Environmental Impact Of Artificial Intelligence

The environmental impact of artificial intelligence includes substantial energy consumption for training and using deep learning models, and the related carbon footprint and water usage. Some scientists have suggested that artificial intelligence (AI) may also provide solutions to environmental problems.

Carbon footprint

AI has a significant carbon footprint due to growing energy usage, especially due to training and usage. Researchers have argued that the carbon footprint of AI models during training should be considered when attempting to understand the impact of AI. One study suggested that by 2027, energy costs for AI could increase to 85–134 Twh, nearly 0.5% of all current electricity usage. Training large language models (LLMs) and other generative AI generally requires much more energy compared to running a single prediction on the trained model. Using a trained model repeatedly, though, may easily multiply the energy costs of predictions. The computation required to train the most advanced AI models doubles every 3.4 months on average, leading to exponential power usage and resulting carbon footprint. Additionally, artificial intelligence algorithms running in places predominantly using fossil fuels for energy will exert a much higher carbon footprint than places with cleaner energy sources. These models may be modified for less environmental impacts at the cost of accuracy, emphasizing the importance of finding the balance between accuracy and environmental impact.

Training a large AI model requires enormous amounts of energy. It is estimated that training a whole AI model produces around 626 000 lbs (283 Tons) of carbon dioxide. It is the equivalent of 300 round-trip flights between New York and San Francisco, or nearly 5 times the lifetime emissions of the average car.

BERT, a language model trained in 2019, required "the energy of a round-trip transcontinental flight" to train. GPT-3 released 552 metric tons of carbon dioxide into the atmosphere during training, "the equivalent of 123 gasoline-powered passenger vehicles driven for one year". Much of the energy cost is due to inefficient model architectures and processors. One model named BLOOM, from Hugging Face, trained with more efficient chips and, therefore, only released 25 metric tons of CO₂. Incorporating the energy cost of manufacturing the chips for the system doubled the carbon footprint, to "the equivalent of around 60 flights between London and New York." Operating BLOOM daily was estimated to release the equivalent carbon footprint as driving 54 miles.

Algorithms which have lower energy costs but run millions of times a day can also have significant carbon footprints. The integration of AI into search engines could multiply energy costs significantly, with some estimates suggesting energy costs rising to nearly 30 billion kWh per year, an energy footprint larger than many countries. Another estimate found that integrating ChatGPT into every Google search query would use 10 TWh each year, the equivalent yearly energy usage of 1.5 million European Union residents.

Once the model is trained, it consumes significantly less energy, however it still requires a high amount of electricity. Researchers have estimated that a ChatGPT query consumes about five times more electricity than a simple web search. In June 2025, OpenAI executive Sam Altman stated that the average ChatGPT query used about of electricity and of water.

Increased computational demands from AI caused both increased water and energy usage, leading to significantly more demands on the grid. Due to increased energy demands from AI-related projects, coal-fired plants in Kansas City and West Virginia pushed back closing. Other coal-fired plants in the Salt Lake City region have pushed back retirement of their coal-fired plants by up to a decade. Environmental debates have raged in both Virginia and France about whether a "moratorium" should be called for additional data centers. In 2024 at the World Economic Forum, Sam Altman gave a speech in which he said that the AI industry can only grow if there is a major technology breakthrough to increase energy development.

In 2024, Google failed to reach key goals from their net zero plan as a result of their work with AI, and had a 48% increase in greenhouse gas emission attributable to their growth in AI. A request made via ChatGPT, an AI-based virtual assistant, uses 10 times as much electricity as a Google Search. Microsoft and Meta had similar increases in their carbon footprint, similarly attributed to AI. Carbon footprints of AI models depends on the energy source used, with data centers using renewable energy lowering their footprint. Many tech companies claim to offset energy usage by buying energy from renewable sources, though some experts argue that utilities simply replace the claimed renewable energy with increased non-renewable sources for their other customers. Analysis of the carbon footprint of AI models remains difficult to determine, as they are aggregated as part of datacenter carbon footprints, and some models may help reduce carbon footprints of other industries, or due to differences in reporting from companies.

