{"id":"theaimeters-environmental-impact","title":"TheAIMeters environmental impact data pack","description":"Versioned annualized estimates for global AI electricity consumption, operational water consumption and operational carbon emissions.","figureType":"modelled_estimate","reviewedOn":"2026-11-12","methodologyPath":"/methodology","sourcesPath":"/sources","reuse":{"permission":"These published estimates and their JSON or CSV metadata may be reused with attribution.","attribution":"Credit TheAIMeters and link to the relevant methodology page.","exclusions":"This permission does not extend to site articles, images, branding or visual design."},"indicators":[{"id":"electricity","label":"AI electricity consumption","description":"Modelled global annualized electricity consumption associated with AI workloads.","canonicalMeterSlug":"electricity-ai-this-year","indicatorPath":"/ai-electricity-consumption","methodologyPath":"/methodology/electricity-ai-this-year","methodologyId":"electricity","figureType":"modelled_estimate","scope":"global","period":"annualized","evidenceYear":2025,"estimate":{"lower":100000000000,"central":150000000000,"upper":200000000000,"unit":"kWh"},"rangeBasis":"Indicative range synthesized from recent public estimates of AI-focused datacenter power demand and capacity. It is not a statistical confidence interval.","confidenceLevel":"medium","reviewedOn":"2026-11-12","modelVersion":"2026.11","sourceNote":"Modelled global annualized estimate using a 2025 evidence base. Central estimate: 150 TWh/year; indicative range: 100–200 TWh/year. Not official telemetry or a confidence interval.","sourceIds":["iea-key-questions-2026","stanford-ai-index-2026","de-vries-gao-2025"],"sources":[{"id":"iea-key-questions-2026","name":"IEA — Key Questions on Energy and AI (2026)","url":"https://www.iea.org/reports/key-questions-on-energy-and-ai","category":"research"},{"id":"stanford-ai-index-2026","name":"Stanford HAI — AI Index Report 2026","url":"https://hai.stanford.edu/ai-index/2026-ai-index-report","category":"research"},{"id":"de-vries-gao-2025","name":"de Vries and Gao — The environmental footprint of AI (2025)","url":"https://doi.org/10.1016/j.patter.2025.101430","category":"research"}],"revisions":[{"version":"2026.04","effectiveOn":"2026-04-30","centralValue":60000000000,"unit":"kWh","reason":"initial_baseline","summary":"Initial conservative AI-attributable electricity baseline."},{"version":"2026.11","effectiveOn":"2026-11-12","centralValue":150000000000,"unit":"kWh","reason":"evidence_recalibration","summary":"Recalibrated against newer public evidence on AI-focused datacenter electricity demand and installed capacity."}]},{"id":"water","label":"AI operational water consumption","description":"Modelled global annualized operational water footprint associated with AI workloads.","canonicalMeterSlug":"water-ai-this-year","indicatorPath":"/ai-water-consumption","methodologyPath":"/methodology/water-ai-this-year","methodologyId":"water","figureType":"modelled_estimate","scope":"global","period":"annualized","evidenceYear":2025,"estimate":{"lower":312500000000,"central":540000000000,"upper":764600000000,"unit":"L"},"rangeBasis":"Published estimate range for the operational water footprint of AI systems, including direct cooling water and indirect water associated with electricity generation. It is not a statistical confidence interval.","confidenceLevel":"low","reviewedOn":"2026-11-12","modelVersion":"2026.11","sourceNote":"Modelled global annualized operational water footprint using a 2025 evidence base. Central estimate: 540 billion L/year; indicative range: 312.5–764.6 billion L/year. Not official telemetry or a confidence interval.","sourceIds":["de-vries-gao-2025","li-making-ai-less-thirsty","iea-key-questions-2026"],"sources":[{"id":"de-vries-gao-2025","name":"de Vries and Gao — The environmental footprint of AI (2025)","url":"https://doi.org/10.1016/j.patter.2025.101430","category":"research"},{"id":"li-making-ai-less-thirsty","name":"Li et al. — Making AI Less Thirsty","url":"https://arxiv.org/abs/2304.03271","category":"research"},{"id":"iea-key-questions-2026","name":"IEA — Key Questions on Energy and AI (2026)","url":"https://www.iea.org/reports/key-questions-on-energy-and-ai","category":"research"}],"revisions":[{"version":"2026.04","effectiveOn":"2026-04-30","centralValue":30000000000,"unit":"L","reason":"initial_baseline","summary":"Initial conservative baseline using a limited water-intensity factor."},{"version":"2026.11","effectiveOn":"2026-11-12","centralValue":540000000000,"unit":"L","reason":"evidence_recalibration","summary":"Expanded to a total operational footprint consistent with the stated scope, including direct and electricity-linked indirect water use."}]},{"id":"carbon","label":"AI operational carbon emissions","description":"Modelled global annualized operational carbon emissions associated with AI electricity use.","canonicalMeterSlug":"co2-emitted-this-year","indicatorPath":"/ai-carbon-emissions","methodologyPath":"/methodology/co2-emitted-this-year","methodologyId":"co2","figureType":"modelled_estimate","scope":"global","period":"annualized","evidenceYear":2025,"estimate":{"lower":32600000000,"central":52500000000,"upper":79700000000,"unit":"kg CO2e"},"rangeBasis":"Published estimate range for operational emissions associated with AI electricity use. It is not a statistical confidence interval and excludes a complete lifecycle inventory of hardware and construction.","confidenceLevel":"low","reviewedOn":"2026-11-12","modelVersion":"2026.11","sourceNote":"Modelled global annualized operational emissions using a 2025 evidence base. Central estimate: 52.5 Mt CO2e/year; indicative range: 32.6–79.7 Mt/year. Not official telemetry or a confidence interval.","sourceIds":["de-vries-gao-2025","iea-key-questions-2026","stanford-ai-index-2026"],"sources":[{"id":"de-vries-gao-2025","name":"de Vries and Gao — The environmental footprint of AI (2025)","url":"https://doi.org/10.1016/j.patter.2025.101430","category":"research"},{"id":"iea-key-questions-2026","name":"IEA — Key Questions on Energy and AI (2026)","url":"https://www.iea.org/reports/key-questions-on-energy-and-ai","category":"research"},{"id":"stanford-ai-index-2026","name":"Stanford HAI — AI Index Report 2026","url":"https://hai.stanford.edu/ai-index/2026-ai-index-report","category":"research"}],"revisions":[{"version":"2026.04","effectiveOn":"2026-04-30","centralValue":21000000000,"unit":"kg CO2e","reason":"initial_baseline","summary":"Initial conservative electricity-related emissions baseline."},{"version":"2026.11","effectiveOn":"2026-11-12","centralValue":52500000000,"unit":"kg CO2e","reason":"evidence_recalibration","summary":"Recalibrated against newer public estimates of operational AI emissions while retaining an explicit global-average emissions factor."}]}]}