AI4AI for AI platforms

Show your customers the footprint of their AI.

Your enterprise customers are starting to ask what their AI use costs the planet. Send us your usage and AI4AI gives each customer their electricity, water and CO2e, a verified monthly statement and a badge for your product, all on a peer-reviewed method.
API reference

Response, trimmed and rounded

{
  "model": "claude-sonnet-4-5",
  "model_class": "B",
  "region": "eu-west-1",
  "kg_co2e_per_million_tokens": {
    "central": 0.02454
  },
  "high_case": 0.03809,
  "litres_water_per_million_tokens": 0.106,
  "uncertainty_pct": 39,
  "factor_version": "soma-ai-ef-2026.11-dataset-v3.1.0",
  "dataset_doi": "10.5281/zenodo.20443585"
}

Not just numbers. A footprint you can put your name next to.

Your product · this monthExample

4.2 kg

CO2e

13 kWh

Electricity

24 L

Water

carbon data powered by SOMA Verified

Usage in, footprint out

Send your token logs. We compute every request and keep the totals per customer, so your team doesn't build or store any of it.

Verified statements

Issue a monthly AI footprint statement for each customer, ready for their Scope 3 report, with a public page their auditor can check.

The SOMA badge

Embed the verified footprint in your product with one line. "Carbon data powered by SOMA", linked to a method published by Elsevier.

Maintained for you

We map new models and update factors on our side. Signed webhooks tell your system the moment a statement is issued or a factor changes.

Factors for your own stack

Share data from your own hardware or data centre and we build factors for it, so your numbers reflect how you actually run.

A team behind it

When your customer or their auditor asks how a number was made, the people who wrote the method answer.

Three steps to proof.

  1. 1

    Send your usage

    Token counts per request or per day, tagged with your customer's id. Any model, any provider, any cloud region.

  2. 2

    We do the accounting

    Carbon, electricity and water for every record, totalled per customer, with the factor version frozen.

  3. 3

    Give your customers proof

    Issue a verified statement each month and show the live footprint in your product with the SOMA badge.

1 · Send usage

curl -X POST "https://app.somaai.earth/api/v1/ai-usage" \
  -H "Authorization: Bearer sk_soma_..." \
  -d '{"records":[{"customer_ref":"acme","usage_date":"2026-11-20",
       "model":"claude-sonnet-4-5","region":"eu-west-1",
       "input_tokens":184000,"output_tokens":52000}]}'

3 · Issue a verified statement

POST /api/v1/ai-statements
{ "customer_ref": "acme", "period_start": "2026-11-01", "period_end": "2026-11-30" }

→ { "code": "K7M2Q9XRTB",
    "verify_url": "https://app.somaai.earth/verify/K7M2Q9XRTB",
    "classification": "Scope 3 Category 1", ... }

Any model id, any AWS, GCP or Azure region, or a country. Factor lookups and a change feed are included.

For teams that ship AI to other people.

AI platforms and gateways

Show the footprint of every request next to its cost and latency.

Carbon accounting software

Add a traceable AI line to your customers' Scope 3 inventory.

FinOps and observability

Put electricity, water and CO2e beside the token bill.

Numbers your customers can trust.

Every figure comes from a method published by Elsevier in a peer-reviewed Q1 journal. Tell us about your platform and we'll show you what it looks like with your models.

Prototyping or doing research? The method and the factors are open. Start with the saef Python package, then talk to us when you go to production.

PythonLangChainLlamaIndexHaystackMCPDeveloper tools