Some applications of ML, such as for fossil fuel discovery and exploration, may worsen climate change. Use of AI for personalized marketing online may also lead to increased consumption of goods, which could also increase global emissions.

Energy use and efficiency

AI chips, (i.e. GPUs) use more energy and emit more heat than traditional CPU chips. AI models with inefficiently implemented architectures, or trained on less efficient chips may use more energy. Since the 1940's the energy efficiency of computation has doubled every 1.6 years. Some skeptics argue that improvements of AI efficiency may only increase AI usage and therefore carbon footprint due to Jevons paradox.

In September 2024, Microsoft announced an agreement with Constellation Energy to re-open the Three Mile Island nuclear power plant to provide Microsoft with 100% of all electric power produced by the plant for 20 years. Reopening the plant, which suffered a partial nuclear meltdown of its Unit 2 reactor in 1979, will require Constellation to get through strict regulatory processes which will include extensive safety scrutiny from the US Nuclear Regulatory Commission. If approved (this will be the first ever US re-commissioning of a nuclear plant), over 835 megawatts of power – enough for 800,000 homes – of energy will be produced. The cost for re-opening and upgrading is estimated at $1.6 billion (US) and is dependent on tax breaks for nuclear power contained in the 2022 US Inflation Reduction Act. The US government and the state of Michigan are investing almost $2 billion (US) to reopen the Palisades Nuclear Reactor on Lake Michigan. Closed since 2022, the plant is planned to be reopened in October 2025. The Three Mile Island facility will be renamed the Crane Clean Energy Center after Chris Crane, a nuclear proponent and former CEO of Exelon who was responsible for Exelon spinoff of Constellation.

In 2025, Microsoft unveiled plans to invest $80 billion in the development and expansion of data centers designed to support AI technologies. These facilities, critical to the advancement of AI, depend on vast networks of interconnected chip clusters and significant electrical power to operate efficiently.

The International Energy Agency (IEA) released its 2025 Electricity Analysis and Forecast in February 2025, projecting 4% growth in global electricity demand over the next three years due to data center growth, increased industrial production, increased electrification, and increased use of air conditioning. By 2027, the US's energy consumption is expected to grow by an amount equivalent to California's entire annual power usage, largely driven by energy-hungry data centers and manufacturing operations. In 2024, U.S. electricity generation rose by 3%, with data centers emerging as a dominant force behind the increase. The trend is expected to continue as semiconductor and battery manufacturing plants ramp up operations, further intensifying demand.

In 2024, a US public policy group reported that AI and other technologies and industries poised to dominate the global economy are characterized by their high electricity demands. As such, the foundation of US energy strategy and policymaking will be to prioritize the reliable and abundant provision of electricity to support these critical sectors, which are needed to maintain the US economic and technological leadership in the twenty-first century. The rapid proliferation of AI has created unprecedented demand for electrical power, presenting a major obstacle to the sector’s growth. E.g., in Northern Virginia, the largest global hub for AI data centers, the timeline for connecting bigger facilities—those requiring over 100 megawatts of power—to the electrical grid has extended to seven years, highlighting the strain on the energy infrastructure and the challenge of meeting AI’s escalating power needs. Across the United States, utilities are experiencing the most substantial surge in electrical demand in decades. This strain is directly contributing to longer wait times for grid connections, complicating efforts to maintain the country’s technological leadership in AI. The significance of these energy challenges extends beyond logistics. ''A New York Times'' editorial emphasized the critical role of energy infrastructure, stating that "Electricity is more than just a utility; it’s the bedrock of the digital era. If the United States truly wants to secure its leadership in A.I., it must equally invest in the energy systems that power it."

Globally, the electricity consumption of data centers rose to 460 terawatts in 2022. This would have made data centers the 11th largest electricity consumer in the world, between the nations of Saudi Arabia (371 terawatts) and France (463 terawatts), according to the Organization for Economic Co-operation and Development.

