AI inference emission factors by model class and cloud region

Factor version soma-ai-ef-2026.09-lifecycle-v2, dated . Location-based grid averages, not renewable contracts. Estimates from published benchmarks.

What is the carbon footprint of AI usage per query and per million tokens?

A typical query of 400 tokens to a mid-size AI model (class B, the gpt-4o class) served from the global-average grid emits about 0.029 g CO₂e, and a million tokens emit 0.0721 kg CO₂e, counting electricity, the share of server hardware and amortised training. The same query served from EU Sweden emits 0.0052 g and from Japan 0.035 g, a 6.6 times difference from region alone.

The table below is the full factor set: three model classes across 16 regions, in kg CO₂e per million tokens (all three terms) and litres of water per million tokens. Each row links to a page with the per-query figures, the split between electricity, hardware and training, and the comparison with the other regions or classes.

Factor table

Carbon in kg CO₂e per million tokens (electricity + embodied hardware + amortised training). Water in litres per million tokens. Class A = small models, B = mid-size, C = frontier.
RegionCO₂e ACO₂e BCO₂e CWater AWater BWater C
EU Belgium0.0070.0290.0500.1380.5580.710
EU Finland0.0050.0200.0390.1380.5580.710
EU France0.0040.0160.0330.1380.5580.710
EU Germany0.0170.0660.0970.1150.4640.590
EU Ireland0.0130.0530.0800.0600.2410.307
EU Netherlands0.0130.0500.0770.1400.5650.719
EU Sweden0.0040.0130.0300.2440.9851.253
Global average0.0180.0720.1040.1440.5820.740
Japan0.0220.0870.1230.0870.3500.446
Singapore0.0210.0830.1190.1290.5200.661
UK (London)0.0110.0450.0710.0900.3620.461
US Central (Iowa)0.0190.0740.1080.1130.4580.583
US East0.0130.0510.0780.1000.4040.514
US South0.0170.0690.1010.1440.5820.740
US Texas0.0150.0610.0900.0600.2420.308
US West (Oregon)0.0130.0530.0810.1440.5820.740

Electricity-only values (the term most other sources publish) are lower: for class B they run from 0.006 kg in EU Sweden to 0.080 kg in Japan per million tokens. The methodology page shows the electricity term on its own.

Which model class does my model belong to?

  • Class A — Small: gpt-4o-mini, claude-haiku, gemini-flash, llama-8b. About 0.007 g CO₂e and 0.058 mL of water per 400-token query on the global-average grid.
  • Class B — Mid: gpt-4o, claude-sonnet, llama-70b. About 0.029 g CO₂e and 0.23 mL of water per 400-token query on the global-average grid.
  • Class C — Frontier: claude-opus, gpt-4, o1, large MoE models. About 0.041 g CO₂e and 0.3 mL of water per 400-token query on the global-average grid.

Providers do not publish model sizes, so the class comes from the published tier and benchmark behaviour. Misclassifying a model is the largest single source of error, about four to five times, which is why SOMA maintains the mapping. Reasoning models (o1, o3, DeepSeek-R1, extended thinking modes) generate many more tokens per answer; the token count captures that when it comes from billing.

Which cloud region has the lowest AI emissions?

EU Sweden, EU France and EU Finland have the lowest carbon per token for every model class, because their grids run largely on hydro, nuclear and wind. Japan and Singapore are the highest. Water follows a different pattern: cooler grids can use more on-site water per kWh, so the lowest-carbon region is not always the lowest-water region.

Where do these factors 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 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). Uncertainty on the electricity term is ±40% for class A and ±50% for classes B and C.

Dataset: AI Inference Emission and Resource Factors for Corporate GHG Inventories, Guillermo Llopis, SOMA AI, September 2026, CC BY 4.0, concept DOI 10.5281/zenodo.20443585 (this version 10.5281/zenodo.22767475). Method: Accounting for AI Inference in Corporate GHG Inventories: A Four-Tier Methodology for Scope 3 Category 1 Reporting, arXiv:2606.10660, under peer review.

Partners can fetch any row from the AI factors API.

Read how to account for AI usage in a Scope 3 inventory under CSRD and why published estimates of a ChatGPT query's footprint differ by 150 times.