Artificial intelligence is increasingly seen as a powerful lever in the fight against climate change. It helps to model climate systems, optimize energy usage, forecast renewable energy supply, and detect environmental hazards. Yet at the same time AI itself consumes enormous amounts of energy, often powered by fossil fuels, and contributes measurable greenhouse gas emissions. This creates a paradox: a tool meant to save the planet may also be harming it.
One clear dimension of this paradox is the energy cost of training and running large AI models. For example, the training of GPT-3 has been estimated at 1,287 megawatt-hours (MWh) of electricity and over 500 metric tons of carbon dioxide (CO₂) emissions; which is roughly equivalent to the annual emissions of more than 100 gasoline cars (Columbia’s State of the Planet, 2023). Inference, meaning the phase when a model responds to user queries, also adds ongoing carbon emissions and energy consumption (Columbia’s State of the Planet, 2023; Stanford researchers, 2024).
At the same time there are promising opportunities where AI does appear to reduce emissions and environmental harm. For instance, AI tools are being applied to improve energy efficiency in buildings. A case study in Manhattan found that AI-driven HVAC (heating, ventilation, air conditioning) systems cut energy consumption of a large building by nearly 16 percent which in turn reduced carbon emissions by tens of metric tons per year (Time, 2024). AI is similarly used in climate modeling and renewable energy forecasting to help policymakers design more effective mitigation strategies (NenPower, 2024).
The moral and policy question becomes how to balance these costs and benefits. Should governments regulate AI development to limit environmental harm even if that slows innovation? What standards of transparency, metrics of “carbon cost,” and accountability should apply? There is also the risk of greenwashing, when AI providers or users promote small efficiency gains while ignoring the larger footprint of expansive deployment. Additionally there are equity issues: data centers and AI infrastructure tend to cluster in places with cheaper electricity, which may have high carbon intensity, or limit resource access in less developed regions.
Some researchers and developers are already proposing technical and policy solutions. On the technical front there are efficiency techniques such as model pruning or distillation, better scheduling of training (using times when grid power is cleanest), using more energy-efficient hardware, or developing algorithms that are “carbon aware” (Forbes, 2025; University of Michigan work, 2023). On the policy side governments can mandate environmental impact disclosure for large AI deployments, provide incentives for use of renewables in data centers, adopt regulations to limit carbon intensity, and support research into “Green AI” (Stanford, 2024).
The crux is that AI’s potential to help with environmental issues, such as climate modeling, disaster response, and renewable energy optimization is real. But it cannot be separated from the consequences of its own environmental burden. Morally, it seems irresponsible to deploy AI at scale without internalizing its climate costs. Practically, failing to do so could undermine public trust, provoke regulatory backlash, and worsen climate change even as we try to combat it.
AI may be a high-powered tool in our climate toolbox. For it to help more than it harms we need rigorous policies, transparency, ethical design, and constant attention to environmental externalities. Otherwise, the very technologies we hope will save us may instead deepen the crises we seek to resolve.
Sources
University of Michigan, Chowdhury, M., & team. (2023, April 17). Optimization could cut the carbon footprint of AI training by up to 75%. University of Michigan Engineering. https://ece.engin.umich.edu/stories/optimization-could-cut-the-carbon-footprint-of-ai-training-by-up-to-75 ece.engin.umich.edu
Columbia University. (2023, June 9). AI’s growing carbon footprint. State of the Planet. https://news.climate.columbia.edu/2023/06/09/ais-growing-carbon-footprint State of the Planet
Stanford University. (2024). AI’s carbon footprint problem. Stanford Doerr School of Sustainability. https://sustainability.stanford.edu/news/ais-carbon-footprint-problem Stanford Doerr School of Sustainability
Time. (2024). How AI is making buildings more energy-efficient. Time Magazine. https://time.com/7201501/ai-buildings-energy-efficiency/ TIME
NenPower. (2024). What role does AI play in climate modeling for renewable energy? NenPower Blog. https://nenpower.com/blog/what-role-does-ai-play-in-climate-modeling-for-renewable-energy/ NenPower
Forbes Technology Council. (2025, May 21). 19 practical ways to reduce AI’s environmental impact. Forbes. https://www.forbes.com/councils/forbestechcouncil/2025/05/21/19-practical-ways-to-reduce-ais-environmental-impact/ Forbes
Leave a Reply