Water usage

Cooling AI servers can demand large amounts of fresh water which is evaporated in cooling towers. In fact, data centers housing AI are globally expected to consume six times more water than the country of Denmark. By 2027, AI may use up to 6.6 billion cubic meters of water. One professor has estimated that an average session on ChatGPT, with 10–50 responses, can use up to a half-liter of fresh water. Training GPT-3 may have used 700,000 liters of water, equivalent to the water footprint of manufacturing 320 Tesla EVs.

One data center that Microsoft had considered building near Phoenix, due to increasing AI usage, was likely to consume up to 56 million gallons of fresh water each year, equivalent to the water footprints of 670 families. Microsoft may have increased water consumption by 34% due to AI, while Google increased its water usage by 20% due to AI. Due to their Iowa data center cluster, Microsoft was responsible for 6% of the freshwater use in a local town.

A possible solution for reducing water consumption is to build data centers in colder countries that can offer a natural cooling system. For example, Facebook (now Meta) built a data center in Luleå, northern Sweden, in 2011. Google invested one more billion euros into the expansion of its data centre campus in Hamina in Finland in 2024, for a total of 4.5 billion euros invested in the site which also uses seawater for cooling its servers.

E-waste

E-waste due to production of AI hardware may also contribute to emissions. The rapid growth of AI may also lead to faster deprecation of devices, resulting in hazardous e-waste. Among the 62 million tonnes (Mt) of e-waste produced in 2022, less than one quarter of the total mass was properly recycled. Worldwide, the annual generation of e-waste is rising by 2.6 million tonnes annually, on track to reach 82 million tonnes by 2030, a further 33% increase from the 2022 figure. AI could have an important role because it is expected to add 1.2 million to 5 million metric tons of e-waste in total by 2030, which would represent up to 12% of global e-waste. Some applications of AI, such as for robot recycling, may reduce e-waste.

Mining

Large-scale AI is typically housed in data centres, which can exact heavy tolls on the planet. Their electronics rely on huge amounts of raw materials: making a 2 kg computer requires 800 kg of raw materials. Also, the microchips that power AI require rare earth elements, often mined in environmentally destructive ways.

Climate solutions

AI has significant potential to help mitigate effects of climate change, such as through better weather predictions, disaster prevention and weather tracking. Some climate scientists have suggested that AI could be used to improve efficiencies of systems, such as renewable-energy systems. Google has claimed AI could help mitigate some effects of climate change such as predicting floods or making traffic more efficient. Some algorithms may help predict the impacts of more severe hurricanes, measure the melting of polar ice, deforestation, and help monitor emissions from sources. AI is used to model and analyze extreme weather events such as floods, droughts, and heatwaves by processing vast amounts of climate data and identifying patterns that may not be easily detected by traditional methods. These advanced predictive capabilities help governments, emergency responders, and policymakers improve disaster preparedness, optimize resource allocation, and develop early warning systems to mitigate the impact of natural disasters on communities. AI is also being applied to genetic engineering. An AI tool called Social LEAP Estimates Animal Poses (SLEAP) is being used to improve the carbon sequestration of plant root systems. One machine learning project, the Open Catalyst project, has been used to identify "suitable low-cost electrocatalysts" for battery storage of renewable energy sources. AI may also improve the efficiencies of supply chains and productions for environmentally detrimental industries such as food and fast fashion. However, as yet there are no widely accepted frameworks which evaluate AI systems' total climate impacts, factoring in both costs and benefits.

Policy and regulation

United States

The environmental impacts of AI have been a blindspot in the range of AI legislation proposed in the US Congress. As of November 2024, the Artificial Intelligence Environmental Impacts Act of 2024 introduced in the Senate by Massachusetts Senator Ed Markey was the only federal bill to make environmental recommendations for the use of AI. The Act would have required the administrators of the Environmental Protection Agency, the National Institute of Standards and Technology, and the Department of Energy to study the environmental effects of AI's development, deployment, and post-deployment and enact a voluntary reporting system for AI-related environmental impacts. The bill has not been reintroduced in the 119th Congress.

