Skip to content
by Projet Celsius
Projet Celsius
Key figureClimateUpdated on 11 November 2026

1.14gCO2e

Carbon footprint of one Le Chat answer, over its life cycle

An answer of about one page · full life cycle · Mistral AI, July 2025

An answer of about one page from Le Chat, Mistral AI's assistant, emits 1.14 gCO2e over its full life cycle and consumes 45 mL of water, according to the life cycle assessment published by Mistral AI in July 2025 with Carbone 4 and ADEME.

The broadest scope published by a provider: it cannot be compared with an electricity-only figure.

See the source
A fully dismantled laptop, its parts lined up on a wooden floor
A dismantled laptop: the life cycle also counts manufacturing. Photo Siarhei V, Wikimedia Commons, CC BY-SA 4.0

Definition and scope

Greenhouse gas emissions of an answer of 400 tokens from Mistral AI's assistant, counting the full life cycle: training, server manufacturing, electricity, data centre.

What is included

  • The manufacturing of servers and chips
  • Training the model, spread over its answers
  • Data centre electricity
  • The building and the network

What is excluded

  • The user's device

How the value is obtained

Life cycle assessment conducted by Mistral AI with Carbone 4 and ADEME, reviewed by Resilio and Hubblo, covering training and 18 months of use of Mistral Large 2.

Limits and uncertainties

  • A single model and a single provider.
  • The full report is not public: only the page and its infographic are.

The order of magnitude

All figures

1.14 gCO2e on an emissions scale, with 3 values from the linked chart, brought to the same unit. Each tick multiplies by 10.

0.01 g0.1 g1 g10 g1.14 gCO2eLe Chat answer0.012 g0.3 Wh, in France0.03 gGoogle, Gemini0.14 g0.3 Wh, world
1 g1.14 gCO2eLe Chat answer0.14 g0.3 Wh, world0.03 gGoogle, Gemini0.012 g0.3 Wh, in France
This figureA value from the chart "One question, 4 ways of counting CO2e"Logarithmic scale

The other published estimates

The chart and its sources

The chart "One question, 4 ways of counting CO2e" compares 4 estimates of the same quantity. This entry uses 1.14 gCO2e. The differences stem first from what each one counts.

  1. 0.3 Wh, in FranceServer electricity, 41.4 gCO2e/kWh0.012 g92 times less
  2. Google, GeminiElectricity according to its contracts, amortised hardware included0.03 g38 times less
  3. 0.3 Wh, worldServer electricity, 458 gCO2e/kWh0.14 g8.3 times less
  4. Mistral AI, Le ChatFull life cycle, excluding the user's device1.14 gSelected value

In context

See the chart and its data

One question, 4 ways of counting CO2e

Emissions of a written question to an AI, in gCO2e (logarithmic scale)

2025

More than 90 times difference between the lowest and the highest: it stems from the electricity grid, purchase contracts and scope.

0.010.1× 101× 1010× 10ScalelogarithmicgCO2e (logarithmic scale)0.3 Wh, in Francecalculation, 41.4 g/kWh0.012 gGoogle, Geminiaccording to its contracts0.03 g0.3 Wh, worldcalculation, 458 g/kWh0.14 gMistral AI, Le Chatlife cycle, half a page1.14 g
Source: Epoch AI; Ember; Google; Mistral AI · projetcelsius.com/en/ai · CC BY 4.0AI Impactby Projet Celsius

Update history

The hub changelog
Answer sentence added at the top of the entry
First version of the entry

More on climate

The whole Climate topic →

Charts and maps

ChartClimateOne question, 4 ways of counting CO2eover90timesgap between the lowest and the highestChartClimateServer electricity, country by country14times as muchin Poland than in France, for the same calculationChartClimateAI emissions, compared99MtCO2eAI in 2025, as much as BelgiumChartClimateGoogle and Microsoft emissions58 to 81%rise in emissions since the baseline year

Analyses

AnalysisClimate12 November 2026Why do estimates of AI's impact differ?Calculated with French electricity, one question to an AI emits 0.012 gram of CO2e; Mistral AI publishes 1.14 gram, more than 90 times as much. Its water use ranges from 0.26 to 45 millilitres depending on the published measurements. These figures are accurate; they do not cover the same scope.Where does this figure come from?Climate20 October 2026"Meta's electricity emits only 1,358 tonnes of CO2e"In 2024, Meta reports 1,358 tonnes of CO2e for the electricity it buys, the annual footprint of 170 people in France. The same report also gives 5.97 million tonnes. Both of the 2 figures are accurate.

Tools

ToolClimateCompare the footprint of an AI use by the country where the servers are21times as muchof CO2e in Poland than in Norway

On the Projet Celsius blog ↗

Blog article ↗ClimateThe carbon footprint of AI: what a ChatGPT query costs and what it weighs in a Bilan Carbone®Blog article ↗WaterHow and why does AI pollute?

Across the hub

AI Impact DatabaseAll the figures, their methodology and their sources.AI Ranking10 assistants rated out of 100: performance, ethics and environment.AnalysesThe signed pieces: analyses, viewpoints, "Where does this figure come from?".Charts and mapsYou are hereThe reference visuals, with their data and sources.ToolsOpen calculations, your assumptions on the left, the result on the right.

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 ↗
  • Institut Pasteur
  • SNCF Gares & Connexions
  • Radio France
  • Institut Curie
  • Institut Polytechnique de Paris
  • ESSEC Business School
  • Qualiopi, certified process