Berget AI Releases Emissions Calculation Method
The open-source tool calculates the carbon footprint of AI requests by measuring the intensity of local electrical grids.
Updated on Oct. 21, 2026 in Artificial Intelligence
Live Poll
Should companies prioritize locations with lower carbon footprints when hosting their AI data centers?
Berget AI has released an open-source methodology designed to quantify the carbon dioxide emissions associated with individual AI model requests. The calculation method accounts for the geographic location of the request, reflecting the varied carbon intensity of local power grids.
Why it matters
As the AI industry grows, this tool provides standardized transparency regarding the climate impact of model usage across different infrastructure environments. It aims to clarify how regional energy sources directly influence the environmental cost of executing AI tasks.
The method uses granular grid-intensity data to estimate emissions, showing that a Kimi K3 request generates 0.4 grams of CO2 in Sweden versus 3.2 grams in Texas. This eight-fold discrepancy highlights how location-specific energy dependencies alter the total footprint per request.
The players
Berget AI
An AI organization focused on developing tools for transparency and environmental impact tracking in the machine learning industry.
Stockholm Environment Institute
A research organization dedicated to environment and development issues that contributed technical analysis to the project.
WattTime
A non-profit tech organization specializing in software solutions that detect and reduce emissions from power grids.
Klimatkollen
An environmental research initiative that provided expertise for this carbon-tracking methodology.
The details
The methodology correlates compute power consumption with the carbon intensity of the local electrical grid where the request is processed. By mapping server location to real-time grid data, the system calculates the carbon output based on the mix of renewable and fossil fuel sources feeding that specific data center. This approach provides a location-aware metric rather than relying on global industry averages.
Timeline
Berget AI released the emissions calculation method in October 2026.
The Tech Race
This effort moves beyond the 2024 AI energy transparency reports from major hyperscalers that historically relied on annual, company-wide carbon disclosures. It advances the field toward precise, unit-based tracking that treats carbon cost as a fundamental attribute of AI performance.
Developers and companies can now integrate this methodology to provide users with carbon estimates per request, making environmental impact a visible metric for AI operations. The accuracy of these estimates remains tied to the quality of local grid data provided by regional energy monitoring systems.
The takeaway
This framework establishes a standard for measuring the climate cost of AI computation based on geographic energy grids. Observers should track whether major cloud providers adopt this location-aware calculation to standardize carbon reporting across the industry.
Further reading
For broader trends in infrastructure efficiency, see the latest in Artificial Intelligence.
Source note: This article includes information reported by CIO.
Live Poll
Should companies prioritize locations with lower carbon footprints when hosting their AI data centers?
