In short
Use regional, time-matched grid data for where and when your AI work ran. In Great Britain the Carbon Intensity API from the National Energy System Operator gives regional estimates and forecasts; Electricity Maps covers more than 350 zones worldwide. Fall back to national yearly averages only when nothing finer exists, and say which you used.
Steps
- Identify the grid region where the work physically ran, which for cloud work is the data centre's region.
- Get carbon intensity for that region at the finest time resolution available, ideally hourly.
- Match the intensity to the hours when the energy was used.
- Use location-based grid figures, not certificates or tariffs, for software calculations.
- Record the data source and resolution with every figure.
Carbon intensity, in grams of CO₂e per kilowatt-hour, turns energy into carbon, and it changes by place and hour. One national yearly average can hide both a real improvement and a real problem.
Where can I get the data?
| Source | Coverage | What it offers |
|---|---|---|
| Carbon Intensity API (National Energy System Operator) | Great Britain, 14 regions | Estimates and forecasts more than 96 hours ahead |
| Electricity Maps | More than 350 zones worldwide | Real-time, forecast and historical data |
| Google Cloud region data | Google Cloud regions | Grid carbon intensity and carbon-free energy share per region |
| Cloud providers' carbon dashboards | Your account | Monthly estimates; see your cloud provider's carbon data |
Which grid counts?
The grid where the computing happens. For a hosted AI service or cloud job, that is the data centre's region, not your home or office. The Green Software Foundation's guidance prefers frequent, regional data, with averages only as a fallback.
Common mistakes
- Using a market-based figure. Renewable certificates and green tariffs don't change the physical grid at a given hour; software carbon calculations use location-based intensity.
- Using annual averages for scheduling. Forecasts are what make shifting work to cleaner hours possible.
- Not saying which data you used. Two honest figures can differ just because of the grid data.
See how to measure AI energy for the full method.
Sources
- Carbon Intensity API, National Energy System Operator (2025)
- Electricity Maps data, Electricity Maps (2026)
- SCI guidance: carbon intensity (I), Green Software Foundation (2024)
- Carbon free energy for Google Cloud regions, Google Cloud (2025)
Related
How do I schedule training when the grid is greener?
Run training and batch AI jobs when grid carbon intensity is low: use forecasts, make jobs pausable with checkpoints, and keep deadlines as hard limits.
How do I pick a cloud region with low carbon?
Choose a lower-carbon cloud region for AI work: compare grid carbon intensity and carbon-free energy by region, then weigh latency, data rules and cost.
How do I calculate a Software Carbon Intensity score for an AI feature?
Work out an SCI score for an AI feature: energy times grid intensity, plus embodied emissions, divided by a functional unit such as per request.
How to measure the energy and carbon of AI
A practical guide to metering or estimating the energy and carbon of AI work: tools, system boundaries, per-task metrics and honest uncertainty.