In short
Report AI inside your existing greenhouse gas inventory, not as a separate figure: electricity for hardware you run goes in Scope 2, and cloud and AI services usually go in Scope 3 purchased goods and services. Use supplier data where you can get it, estimate the rest from usage with a stated method, and give the boundary and uncertainty. Show energy per useful task alongside the total.
Steps
- Identify AI use you run yourself and AI or cloud services you buy.
- Put electricity for your own hardware in Scope 2, reported on both location-based and market-based methods.
- Put purchased cloud and AI services in Scope 3, using supplier emissions data where available.
- Estimate missing figures from tokens, calls or compute hours and write down every assumption.
- Report the boundary, method, uncertainty and an intensity metric such as emissions per task.
AI belongs in the inventory you already keep. It is electricity and purchased services, so it fits the Greenhouse Gas Protocol's existing scopes rather than needing a new category.
Which scope does AI go in?
- Scope 2 covers purchased electricity. If you run AI on your own servers, its electricity is here. The GHG Protocol's Scope 2 guidance asks for both a location-based figure (the local grid) and a market-based figure (your contracts and certificates).
- Scope 3 covers your value chain, in 15 categories. Cloud and hosted AI services are commonly reported under Category 1, purchased goods and services. The standard itself does not single out cloud, so state your choice.
What if suppliers give me nothing?
Estimate, and say so. Multiply your usage (tokens, calls, compute hours) by a published or supplier energy figure and a grid carbon intensity. Cloud providers' carbon dashboards help; see finding your cloud provider's carbon data. Replace estimates with supplier figures as they arrive.
Do I have to report?
That depends on where you are. In the EU, the Corporate Sustainability Reporting Directive was narrowed in 2026 to companies with more than 1,000 employees and more than €450 million net turnover, according to the Council of the EU. Smaller firms may still be asked for data by customers in scope. Elsewhere, check national rules.
Common mistakes
- Reporting only market-based figures, which can make AI look carbon-free.
- Leaving out embodied emissions without saying so.
- Showing only a total. Add an intensity measure, such as emissions per request, in the spirit of the Green Software Foundation's Software Carbon Intensity.
See how to measure AI energy and check wording against honest green claims.
Sources
- Scope 2 guidance, Greenhouse Gas Protocol (2015)
- Corporate value chain (Scope 3) standard, Greenhouse Gas Protocol (2011)
- Council signs off simplification of sustainability reporting and due diligence requirements, Council of the European Union (2026)
- Software Carbon Intensity specification, Green Software Foundation (2024)
Related
How do I set targets for AI energy and emissions?
How to set AI energy and emissions targets that mean something: a measured baseline, intensity and absolute goals together, and a regular review.
How do I find my cloud provider's carbon data?
Where AWS, Azure and Google Cloud put customer carbon data, what it covers, its limits, and how to use it alongside your own finer measurements.
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.
Honest green claims about AI
How to avoid greenwashing about AI: what you can and can't claim, rebound effects, honest reporting, and the UK, EU and US rules on environmental claims.