumption annually (17,464,161 m³ in FY2026, the year ended March 31, 2026, up from 13,479,552 m³ in FY2025). Reports an average WUE of 1.198 L/kWh for Alibaba Cloud's self-built data centers in FY2026: 0.205 for mainly air-cooled sites and 1.799 for mainly evaporative-cooled sites. View source →Energy Usage Reports total energy consumption (9,707,757 MWh in FY2026) and an average PUE of 1.187 for Alibaba Cloud's self-built data centers and 1.224 for its leased data centers in China. Reports clean electricity at 62.2% of the group's electricity mix. View source →Per-Facility Data Reports water and energy metrics as averages across self-built and leased data centers. Names individual sites (e.g., water reclaimed at bases including Ulanqab) but publishes no water or energy figures for individual facilities. View source →AI-Specific Data No published water or energy data for Qwen models or other AI workloads. Checked 2026-10-31 Water Pledge No water replenishment, water positive or quantified water target published. The FY2026 report describes site-specific cooling choices and water-efficiency practices for its data centers. View source → |
| Baidu (ERNIE) | Yes | Yes | No | No | No |
Water Usage Reports total water withdrawal (2,183,887.2 tons in 2025), discharge (636,410.9 tons) and consumption (1,547,476.4 tons), with 2023 and 2024 figures for comparison. Reports an average annual WUE of 1.65 for its self-built data centers in 2025 (the report gives no unit). View source →Energy Usage Reports total energy consumption (1,273,970.0 MWh in 2025) and an average annual PUE of 1.20 for its self-built data centers. Reports 106.16 million kWh of renewable energy consumption in 2025. View source →Per-Facility Data Describes measures at named data centers (e.g., a wastewater reuse program at Yangquan that cut cooling-system water use by 20%) but publishes no water or energy figures for individual facilities. View source →AI-Specific Data No published water or energy data for ERNIE models or other AI workloads. Checked 2026-10-31 Water Pledge States a commitment to reduce water usage and says it sets annual WUE targets for its data centers, but publishes no quantified water target, replenishment or water positive pledge. View source → |
| ByteDance (Doubao) | No | Partial | No | No | No |
Water Usage No sustainability report or water usage data published as of October 2026. Checked 2026-10-31 Energy Usage Reports surpassing 42% carbon-free electricity globally and 98.6% across UK and EU operations in 2025, with a goal of 100% carbon-free electricity by 2030. Does not report total energy consumption or PUE. View source →Per-Facility Data Says some newer data centers, such as its Norwegian data center, run on 100% carbon-free electricity, but publishes no water or energy figures for individual facilities. View source →AI-Specific Data No per-model or per-query energy or water data published for Doubao or other AI models. Checked 2026-10-31 Water Pledge No water-related sustainability commitments published. Its published climate goals cover carbon neutrality and carbon-free electricity by 2030. View source → |
| DeepSeek | No | No | No | No | No |
Water Usage No sustainability report or water usage data published as of October 2026. Checked 2026-10-31 Energy Usage No energy consumption data published. Checked 2026-10-31 Per-Facility Data No facility-level data published. Checked 2026-10-31 AI-Specific Data No energy or water figures published for its models. The DeepSeek-V3 technical report gives training compute in GPU hours (2.788M H800 GPU hours) and an estimated rental cost, but no energy, emissions or water figures. View source →Water Pledge No water-related sustainability commitments published. Checked 2026-10-31 |
| Tencent (Hunyuan) | Yes | Yes | No | No | Partial |
Water Usage Reports total water consumption annually (12,151,404.1 tonnes in 2025, up from 10,370,588.3 tonnes in 2024) and states that its water comes from municipal supply. Reports a WUE of 1.50 L/kWh for its owned data centers in 2025. View source →Energy Usage Reports total energy consumption (8,715,193.3 MWh in 2025) and an average PUE of 1.246 for owned data centers. Reports green electricity at 48.5% of consumption in 2025. View source →Per-Facility Data Reports water and energy metrics in aggregate and as averages across owned data centers. No figures for individual facilities. View source →AI-Specific Data No published water or energy data for Hunyuan models or other AI workloads. Checked 2026-10-31 Water Pledge Targets a 20% improvement in the WUE of its owned data centers by 2030 from a 2025 base year, and reports meeting a 2025 target to cut water use per employee in its owned Chinese Mainland office buildings by 15%. No water replenishment or water positive pledge. View source → |