Green Software Foundation ratifies SCI for AI, a standard for measuring AI emissions

2025-12-17 · Publisher : Green Software Foundation · Domain : Cloud & DevOps

The Green Software Foundation ratifies SCI for AI, a standardized method for measuring the carbon footprint of AI systems across their lifecycle, built on the ISO SCI standard.

### Background On 17 December 2025, the Green Software Foundation announces ratification of **SCI for AI** (Software Carbon Intensity for Artificial Intelligence), a specification extending the SCI method, ISO/IEC 21031:2024, to the measurement of AI systems' emissions across their lifecycle. It was developed under the leadership of Navveen Balani (Accenture) and Henry Richardson (WattTime). The foundation presents it as the first consensus-based standard making the carbon footprint of AI systems transparent, comparable and actionable. ### Key facts - Scope: classical machine learning, computer vision, natural language processing, generative AI and agentic systems. - Functional units: tokens for language models, inferences for classifiers, FLOPs for training efficiency. - Basis: the SCI methodology, ISO/IEC 21031:2024. - Leadership: Navveen Balani (Accenture) and Henry Richardson (WattTime). - Support: Google, through Jonathan Turnbull, member of the GSF Green AI Committee, welcomes a shared standard to track emissions. ### Why it matters Customers increasingly ask for the footprint of their AI uses, for sustainability reports as well as to prepare regulatory requirements. A common method prevents each model or cloud provider from publishing its own figures, hard to compare. For services firms, SCI for AI provides a frame to document the impact of an AI project and compare model choices. ### Takeaway The carbon footprint of AI gets a common measurement method, built on the ISO software standard.

Tags : GreenOps, IA, Open Source, Énergie