AI carbon footprint in EU-27 average: emissions per token and per query
Factor version soma-ai-ef-2026.10-dataset-v3.0.0. Last updated . Location-based grid averages, not renewable contracts. Estimates from published benchmarks.
Part of the SOMA AI Emission Factors cluster.
How much CO₂ does an AI query emit when served from EU-27 average?
A 1210-token query to a class B model (gpt-4o) served from EU-27 average emits about 0.029 g CO₂e, and a million tokens emit 0.0212 kg CO₂e.
It sits between EU North - Sweden (0.00714 kg per million tokens for class B) at the low end and Asia Pacific - Singapore (0.0401 kg) at the high end. The factor applies when a model is served from the European Union and the country is unknown.
Which grid supplies the electricity in EU-27 average?
This page uses the EU row of the dataset's grid table, 236 g CO₂e per kWh in 2023, an aggregate and not a single grid. The source is Ember yearly electricity data, CO2 intensity 2023, 236.0 g/kWh (lifecycle, IPCC AR5). For comparison, the global average is 483.2 g per kWh, 2.0 times the EU-27 figure.
No provider region in the dataset's cloud mapping table resolves to the EU grid. The row is the fallback the dataset applies when no region is known or a model is served from the EU without a country, so it has no AWS, Azure or Google Cloud region id of its own.
Which grids have an intensity close to EU?
| Difference g | Grid | Code | g CO₂e per kWh | Ratio |
|---|---|---|---|---|
| -0.4 | United Kingdom | GBR | 235.6 | 0.998 |
| -3.9 | Lao People's Democratic Republic (the) | LAO | 232.1 | 0.983 |
| 7.2 | Romania | ROU | 243.2 | 1.031 |
| 8.9 | French Guiana | GUF | 244.9 | 1.038 |
| -10.9 | Slovenia | SVN | 225.1 | 0.954 |
| -11.6 | Cameroon | CMR | 224.4 | 0.951 |
| 14.0 | Cook Islands (the) | COK | 250 | 1.059 |
| -16.9 | WECC California | CAMX | 219.1 | 0.928 |
| 22.4 | Colombia | COL | 258.4 | 1.095 |
| 26.4 | Armenia | ARM | 262.4 | 1.112 |
| Region | Class B kg | Difference kg | Ratio |
|---|---|---|---|
| EU North - Sweden | 0.00714 | -0.0141 | 0.34 |
| EU West - France | 0.0082 | -0.0130 | 0.39 |
| EU North - Finland | 0.0102 | -0.0110 | 0.48 |
| EU West - Belgium | 0.014 | -0.0072 | 0.66 |
| EU South - Spain | 0.0165 | -0.0047 | 0.78 |
| EU West - UK (London) | 0.0212 | 0.0000 | 1.00 |
| Netherlands | 0.0235 | 0.0023 | 1.11 |
| EU West - Ireland | 0.0245 | 0.0033 | 1.16 |
| US East - N. Virginia | 0.0261 | 0.0048 | 1.23 |
| US West - Oregon | 0.0275 | 0.0062 | 1.29 |
| EU Central - Germany | 0.0303 | 0.0091 | 1.43 |
| US South Central - Texas | 0.0312 | 0.0099 | 1.47 |
| US South - Southeast | 0.0351 | 0.0139 | 1.66 |
| US Central - Iowa | 0.0381 | 0.0168 | 1.79 |
| Asia Pacific - Japan | 0.0395 | 0.0183 | 1.86 |
| Asia Pacific - Singapore | 0.0401 | 0.0189 | 1.89 |
Emission factors for EU-27 average by model class
| Model class | Electricity kg | Hardware kg | Construction kg | Training kg | Total kg CO₂e | g per query | Water L | mL per query |
|---|---|---|---|---|---|---|---|---|
| Class A (small) | 0.00514 | 0.00027 | 0.00014 | 0.00060 | 0.006 | 0.0084 | not available | not available |
| Class B (mid-size) | 0.01682 | 0.00101 | 0.00046 | 0.00293 | 0.021 | 0.029 | not available | not available |
| Class C (frontier) | 0.06265 | 0.00388 | 0.00169 | 0.01284 | 0.081 | 0.11 | not available | not available |
| Model class | Central | Class range low | Class range high | High case | B200 | Development |
|---|---|---|---|---|---|---|
| Class A | 0.006149 | 0.003901 | 0.00879 | 0.009617 | 0.005439 | 0.007006 |
| Class B | 0.02122 | 0.01416 | 0.02538 | 0.03262 | 0.01544 | 0.02294 |
| Class C | 0.08107 | 0.03574 | 0.118 | 0.1235 | 0.08314 | 0.08543 |
How much does siting matter for a model served from EU-27 average?
Electricity is 79.3% of the class B total here (class A 83.6%, class C 77.3%), with hardware at 4.8%, construction 2.1% and training 14%. Electricity carries most of the 0.0212 kg, so the grid is the biggest lever.
The example company (2,323 million class B tokens a year) emits 49 kg CO₂e from EU-27 average. Moving to EU West - UK (London) would save 0.066 kg (0.13%) or EU South - Spain would save 11 kg (22%).
| Region | Grid g per kWh | Class A | Class B | Class C |
|---|---|---|---|---|
| EU West - UK (London) | 235.6 | 0.0202 | 0.0663 | 0.247 |
| EU South - Spain | 169.7 | 3.35 | 11 | 40.9 |
How much water does AI usage consume in EU-27 average?
Water: not available for EU-27 average. The dataset publishes no off-site water factor for this region, so SOMA prints no water figure rather than a partial one.
Where do these numbers come from and how do I cite them?
Figures per query are one plain exchange, before the Tier 2b reasoning uplift. Electricity term. GPU energy per token from the ML.ENERGY Leaderboard v3, a facility overhead of 1.20, and grid carbon intensity from EPA eGRID 2023 and Ember 2023. Embodied hardware, data-centre construction and amortised training terms are documented on the methodology page. Water combines on-site cooling water (Li et al. 2025) with the water embedded in grid electricity (Reig et al. 2020). The modelled range of the models inside a class, around the class value, is −41% to +48% for class A, −39% to +24% for class B and −67% to +55% for class C.
Dataset: AI Inference Emission and Resource Factors for Corporate GHG Inventories (SOMA AI Emission Factors (SAEF)), Llopis-Garcia, G., SOMA AI, 19 October 2026, CC BY 4.0, concept DOI 10.5281/zenodo.20443585, which resolves to the latest version (this version 10.5281/zenodo.23371447, earlier version 2 values at 10.5281/zenodo.22767475). Method: Accounting for Purchased AI Inference in Corporate GHG Inventories: A Token-Based Method, arXiv:2606.10660, under peer review.
Partners can fetch any row from the AI factors API, and the whole table is a CSV download.
Cite as "0.0212 kg CO₂e per million tokens, class B, EU-27 average, SAEF, 19 October 2026" and link somaai.earth/ai-footprint/factors/eu-27.