What is this Estimator for? 

This is designed to help you estimate the environmental footprint of AI use. These calculations and sources are very complex and data are highly variable, so there will probably be some errors. 

• What types of AI use the most energy and have the biggest carbon footprint? 

• How is this equivalent to other forms of energy use? 

• How could we use this information to help us choose when and how to use AI? 

• How could we use this information to help us reduce the impacts of AI on the environment? 

Find out more about how this app was created on the Methods & About page. Some alternative versions are on the Alternative Apps page, made with Perplexity Pro & Qwen. 

• Disclaimer: This page and the app are designed for educational illustration only. With so many variables and changes to AI models, data can change quickly. This is not for serious technical use, but could support classroom conversations around AI use. If you can make a better, more reliable version of this, please go ahead and share! 

Why Does AI Have Such a Big Environmental Footprint? 

AI needs massive amounts of computing power - imagine thousands of powerful computers all working at once. When we "train" AI (teach it to do things like chat or create images), these computers work hard, using lots of electricity. They also need to be cooled, which has a high water demand. 

Think of it like baking cookies: training AI is like figuring out the perfect recipe through hundreds of attempts (using lots of energy), while using AI after training is like following the recipe once you know it (using less energy, but still significant). Even after training, simple tasks like generating one AI image uses about as much energy as charging your smartphone (MIT Technology Review, 2023).

The environmental impact gets bigger because most of this electricity still comes from fossil fuels like coal and natural gas. At the same time, more and more AI tools are being created, AI is being included in more of the things we use every day and there is more demand for AI. This means ever-increasing demands for computing power, electricity and water - not to mention the precious metals and other materials needed to create the computers. 

So what can we do about it?

We can try to be aware of our impacts of AI use, and make mindful, informed choices about the best time to use AI tools: 

• Do we really need to use AI for this task? 

• Am I missing the point of learning if I use AI for this? 

• If I need to use AI, how can I learn to use more efficiently, for better results and a reduced impact? 

How about using these tools or your own research to estimate the carbon footprints of your class over year, and develop a community engagement project to offset those impacts? 

• Could you plant some trees? 

• Raise awareness about making sensible choices? 

• Purchase Green Energy Certificates (China)/Renewable Energy Certificates (USA)? 

Here's another way to visualise your impacts, with simpler data. 

What about your weekly footprint? Think about how often you use different tools. How might you make more informed decisions? 

Research Sources

Research discovery was carried out using DeepSeek-R1 with R1 Reasoning and Search functions, alongside Perplexity Pro, and supplemented with further search. You can access these sources to learn more. 

Environmental Impacts of AI: Energy and Water

Lots of research papers and articles have been published to work out the per-use impacts of AI tools. With so many variables, this is a complex problem. Many of the sources below might be interesting reading for your projects. Where we know that AI use has a lot of environmental impacts, some use-cases might be counter-intuitive. For example, in the creation of this app and site, AI assistance for coding and research saved many hours of human and computer footprint. 

• Audacia (2023). Sustainable AI: 3 tools to measure the environmental impact of ML solutions. https://audacia.co.uk/technical-blog/tools-for-measuring-the-environmental-impact-of-ml 

• de Vries, Alex. (2023). Joule. The growing energy footprint of artificial intelligence. https://www.cell.com/joule/fulltext/S2542-4351(23)00365-3 

• Freedom, P. (2024). The Misrepresented Environmental Impact of AI: Separating Fact from Fiction. https://medium.com/the-simulacrum/the-misrepresented-environmental-impact-of-ai-separating-fact-from-fiction-7eb60893e4e9 

• Heikkilä, M. (2023). MIT Technology Review. Making an image with generative AI uses as much energy as charging your phone.  https://www.technologyreview.com/2023/12/01/1084189/making-an-image-with-generative-ai-uses-as-much-energy-as-charging-your-phone/ 

• Heikkilä, M. (2023). MIT Technology Review. AI’s carbon footprint is bigger than you think. https://www.technologyreview.com/2023/12/05/1084417/ais-carbon-footprint-is-bigger-than-you-think/ 

• Luccioni & Jernite (HuggingFace, 2024). Power Hungry Processing: Watts Driving the Cost of AI Deployment? https://arxiv.org/pdf/2311.16863 

• MIT Technology Review (2023). Making an image with generative AI uses as much energy as charging your phone. https://www.technologyreview.com/2023/12/01/1084189/making-an-image-with-generative-ai-uses-as-much-energy-as-charging-your-phone/ 

