Skip to main content

COMPEL Glossary / GL-60

Carbon Attribution

Grams of CO2-equivalent emissions attributed to AI model inference, typically reported per 1,000 calls and rolled up per model, workload, and tenant.

What this means in practice

Carbon attribution combines energy consumption from hosting providers or on-premise hardware with grid-intensity factors for the region of execution, producing a defensible sustainability metric that can be trended over time and compared across model choices.

Context in the COMPEL framework

A core metric of the Sustainability dimension. Captured during Evaluate and used in Learn to inform model selection and routing decisions.

Where you see this

Carbon Attribution is most commonly referenced when teams work across the Evaluate and Learn stages — especially within the Operational Readiness layer . It appears in governance artifacts, assessment instruments, and delivery playbooks wherever COMPEL is operationalized.

Related COMPEL stages

Related domains

Synonyms

carbon per inference , gCO2e per 1k calls , inference carbon footprint

See also

  • Trust & Performance Dimensions — The eight continuous-measurement axes against which every AI transformation is evaluated in COMPEL: Value, Reliability, Safety, Responsibility, Compliance, Security, Sustainability, and Adoption.
  • Time-to-Value — The elapsed time from a user being provisioned on an AI system to their first recorded value-generating interaction with it, measured at the cohort level.
  • Operational Readiness — The assessed capability of an organization to sustain AI operations across 10 interdependent dimensions: strategy alignment, governance maturity, operating model, workforce capability, data readiness, technology infrastructure, monitoring and observability, vendor dependency management, compliance readiness, and change and adoption.