Chapter 08 · Finance, data & evidenceData, Technology & Verification Systems
Carbon-aware computing
Definition
The practice of adjusting when and where computation runs in response to the carbon intensity of electricity supply. Flexible workloads (batch processing, AI training, media encoding) are shifted toward hours or data-centre locations with lower grid carbon intensity, using carbon-intensity forecasts, while preserving service-level commitments.
References
Supports definition and framing.
Supports definition and framing.
Overview
What it means
Pioneered at hyperscale by Google's Carbon-Intelligent Computing System (deployed from 2020; published 2021), carbon-aware computing exploits a fact of decarbonising grids: emissions per kWh vary hour to hour and region to region.
Google's fleet-wide implementation demonstrated measurable reductions in power draw at peak-carbon hours, and the approach pairs with the 24/7 carbon-free-energy goal of matching demand to clean supply every hour.
The concept has diffused into open-source practice via the Green Software Foundation's Carbon Aware SDK and into cloud carbon tooling, though measured effects remain modest where flexible load is a small share.
How it is used
Cloud providers and enterprises deploy carbon-aware schedulers; software engineers use carbon-intensity APIs (Electricity Maps, WattTime) to delay or relocate jobs; digital-sustainability reporting credits carbon-aware design as an operational mitigation measure.
Why it matters
As AI-driven compute demand surges, carbon-aware computing is one of the few levers that reduces digital emissions immediately, using the clean electricity already on the grid — a bridge to fully carbon-free operations.