Ratified
Software Carbon Intensity for Artificial intelligence (SCI for AI)
Extending the Software Carbon Intensity (SCI) to Artificial intelligence (AI). Addressing the challenges of measuring Artificial intelligence carbon emissions
Overview
The purpose of this specification is to assist AI practitioners—developers, data scientists, engineers, and decision-makers—in understanding and reducing the carbon footprint of AI systems. By making informed choices about model design, computational efficiency, and deployment strategies, practitioners can minimize emissions while maintaining performance.
We have developed a methodology for calculating the carbon emissions rate (SCI score) of AI software systems, including both classical AI and generative AI applications. This specification provides a reliable, consistent, and comparable measure that practitioners can use to set targets and track progress in reducing carbon emissions throughout the AI lifecycle—from development and training to deployment and inference.
Status: Ratified — This is a published, ready-to-use standard. View the full specification.
Getting Started
1. Understand the Standard
- Read the SCI AI specification for detailed methodology and guidance
- Review the repository for technical documentation and examples
2. Implement the Calculation
- Gather data on your AI system's energy consumption, hardware details, and operational metrics
- Use the formula to calculate your baseline SCI score
- Document assumptions and data sources for transparency
3. Set Targets & Iterate
- Compare your score against similar systems to establish realistic targets
- Optimize model architecture, inference frequency, or compute infrastructure
- Recalculate and track progress over time
Resources
- Specification: sci-for-ai.greensoftware.foundation
- Repository: github.com/Green-Software-Foundation/sci-ai
- Mailing List: [email protected]
- Related Standard: SCI (Software Carbon Intensity) — the foundation for AI-specific adaptations
Get Involved
Subscribe & Participate
- Complete GSF Registration if you haven't already
- Subscribe to the project to access the mailing list and meetings
Contributing
- GitHub: We work in the open via the sci-ai repository. Issues, pull requests, and discussions are welcome.
- Meetings: Join the Software Standards Working Group meetings (details sent to subscribers)
- Email: Reach out to the project mailing list ([email protected]) with questions or contribution ideas
How to Help
- Test the standard on your AI systems and share feedback
- Contribute case studies or reference implementations
- Help translate or adapt the standard for specific use cases
- Participate in discussions about methodology refinements