#compute-emissions-measurement

Consumer-side tools must assume model architecture and a worldwide-average grid because providers publish almost nothing about model size, hardware, or data-centre efficiency. Every downstream estimate therefore carries wide, irreducible error bars.

Running software and AI models consumes electricity whose CO2 cost varies by grid and workload, but the default toolchain reports latency, cost and accuracy and almost never emissions. Invisible to the people who could reduce it, the footprint goes unmanaged and unoptimised.

Benchmark models on the same hardware for the same tasks, then publish results as star-band ratings on a leaderboard with a shareable label. An ENERGY STAR-style signal makes efficiency legible to non-experts and usable in procurement, pressuring providers to disclose.

Every emissions estimate multiplies energy by a carbon-intensity factor, but annual national averages miss regional and hourly variation. An API exposing live, historical, and forecast intensity—both average and marginal—per zone and per hour lets any tool convert energy to accur

Every tool in this space estimates differently, so numbers are not comparable across teams or vendors. A shared specification expresses software carbon as a rate — emissions per functional unit such as a request, user, or training run — so results can be compared and tracked over

Convert itemised cloud billing and usage records into energy and emissions figures using published hardware power coefficients and region-specific grid carbon intensity, producing a per-service, per-region footprint across multiple providers.

When a generative model is called through a provider's API, hardware cannot be metered. Estimate per-request energy and CO2 from model size, token counts, and data-centre assumptions using a life-cycle method, giving API consumers an impact figure they otherwise cannot obtain.

For code you operate, sample the energy drawn by CPU, GPU and RAM during execution, multiply by the carbon intensity of the local electricity grid, and log the resulting CO2 estimate next to other run metrics so it can be tracked and reduced.

AI Energy Score, launched in February 2025 at the Paris AI Action Summit by Hugging Face, Salesforce, Cohere and Carnegie Mellon University, benchmarks models on standardised NVIDIA H100 hardware across 10 tasks and ass…

WattTime, founded in 2014 by UC Berkeley researchers and a subsidiary of Rocky Mountain Institute since 2017, provides Marginal Operating Emissions Rate (MOER, in pounds of CO₂ per MWh) via API in real-time, forecast, a…

Electricity Maps ingests data from TSOs, market operators, and government agencies via an open-source parser system, applies a flow-tracing algorithm to produce consumption-based (rather than production-based) carbon in…

The Software Carbon Intensity (SCI) specification reached ISO accreditation as ISO/IEC 21031:2024 in Q1 2024. It defines software carbon as a rate: SCI = (E × I) + M per functional unit R, using a rate rather than a tot…

Google Cloud Carbon Footprint is a provider-native emissions tool included as a deliberate contrast to open-source multi-cloud approaches. It reports both location-based and market-based scope 2 emissions computed on an…

Cloud Carbon Footprint (CCF) is an open-source, multi-cloud tool launched by Thoughtworks in March 2021. It reads itemised billing and usage data from AWS, Google Cloud, Azure and Alibaba, converts it to energy using he…

EcoLogits is an open-source Python library that estimates energy and multi-criteria environmental impacts of generative AI API calls. It patches provider Python client libraries (OpenAI, Anthropic, Mistral AI, Cohere, G…

Kepler (Kubernetes-based Efficient Power Level Exporter) attributes energy to individual processes, containers, pods, and nodes in orchestrated environments, exporting results as Prometheus metrics. It draws power readi…

CodeCarbon is an open-source Python package that estimates CO2 from code execution. An EmissionsTracker (used as a context manager or decorator) samples CPU energy via Intel RAPL, GPU energy via NVIDIA NVML, and RAM e…