The same AI request emits 8.5 times more CO2 in Texas than in Sweden (berget.ai)

Swedish AI provider Berget AI has introduced an open-source method to measure the carbon emissions associated with individual AI requests, revealing a significant disparity in emissions between regions. A typical request to their model, Kimi K3, emits 0.4 grams of CO₂ in Sweden versus 3.2 grams in Texas—an 8.5 times difference—due to variations in regional electricity grid emissions. Berget AI aims to empower users to track their carbon impact in real-time and encourages the broader AI community to adopt similar transparency, fostering informed choices regarding AI services based on concrete emission data rather than vague estimates. This initiative is vital as the demand for AI technology is expected to surge, with data center electricity consumption projected to double by 2030. The method not only tracks emissions from AI usage but also takes into account the manufacturing impact of hardware, promoting a more nuanced understanding of the environmental implications of AI infrastructure. By making this method publicly accessible, Berget AI facilitates a critical shift towards sustainability in AI, enabling users and organizations to consider the carbon footprint of their cloud-based AI operations and make environmentally responsible decisions.