AI Carbon Footprint Varies Wildly By Location

A new tool reveals an AI query's carbon footprint varies by location, helping companies choose greener cloud regions for their workloads.

A Swedish company, Berget AI, has released a new method and tool to calculate the carbon dioxide generated by individual AI requests, according to a report from CIO.com. Developed with researchers from organizations including the Stockholm Environment Institute and WattTime, the project aims to bring transparency to the AI industry's growing climate impact. The key finding is that the environmental cost of an AI query is not fixed; it changes dramatically based on the geographical location of the data center processing it. This allows companies to directly compare the carbon footprint of different AI providers and cloud regions, turning an abstract environmental concern into a measurable data point. The initiative provides both an open-source methodology for developers and a user-friendly web calculator for broader access.

The tool’s innovation lies in how it connects two distinct sets of data. It first analyzes the computational power required for a specific AI task, then cross-references that energy consumption with real-time information about the carbon intensity of the electrical grid where the computation occurs. This is made possible through partnerships with organizations like WattTime, which tracks the mix of energy sources—such as coal, natural gas, solar, and wind—powering different grids at any given moment. This granular approach explains why the same AI query run in a Texas data center, which relies more heavily on fossil fuels, would generate significantly more CO2 than one run in Sweden, where the grid incorporates a higher percentage of renewable and nuclear energy. It moves beyond simple energy usage metrics to assess the actual environmental impact based on the local power source.

This development arrives as the tech industry grapples with the rapidly growing energy demands of artificial intelligence. The training and operation of large language models consume vast amounts of electricity, placing a significant strain on power grids and contributing to global carbon emissions. While major cloud providers like Amazon Web Services, Google Cloud, and Microsoft Azure have made public commitments to sustainability and offer their own carbon footprint dashboards, this new tool offers a more specialized and independent perspective. It focuses specifically on the per-query impact of AI workloads, a level of detail often missing from broader cloud-level reports. By open-sourcing the methodology, the creators are encouraging a standardized, transparent approach to measurement that can be adopted across the industry, moving beyond proprietary calculations and corporate sustainability reports.

For CTOs, developers, and IT teams, this tool provides an immediately actionable way to integrate sustainability into technical decision-making. Instead of selecting a cloud region based solely on latency, cost, and data sovereignty, teams can now add carbon intensity as a key criterion. This could lead to new “carbon-aware” application architectures where AI workloads are dynamically routed to data centers in regions with the cleanest available energy. While not always feasible for latency-sensitive applications, it presents a powerful option for asynchronous tasks like model training or batch processing. The next step will be to see how this data is integrated into major cloud platforms and orchestration tools, potentially enabling automated, environmentally-optimized workload scheduling. This shifts the responsibility for green computing from the cloud provider to the application developer, empowering them to build more sustainable software.

For developers and IT leaders, this provides a concrete, open-source method to measure and reduce the environmental impact of AI applications. It moves sustainability from a vague corporate goal to a quantifiable engineering metric, directly influencing infrastructure choices and application architecture.

Companies can now use this data to lower their carbon footprint, which is crucial for ESG reporting and attracting environmentally conscious customers. Strategically choosing cloud regions for AI can also lead to long-term operational resilience and potentially lower energy costs.