• NPR (2024). AI brings soaring emissions for Google and Microsoft, a major contributor to climate change. https://www.npr.org/2024/07/12/g-s1-9545/ai-brings-soaring-emissions-for-google-and-microsoft-a-major-contributor-to-climate-change 

• OECD (2024). AI Emissions Scenario Generator. https://oecd.ai/en/catalogue/tools/ai-emissions-scenario-generator (Direct link to tool: https://borisruf.github.io/carbon-footprint-modeling-tool/ai-scenarios.html) 

• PlanBeEco (2024). AI’s carbon footprint – how does the popularity of artificial intelligence affect the climate? https://planbe.eco/en/blog/ais-carbon-footprint-how-does-the-popularity-of-artificial-intelligence-affect-the-climate/ 

• SlashGear (2024). Just How Much Energy Does Generating An AI Image Actually Use? https://www.slashgear.com/1696332/ai-image-generation-how-much-energy-used/ 

• Tomlinson et al. (2024). Nature. The carbon emissions of writing and illustrating are lower for AI than for humans. https://www.nature.com/articles/s41598-024-54271-x 

• Unversity of California, Riverside (2023). AI programs consume large volumes of scarce water. https://news.ucr.edu/articles/2023/04/28/ai-programs-consume-large-volumes-scarce-water 

• Yu et al. (2024) Revisit the environmental impact of artificial intelligence: the overlooked carbon emission source? https://link.springer.com/article/10.1007/s11783-024-1918-y 

Carbon Emissions of Generating Electricity

The source of electricity is a key factor in the environmental footprint of AI tools. For example: at the moment, it looks like China's DeepSeek is much more efficient as a tool than ChatGPT - but more of China's electricity comes from fossil fuels. However, China is moving to renewable energy faster than any other country. Big energy savings could be made with locally-hosted AI models or data centers powered directly from renewable energy. 

• Carbon Brief (2023). China Carbon Emissions Set To Fall in 2024. https://www.carbonbrief.org/analysis-chinas-emissions-set-to-fall-in-2024-after-record-growth-in-clean-energy/ 

• EIA (2024). How much carbon dioxide is produced per kilowatthour of U.S. electricity generation? https://www.eia.gov/tools/faqs/faq.php?id=74&t=11 

• Ember Research (2024). China is a leader in solar, wind and electric transportation, responsible for half of global growth in 2023. https://ember-energy.org/countries-and-regions/china/ 

• Our World in Data (2024). Renewable energy sources are growing quickly and will play a vital role in tackling climate change. https://ourworldindata.org/renewable-energy 

• Statista (2024). Carbon intensity of the power sector in China from 2000 to 2023. https://www.statista.com/statistics/1300419/power-generation-emission-intensity-china/ 

• Zhang et al. (2024).  Study on life-cycle carbon emission factors of electricity in China. International Journal of Low-Carbon Technologies.  https://academic.oup.com/ijlct/article/doi/10.1093/ijlct/ctae181/7762370 

Finding Equivalencies: How Many ___ Am I Using? 

This is a bit of a challenge, but we see it in loads of articles (such as the MIT piece below). In the calculator I wanted students to be able to see rough equivalents. Working out the tree equivalent was trickier, but hopefully creates a visual around offsetting and the services that nature provides. 

• Apple (2024). Environmental Progress Report. https://www.apple.com/environment/ 

• EPA (2024). Greenhouse Gas Equivalencies Calculator. https://www.epa.gov/energy/greenhouse-gas-equivalencies-calculator 

• EPA eGrid (2024). Emissions & Generation Resource Integrated Database (eGRID). https://www.epa.gov/egrid 

• ForTomorrow (2023). How much CO2  does a tree absorb per year? https://www.fortomorrow.eu/en/blog/co2-tree 

• Jackery (2024). How Many Watts Does A Laptop Use: MacBook, Dell, Asus and More [With Data Table]. https://www.jackery.com/blogs/knowledge/how-many-watts-a-laptop-uses 

• MIT Technology Review (2023). Making an image with generative AI uses as much energy as charging your phone. https://www.technologyreview.com/2023/12/01/1084189/making-an-image-with-generative-ai-uses-as-much-energy-as-charging-your-phone/ 

AI Model Research

• DeepSeek (2024). DeepSeek-R1:  Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. https://github.com/deepseek-ai/DeepSeek-R1/blob/main/DeepSeek_R1.pdf 

• Our World in Data (2024). Interactive Charts on AI. https://ourworldindata.org/artificial-intelligence