Details
- Salesforce is updating its AI model cards to include standardized metrics on energy use and carbon emissions across the full model lifecycle, from pre-training to inference.
- The initiative builds on Salesforce’s existing model cards, in place since 2020, and aligns with its trusted AI principles and ISO 42001-certified AI governance standards.
- A new Environmental Impact section is initially available for select proprietary models, including First Name Match, Account Match, and TextEval, using the AI Energy Score methodology to estimate energy consumption and emissions.
- The AI Energy Score framework, which Salesforce helped develop with ecosystem partners, standardizes AI energy reporting by factoring in hardware type, GPU utilization, runtime, and data center region, and complements its broader benchmarking efforts for AI energy efficiency.
- By embedding these disclosures directly into existing model evaluation workflows, Salesforce aims to make environmental impact a routine, comparable metric for customers choosing between models with similar performance but different carbon footprints.
Impact
This move operationalizes environmental transparency in AI at a time when energy and carbon costs are becoming board-level concerns for enterprise adopters. Standardized lifecycle metrics could accelerate demand for more energy-efficient models, influence procurement criteria, and push competitors to match or exceed Salesforce’s disclosure practices over the next 12–24 months.