In lieu of federal legislation on the subject, certain state governments have introduced policy on the environmental cost of AI. Virginia is considering legislation requiring data centers to submit water use estimates, reflecting growing concerns about resource consumption, sustainability, and land use. For instance, Virginia's Joint Legislative Audit and Review Commission (JLARC) has recommended that data centers report their energy and water usage to address the strain these facilities place on infrastructure and resources. Another Virginia bill proposed a mandatory review and approval process from the State Corporation Commission (SCC) for data center developments exceeding 100 megawatts to ensure grid reliability. However, the House Labor and Commerce Committee unanimously voted against the bill, expressing concerns that it might deter data center investments in the state. Additionally, House Bill 2035, introduced in the Virginia General Assembly, would require data centers to report quarterly on water and energy use to the Department of Environmental Quality, with the information made publicly accessible.

European Union

The European Union (EU) intends to regulate the environmental impact of artificial intelligence on multiple levels of government. The European Green Deal (EGD), set forth by the European Commission and approved in 2020, states its intention to utilize AI and other information and communication technology (ICT) to advance sustainability goals. Partnerships between the Commission and various European environmental and IT groups culminated in the development of th
Green Deal Data Space
(GDDS). As part of the European Data Portal, the GDDS project aims to aggregate cross-sectorial data in environmental and climate science to support the policy goals set forth in the EGD. AI agents can use the portal to find trends in the data and make recommendations to policymakers.

Policy research undertaken at the request of the European Commission recommends the EU take a bolder stance against the environmental costs of AI. One ethics report advocates that AI systems "should take into account the environment, including other living beings, and their social and societal impact should be carefully considered." Other reports by EU sources and independent watchdogs point to the lack of environmental considerations in AI model prohibitions set forth in the European Artificial Intelligence Act, and advocate for the assessment of environmental risks posed by the proliferation of AI systems. A 2022 case study recommends the EU restrict market access for AI systems that fail to implement global emissions monitoring or reduction strategies, along with mandated efficiency improvements, industry-wide sustainability reporting, and standardized life-cycle assessments (LCAs).

EU member states maintain individualized national AI strategies, many of which include sustainability goals.

France

France devises general environmental priorities in its national AI strategy report. Notably, it advocates that AI and data system development should be sustainably designed from the onset to support "the ecological transition of the European cloud industry." Furthermore, the report advocates for the publication of "ecological data" to promote AI-driven environmental solutions. The report also includes France's intention to support the adoption of AI for a more efficient grid and renewable energy transition.

Germany

Germany published its national AI strategy in December 2020 which includes dedicated sections on the environmental impacts of AI. These begin with the federal government's intention to thoroughly research and develop AI systems that can be used to promote energy efficiency, conservation, a circular economy, partnerships with higher education, natural resource management, and progress toward United Nations Sustainable Development Goals (SDGs). A subsequent section emphasizes the need for a reduction of energy consumption and a standardized environmental impact assessment (EIA) to create a net carbon-negative AI ecosystem.

Germany also operates the "AI Lighthouses" program which issues grants directly to businesses, non-profits, and researchers utilizing AI to develop environmental solutions. As of 2024, the German German Federal Environmental Ministry (BMUv) has disbursed upwards of 70 million euros in funding through the initiative.

Italy

The Italian national strategy for AI, crafted by a dedicated working group appointed by the Ministry of Economic Development (MISE), aims to leverage Italy's advantage in AI research and development to regain momentum in achieving its SDGs. The report principally advocates for government-backed stimulus in the development of AI for sustainability.

See also

* Environmental impact of bitcoin
* Environmental impact of computers

References

{{Reflist

Artificial intelligence
Human impact on the environment