AI carbon footprint in Asia Pacific - Japan: 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 Asia Pacific - Japan?
A 1210-token query to a class B model (gpt-4o) served from Asia Pacific - Japan emits about 0.054 g CO₂e, and a million tokens emit 0.0395 kg CO₂e.
Rank 15 of 16 for class B, lowest carbon first (full table). 5.5x the level of EU North - Sweden, Asia Pacific - Singapore is 1.0x higher.
Which grid supplies the electricity in Asia Pacific - Japan?
Asia Pacific - Japan maps to Japan, a national grid, at 492.6 g CO₂e per kWh for 2023. That is 2.1 times the EU-27 average (236 g) and 1.0 times the global average (483.2 g). The source is Ember yearly electricity data, CO2 intensity 2023, 492.6 g/kWh (lifecycle, IPCC AR5).
Cloud regions on this grid, 6 in all (2 AWS, 2 Azure, 2 Google Cloud).
| Provider | Region id | Region name | Location | Mapping confidence | Mapping note |
|---|---|---|---|---|---|
| AWS | ap-northeast-1 | Asia Pacific (Tokyo) | Tokyo | high | |
| AWS | ap-northeast-3 | Asia Pacific (Osaka) | Osaka | high | |
| Azure | japaneast | Japan East | Tokyo, Saitama | high | |
| Azure | japanwest | Japan West | Osaka | high | |
| Google Cloud | asia-northeast1 | asia-northeast1 | Tokyo, Japan | high | |
| Google Cloud | asia-northeast2 | asia-northeast2 | Osaka, Japan | high |
Which grids have an intensity close to Japan?
| Difference g | Grid | Code | g CO₂e per kWh | Ratio |
|---|---|---|---|---|
| 1.0 | Malta | MLT | 493.6 | 1.002 |
| -1.4 | Cambodia | KHM | 491.2 | 0.997 |
| 2.3 | Turkey | TUR | 494.9 | 1.005 |
| -3.6 | Mexico | MEX | 489 | 0.993 |
| 4.4 | Guadeloupe | GLP | 497 | 1.009 |
| 4.6 | RFC Michigan | RFCM | 497.2 | 1.009 |
| 7.4 | Kiribati | KIR | 500 | 1.015 |
| 7.4 | Vanuatu | VUT | 500 | 1.015 |
| -7.7 | Macao | MAC | 484.9 | 0.984 |
| 8.3 | Singapore | SGP | 500.9 | 1.017 |
Emission factors for Asia Pacific - Japan 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) | 0.01072 | 0.00027 | 0.00014 | 0.00060 | 0.012 | 0.016 | 0.065 | 0.09 |
| Class B (mid-size) | 0.03512 | 0.00101 | 0.00046 | 0.00293 | 0.040 | 0.054 | 0.213 | 0.29 |
| Class C (frontier) | 0.13077 | 0.00388 | 0.00169 | 0.01284 | 0.149 | 0.2 | 0.793 | 1.1 |
| Model class | Central | Class range low | Class range high | High case | B200 | Development |
|---|---|---|---|---|---|---|
| Class A | 0.01173 | 0.007238 | 0.01702 | 0.01885 | 0.01028 | 0.01462 |
| Class B | 0.03952 | 0.02544 | 0.048 | 0.06284 | 0.02762 | 0.04789 |
| Class C | 0.1492 | 0.05834 | 0.2236 | 0.2361 | 0.1535 | 0.1783 |
How much does siting matter for a model served from Asia Pacific - Japan?
Electricity is 88.9% of the class B total here (class A 91.4%, class C 87.7%), with hardware at 2.6%, construction 1.2% and training 7.4%. Electricity carries most of the 0.0395 kg, so the grid is the biggest lever.
The example company (2,323 million class B tokens a year) emits 92 kg CO₂e from Asia Pacific - Japan. Moving to US Central - Iowa would save 3.4 kg (3.7%) or US South - Southeast would save 10 kg (11%).
| Region | Grid g per kWh | Class A | Class B | Class C |
|---|---|---|---|---|
| US Central - Iowa | 472.1 | 1.04 | 3.4 | 12.7 |
| US South - Southeast | 431 | 3.11 | 10.2 | 38 |
How much water does AI usage consume in Asia Pacific - Japan?
A class B query of 1210 tokens from Asia Pacific - Japan consumes about 0.29 mL of water and a million tokens 0.213 litres, or 0.155 litres without the water attributed to hydropower.
| Input | Value |
|---|---|
| On-site WUE, L per kWh | 1.22 |
| WUE denominator | IT load |
| WUE basis as published | withdrawal |
| Withdrawal to consumption | 0.6667 |
| EWIF, L per kWh | 2.309 |
| EWIF without hydro, L per kWh | 1.494 |
| AWARE factor | 0.64 |
| Confidence | published-site |
| Source | AWS; AWS Region Asia-Pacific (Tokyo) ap-northeast-1; 2025 |
| Model class | On site | Grid electricity | Total | Total without hydro | AWARE-weighted on site |
|---|---|---|---|---|---|
| Class A | 0.0148 | 0.0503 | 0.065 | 0.0473 | 0.0094 |
| Class B | 0.0483 | 0.165 | 0.213 | 0.155 | 0.0308 |
| Class C | 0.18 | 0.613 | 0.793 | 0.577 | 0.115 |
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 "0.0395 kg CO₂e per million tokens, class B, Asia Pacific - Japan, SAEF, 19 October 2026" and link somaai.earth/ai-footprint/factors/japan.