Skip to content
ChartUsesUses · kWh for 1,000 inferences
Published on 20 October 2026

Electricity by type of AI task

For 1,000 inferences, in kilowatt-hours (logarithmic scale)

Electricity by type of AI task

0.0010.01× 100.1× 101× 1010× 10ScalelogarithmickWh (logarithmic scale)Image generation2.9Text summarisation0.049Text generation0.047Object detection0.038Image classification0.007Text classification0.002
Source: Luccioni et al., Power Hungry Processing, 2024. projetcelsius.com/en/ai · CC BY 4.0AI Impact, by Projet Celsius

The analysis

  • Generating costs far more than classifying: image generation consumes more than 1,000 times text classification.
  • The models measured are open models, smaller than today's large assistants.

The figures

  • The scale is logarithmic: each graduation multiplies by 10.
  • Image generation: 2.9 kWh for 1,000 inferences.
  • Text summarisation: 0.049 kWh for 1,000 inferences.
  • Text generation: 0.047 kWh for 1,000 inferences.
  • Object detection: 0.038 kWh for 1,000 inferences.
  • Image classification: 0.007 kWh for 1,000 inferences.
  • Text classification: 0.002 kWh for 1,000 inferences.

Assumptions

  • Electricity measured on open models, average per task.
  • Averages over several models per task.
  • 2023 hardware and software: orders of magnitude remain useful, not the exact values.

Published on 20 October 2026MethodologyAll sourcesLicence CC BY 4.0

Charts and maps

A question about a figure?

An error, a source to add, a methodology question, a press request: write to us. Guillaume Pakula, Clément Reynaud, Sébastien Pierfederici and Alexandre Veilly will reply.

Write to us →

Projet Celsius, which runs this hub, also helps companies measure and reduce the footprint of their AI use.

Accounting for AI in your Bilan Carbone® ↗
Projet Celsius, environmental strategy consultancy

We support companies that account for AI in their Bilan Carbone®

Visiter Projet Celsius ↗