Sustainability & AI
Inference
Using a trained model to generate outputs; for widely deployed AI services, cumulative inference is a major and growing share of energy use.
Definition
Using a trained model to produce outputs for new inputs — each query, prediction or generation — as distinct from the training phase.
Quick reference
At a glance
- Subject
- Sustainability & AI
- Editorial status
- Editorial draft
- Definition status
- Established
- Last updated
- 21 August 2026
References
This source provides part of the technical or institutional basis for the definition.
This source supports the explanation of how the term is applied, measured or governed in practice.
This source supports the wider sustainability significance and context described in the entry.
Overview
What it means
Individually cheap, inference aggregates: a model serving millions of users consumes energy with every request. Research such as 'Power Hungry Processing' shows inference can dominate a popular model's lifetime footprint, shifting attention from one-off training costs to continuous operating costs.
How it is used
Inference is the phase sustainability teams encounter as users: every automated classification, summary or forecast has a marginal energy cost. Efficient architectures, smaller models and batching reduce it; so does simply not using a large model where a small one suffices.
Why it matters
As AI embeds into daily operations, inference turns its environmental cost into a running metre. Accounting regimes and procurement criteria increasingly need to cover deployment, not just development.
Review
Help keep this definition useful and accurate.
Submitted reviews stay private until an editor decides whether to accept and attribute them.
Endorse this definition
Confirm what works
Endorse the current wording when it is accurate and useful. Suggested changes use the separate editorial form below.