Environmental impact of artificial intelligence

AI-related energy use arises at multiple stages, including model training, fine-tuning, inference, storage, networking, and supporting infrastructure such as cooling and power conversion. Published estimates of energy use per AI request vary widely across models, tasks and measurement methods. In that benchmark, simple classification tasks consumed about 0.002–0.007 Wh per prompt on average (about 9% of a smartphone charge for 1,000 prompts), while text generation and text summarisation each used about 0.05 Wh per prompt; image generation averaged 2.91 Wh per prompt, and the least efficient image model in the study used 11.49 Wh per image (roughly equivalent to half a smartphone charge). Comparisons between AI systems and human labour for specific tasks have produced mixed results and remain sensitive to assumptions about output quality, workload and system boundaries. A 2024 study in Scientific Reports reported 130 to 2900 times lower estimated carbon emissions for selected AI systems than for human writers and illustrators under its assumptions. A later Scientific Reports paper reported a counterexample for programming tasks under its assumptions, finding 5 to 19 times higher estimated emissions for the evaluated AI system than for human programmers on the benchmark used in that study. System level Energy use and efficiency }} AI electricity intensity depends not only on model architecture but also on hardware and facility efficiency. Data-centre operators commonly report Power usage effectiveness (PUE), which measures the ratio of total facility energy to IT equipment energy; a lower PUE indicates less overhead energy for cooling and other supporting infrastructure. The International Energy Agency has also reported that data centres remain a relatively small share of global electricity use overall, but that their local effects can be much more pronounced because demand is geographically concentrated. Accounting methods that include upstream or embodied impacts, such as hardware manufacture and facilities construction, can materially affect estimates of AI-related emissions. Decisions and strategies by individual companies Large technology companies have reported that the expansion of AI and cloud infrastructure affects their sustainability targets, electricity demand, and resource use. Google, for example, attributed part of its emissions growth in 2023 to increased data-centre energy consumption and supply-chain emissions in its 2024 environmental report. Cloud and AI companies have also announced measures intended to reduce environmental impacts, including investment in more efficient hardware, low-carbon electricity procurement, alternative cooling systems, and water stewardship programmes. The extent, comparability, and third-party verification of such disclosures vary between firms and jurisdictions. == Water usage ==

Data centres can use water directly for cooling and indirectly through the water used in electricity generation, depending on the local energy mix. WUE does not by itself measure local water stress, source sustainability, or all upstream water impacts. Research on AI-specific water use has argued that the water footprint of AI systems can be difficult to observe and may vary substantially by location, cooling design, and electricity source. A 2025 Communications of the ACM article summarised methods for estimating AI water footprints and emphasised the distinction between water withdrawal and water consumption. A 2025 Reuters report, citing data from Verisk Maplecroft and NatureFinance, said that an average mid-sized data centre uses about 1.4 million litres of water per day for cooling and that Phoenix would experience a 32% increase in annual water stress if currently planned data centres come online. Water use also occurs upstream in semiconductor fabrication, which relies on large quantities of ultrapure water. == E-waste ==

AI systems depend on specialised computing hardware, and rapid turnover in servers and accelerators may contribute to rising e-waste. The World Health Organization has also identified e-waste as a growing environmental and public-health issue. A 2024 study in Nature Computational Science estimated that generative AI could add between 1.2 and 5 million tonnes of e-waste by 2030 under the scenarios examined by the authors. In the study's higher-end scenarios, this would represent up to 12% of projected global e-waste by 2030. The authors also estimated that circular-economy strategies along the generative-AI value chain could reduce AI-related e-waste generation by 16–86%. == Mining ==

AI hardware depends on complex supply chains for metals, minerals and manufactured components. UNCTAD has reported that the expansion of digital infrastructure increases demand for raw materials and raises environmental and distributional concerns linked to extraction, processing and manufacturing. Specialised chips used in AI systems can depend on supply chains involving critical minerals and other materials whose extraction and processing may have significant environmental and social effects. These impacts are not unique to AI, but may increase as demand for AI-related hardware grows. == Social impact and environmental justice ==

The environmental effects of AI-related infrastructure are not distributed evenly. Research on U.S. data centres has found that their environmental footprints vary by region and may intersect with local electricity systems, water availability and existing environmental burdens. == Climate solutions ==

Despite concerns about its environmental footprint, AI has been used in environmental and climate-related applications, including weather forecasting, Earth observation, and optimisation in transport and energy systems. In weather forecasting, peer-reviewed studies have reported strong results for some AI-based forecasting systems under specific evaluation frameworks. A 2023 Nature paper on Pangu-Weather reported strong medium-range forecasting performance relative to a leading numerical weather prediction system in the study's evaluation. AI has also been used in research on extreme weather and climate-event modelling. AI has also been proposed for mitigation-oriented optimisation. Google's Green Light project, for example, uses traffic data and machine learning to recommend traffic-signal timing adjustments intended to reduce stop-and-go traffic and associated emissions at intersections. Whether AI produces net environmental benefits at large scale remains an open question, because outcomes depend on deployment choices, rebound effects, additional infrastructure demand and the extent to which electricity and cooling systems are decarbonised. Conflict on the use of AI for environmental research There is ongoing debate over the balance between the possible environmental benefits of AI applications and the environmental costs of scaling AI systems. This includes discussion of transparency, efficiency, rebound effects, and the extent to which AI-related infrastructure growth may offset environmental gains from specific applications. == Policy and regulation ==

United States In the United States, proposals have been introduced to study and standardise reporting on AI's environmental impacts. The Artificial Intelligence Environmental Impacts Act of 2024 (S. 3732), introduced in the Senate in February 2024, would require a federal study on the environmental impacts of AI, direct the National Institute of Standards and Technology to convene a consortium on measurement and standards, and establish a voluntary reporting system. State Policy Local and state governments have looked to address environmental and infrastructural impacts. As of 2026, at least 27 states are considering or have put legislation related to data center development, with California, Ohio and Utah being the first to pass legislation. This legislation requires data center developers to bear the costs of new energy infrastructure. Some states are also pushing for legislation requiring data centers to report water use, an issue not addressed by the federal government. States are looking to focus on data collection related to data center water use. A 2025 study published by Nature Sustainability estimated that AI servers in the United States could require approximately 731 to 1,125 million cubic meters of water annually by the year 2030, with 24 to 44 million metric tons of carbon emissions. This study found that current growth trajectories are unlikely to meet net zero targets without policy. European Union In the European Union, the Energy Efficiency Directive introduced reporting obligations for large data centres. The European Commission has stated that a European database collects information relevant to the energy performance and water footprint of data centres, and that a delegated regulation sets out the information and key performance indicators for the reporting scheme. Germany Germany's national AI strategy includes sections on the environmental impacts of AI and on research into energy-efficient and sustainable AI applications. Italy Italy's national AI strategy documents include sustainability-related priorities and discuss AI applications in areas such as environment, infrastructure, and sustainable development goals. == See also ==