Skip to content

Compare the footprint of an AI use by the country where the servers are

The same use, converted into CO2e with the electricity of 12 countries and regions.

Calculation
On the AI Impact Database, its factors and its inputs
Published
20 October 2026, CC BY 4.0 licence
1

Your use

The task to convertSame task, same energy in the 12 countries
1,000 written questions300Wh100 images291Wh10 videos of 5 seconds900Wh1 year of regular use6.5kWh
2

Your assumptions

Electricity of the 12 countries and regionsIn gCO2e per kWh, 2025, over the life cycle (Ember)
Poland589 gWorld (default)458 gUnited States384 gGermany330 gIreland257 gNetherlands254 gEuropean Union210 gFinland57.5 gFrance41.4 gSwitzerland39.2 gSweden35.3 gNorway28.1 g

Default : 1 year of regular use (10 questions a day, 5 images and 1 video a week), and the global average.

3

Results, country by country

Gap21times as muchof CO2e in Poland than in Norway, for the same use589 g versus 28.1 g per kWh
The most carbon-intensive3.8kgCO2ePoland, 1 year of regular use6.5 kWh x 589 g per kWh
The lowest180gCO2eNorway, for the same use6.5 kWh x 28.1 g per kWh
The same use, 12 electricity gridsin CO2e, electricity of 2025
↑Poland emits 21 times more than Norway, for the same use

Assumptions and sources

6.5 kWh1 year of regular useAI Impact Database28.1 to 589 g per kWhfrom Norway to Poland, 2025Ember180 gCO2e to 3.8 kgCO2ea gap of 21 times

Not countedRenewable electricity contracts

The calculation in detail

The values and the limits

The values used

1,000 written questions
300 Wh
100 images
291 Wh
10 videos of 5 seconds
900 Wh
1 year of regular use
6.5 kWh
Electricity of countries
Ember, 2025

The limits of the calculation

  • Average annual intensity of the electricity produced in the country, and worldwide (Ember, life cycle, 2025: the same source for all, that of the AI Impact Database).
  • A data centre can buy renewable electricity by contract; this calculation follows the real grid.
  • The large assistants do not say where their servers run: more than half of AI data centres are in the United States, according to GreenIT.

The 5 sources

One line per document
  1. 1Ember, Global Electricity Review 2026, and annual data via Our World in Data
  2. 2Our World in Data, based on Ember, Lifecycle carbon intensity of electricity
  3. 3Epoch AI, How much energy does ChatGPT use?, February 2025
  4. 4Luccioni, Jernite and Strubell, Power Hungry Processing: Watts Driving the Cost of AI Deployment?, 2024
  5. 5Delavande, Pierrard and Luccioni, Video Killed the Energy Budget, September 2025

The source library

Methodology

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 ↗