Examples
Worked examples
- Is an instance
A research workflow queues simulation jobs to run when forecast UK grid intensity drops below a defined threshold.
- Is an instance
A cluster job scheduler offers a low-carbon queue with priority drop in exchange for delayed start.
Counter-examples
Looks similar, but isn't
- Not an instance
Running interactive notebook sessions on a high-carbon grid is unavoidable and not in scope.
- Not an instance
Marketing 'green' compute without runtime carbon-intensity sensing is not carbon-aware.
Editorial commentary
Carbon-aware computing shifts when and where a computational workload runs in response to how carbon-intensive the electricity grid is at that place and time, rather than treating compute as instantaneous and location-agnostic. Grid carbon intensity (grams CO2e per kWh) varies substantially by region, driven by the local generation mix, and by hour, as renewable output (especially solar and wind) fluctuates — the same job can carry a markedly different carbon cost depending purely on when and where it is scheduled, without any change to the code itself.
How it’s implemented
The Green Software Foundation’s open-source Carbon Aware SDK is the most widely referenced tooling in this space: it exposes real-time and forecast grid carbon-intensity data, drawing on sources such as the Electricity Maps API, so that an application or job scheduler can decide to run now, delay, or route a workload to a lower-carbon region. In research computing this is most applicable to flexible, non-urgent, interruptible jobs — batch simulation runs, non-time-critical model training, overnight data-processing pipelines — rather than interactive work, which by definition has to run when the researcher is at the keyboard regardless of grid intensity at that moment. Region-shifting is subject to data-sovereignty and institutional data-governance constraints and cannot be applied indiscriminately.
Distinguishing it from adjacent concepts
Carbon-aware computing is about when and where a fixed workload runs; it complements, rather than substitutes for, reducing the workload itself (see training carbon footprint and inference carbon footprint for the demand side) and energy proportionality, which is about hardware efficiency at a given utilisation level rather than scheduling.
What doesn’t count
Marketing compute as “green” on the basis of renewable energy purchase agreements alone, without any runtime sensing of actual carbon intensity or ability to shift load, is not carbon-aware computing in the operational sense this term describes — it is a separate, legitimate offsetting mechanism, but a different one. See sustainable HPC for how scheduling fits alongside cooling and procurement decisions.
References
- Green Software Foundation, Carbon Aware SDK documentation and GitHub repository.
Also known as
Carbon-aware scheduling · Grid-aware computing
Machine-readable encodings
Use in your systems
<role vocab="credit"
vocab-identifier="https://casrai.org/dictionary/"
vocab-term="Carbon-aware computing"
vocab-term-identifier="https://casrai.org/dictionary/term/carbon-aware-computing" />{
"@context": "https://schema.org",
"@type": "DefinedTerm",
"@id": "https://casrai.org/dictionary/term/carbon-aware-computing",
"name": "Carbon-aware computing",
"identifier": "https://casrai.org/dictionary/term/carbon-aware-computing",
"description": "The practice of scheduling, shifting, or routing computational workloads in response to real-time grid carbon intensity, so that flexible jobs run when and where electricity emissions are lowest.",
"inDefinedTermSet": "https://casrai.org/dictionary/domain/sustainable-research#set",
"url": "https://casrai.org/dictionary/term/carbon-aware-computing",
"sameAs": [
"Carbon-aware scheduling",
"Grid-aware computing"
],
"license": "https://creativecommons.org/licenses/by/4.0/",
"publisher": {
"@id": "https://casrai.org/#organization"
},
"author": {
"@id": "https://casrai.org/#editorial-team"
},
"datePublished": "2026-05-21T02:23:02",
"dateModified": "2026-08-22T16:25:14",
"inLanguage": "en-GB",
"isAccessibleForFree": true
}
