It bases estimates on AI tokens, since many companies pay for licenses that way. The approach, detailed in a white paper published in mid-July, advocates for estimating the average emissions associated with AI tokens, the smaller bits of code that make up an AI prompt. Hard to findAI emissions data is still scarce and often very high-level, such as this new AI emissions tracker, which rates eight AI data center companies. Request more data from AI providers , including (but not limited to) energy-intensity per token, training-related emissions, embodied carbon for the hardware and the physical region where the AI is hosted. , including (but not limited to) energy-intensity per token, training-related emissions, embodied carbon for the hardware and the physical region where the AI is hosted.