Berget AI opens its emissions method after correcting its Texas estimate

Berget AI released an open-source method on October 21 for estimating AI emissions per request, and corrected its Kimi K3 estimate for Texas from 3.9 to 3.2 grams of CO2e per request after reviewing grid data. The method, announced by founder Andreas Lundmark, still excludes training emissions, and Berget AI plans a marginal-emissions estimate. The correction shows request-level carbon accounting is only as comparable as its underlying assumptions.

Berget AI opens its emissions method after correcting its Texas estimate Berget AI revised its Kimi K3 estimate for Texas from 3.9 to 3.2 grams of CO2e per request after reviewing grid data; its method still excludes training. By RuntimeWire Staff https://runtimewire.com/author/runtimewire-staff · Published Primary source: Berget AI https://berget.ai/en/blog/per-request-co2-press-release Why it matters Berget is turning its Swedish infrastructure into part of the product comparison: buyers can weigh jurisdiction and emissions together. But the corrected figures and planned marginal-emissions estimate show that request-level accounting is only as comparable as its assumptions. Andreas Lundmark https://www.luminarventures.com/post/founder-series-a-conversation-with-andreas-lundmark?ref=runtimewire spent nearly two decades building AI work at Boston Consulting Group before leaving to start Berget AI https://berget.ai/?ref=runtimewire . On October 21st, Berget AI released an open-source method for estimating emissions request by request, arguing that buyers need comparable figures to choose where their AI runs. The announcement https://berget.ai/en/blog/per-request-co2-press-release?ref=runtimewire also includes a correction to its headline comparison.… When you purchase through links in our articles, we may earn a small commission https://runtimewire.com/affiliate-monetization-standards . This doesn’t affect our editorial independence.