Estimate the carbon emissions produced by training or running an AI model.
The carbon footprint of an AI workload is a function of how much energy it consumes and how carbon-intensive the electricity grid powering it is. This calculator multiplies energy use (kWh) by the grid's carbon intensity (grams of CO2 per kWh) to estimate total emissions, and translates the result into an equivalent driving distance for intuitive comparison. Choosing a renewable-heavy region or data center can cut emissions by an order of magnitude for the same energy usage.
co2_kg = (energy_kwh × carbon_intensity_g_per_kwh) / 1000
Significantly — a renewable-heavy grid can emit roughly 1/16th the CO2 per kWh compared to a coal-heavy grid, so choosing a green region for training can be one of the highest-leverage emissions reductions available.
Use the Energy Consumption Calculator (AI) to derive kWh from GPU count, TDP, hours run, and data center PUE, then feed that result into this calculator.
Major providers (AWS, GCP, Azure) publish regional carbon intensity and renewable energy percentages for their data centers, which you can use in place of the presets here.