Calculate and Offset Your AI Carbon Footprint
Every chat, image, and video you generate with AI runs on electricity in a data center. This free calculator estimates the carbon footprint of your AI use based on how you use it and where your electricity comes from. It then shows how many trees’ worth of forest it would take to offset it. Your gift helps us buy and protect rainforest land in Sumatra, where the forest is left to regrow naturally.
Your donation supports the purchase, restoration, and protection of a vital wildlife corridor for endangered elephants and countless other species. Use the calculator to estimate your Ai carbon emissions and suggested offset. Then make your gift to help restore and protect rainforest in Sumatra.
How does AI use energy?
AI models run on powerful computers in data centers, and every request uses electricity. A typical chat message uses a small amount, roughly a third of a watt-hour by published estimates. Image generation uses more, and short AI-generated videos use far more than anything else. The carbon impact also depends on how your local electricity is produced. A grid with more coal and gas means more emissions per request.
How do we calculate your AI footprint?
We multiply your daily AI use by an estimated energy cost per request, then by the carbon intensity of your electricity grid. We convert the result into the acres of regrowing forest and trees needed to absorb that carbon. Every assumption is listed in the sources and notes below the calculator, and you can see the full formula under “How we calculate this.”
FAQ's
Does using AI really have a carbon footprint?
Yes. Every AI request runs on servers that use electricity, and most electricity still produces some carbon emissions. For light chat use the footprint is small. For heavy image, video, or agent use it adds up.
How accurate is this calculator?
It’s an estimate. Energy use varies by AI model, the task, and the data center, so the calculator shows a likely range. Chat figures are based on published data from Google. Image, video, and agent figures are rougher ranges.
Can I offset my AI use for less than a few dollars a year?
For light chat use, yes. Your suggested donation is calculated per tree, so small footprints give small amounts. Many people round up or give for the whole year.
Is my donation tax-deductible?
Add your own answer here, including your organization’s legal name and EIN. I only know the “tax-deductible” line from your page, and I don’t want to put specifics in your mouth.
Does offsetting replace reducing my emissions?
No. Offsetting works best alongside cutting what you can. Protecting regrowing forest absorbs carbon over time, and the benefits for wildlife and biodiversity are real, but it doesn’t erase emissions on the day you generate them. Saying this plainly builds trust and protects you from greenwashing criticism.
- https://cloud.google.com/blog/products/infrastructure/measuring-the-environmental-impact-of-ai-inference
- https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/
- https://www.epa.gov/egrid
- https://www.epa.gov/energy/greenhouse-gas-equivalencies-calculator-calculations-and-references
- Energy per request is an estimate. Chat figures are based on Google’s published median of about 0.24 Wh for a text prompt, with a range of 0.2 to 0.5 Wh. Video figures are based on reporting that a short AI-generated clip uses roughly 950 Wh. Image, long-document, and agent figures are rough ranges, because energy use varies widely by model and task.
- Emissions are calculated from the electricity grid you select, using approximate carbon intensity of 0.8 lbs of CO2 per kWh for the United States and about 1.0 lb for the world average. Results are shown in US short tons (2,000 lbs).
- Offset calculations assume natural regeneration sequesters an estimated 4 US tons of CO2 per acre per year, equal to about 675 trees per acre. Suggested donations are $4 per tree. Regrowing forest absorbs less carbon in its early years and more as it matures, so these figures are long-term averages.