AI carbon footprint in EU-27 average: emissions per token and per query

Factor version soma-ai-ef-2026.10-dataset-v3, dated October 2026. Location-based grid averages, not renewable contracts. Estimates from published benchmarks.

How much CO₂ does an AI query emit when served from EU-27 average?

A query of 1256 tokens to a mid-size AI model (class B, such as gpt-4o) served from EU-27 average emits about 0.026 g CO₂e, counting electricity, the share of server hardware, data-centre construction and amortised training; a million tokens emit 0.0181 kg CO₂e. A small model (class A) emits 0.0077 g per query and a frontier model (class C) 0.1 g.

The EU-27 average is the factor to use when a model is served from the European Union but the country is unknown. It sits between the cleanest region, EU North - Sweden (0.00642 kg per million tokens for class B), and the highest, Asia Pacific - Singapore (0.0338 kg).

Emission factors for EU-27 average by model class

Per million tokens unless stated. Total = electricity + embodied hardware + data-centre construction + amortised training. Per query = 1256 tokens.
Model class	Electricity kg	Hardware kg	Construction kg	Training kg	Total kg CO₂e	g per query	Water L	mL per query
Class A (small)	0.00431	0.00022	0.00012	0.00060	0.005	0.0077	not available	not available
Class B (mid-size)	0.01396	0.00084	0.00038	0.00293	0.018	0.026	not available	not available
Class C (frontier)	0.05212	0.00323	0.00141	0.01284	0.070	0.1	not available	not available

Example: a company of 100 people where 80 use a class B assistant for 2,000 messages a month each sends about 2,412 million tokens a year. Served from EU-27 average, that is 44 kg CO₂e a year.

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.

How do the other regions compare?

Class B, total kg CO₂e per million tokens, lowest first: EU North - Sweden 0.00642, EU West - France 0.0073, EU North - Finland 0.00896, EU West - Belgium 0.0121, EU West - UK (London) 0.0181, Netherlands 0.02, EU West - Ireland 0.0209, US East - N. Virginia 0.0221, US West - Oregon 0.0233, EU Central - Germany 0.0257, US South Central - Texas 0.0264, US South - Southeast 0.0297, US Central - Iowa 0.0321, Asia Pacific - Japan 0.0333, Asia Pacific - Singapore 0.0338.

Where do these numbers come from and how do I cite them?

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 −42% to +45% for class A, −40% to +24% for class B and −67% to +58% for class C.

Dataset: AI Inference Emission and Resource Factors for Corporate GHG Inventories, Llopis, G., SOMA AI, October 2026, CC BY 4.0, concept DOI 10.5281/zenodo.20443585, which resolves to the latest version; the earlier version 2 values are record 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.

When citing a single value, name the region and class, for example "0.0181 kg CO₂e per million tokens, class B, EU-27 average, SOMA AI factors October 2026", and link to app.somaai.earth/ai-factors/region/eu-27.