CDP Climate Compliance

Guide for reporting AI emissions in CDP Climate Change questionnaire.

What is CDP?

CDP (formerly Carbon Disclosure Project) runs the global environmental disclosure system. Companies report emissions through annual questionnaires, which are scored and published.

Relevant Module

AI inference emissions belong in:

Module 7: Emissions Breakdown - Section: Scope 3 emissions - Category: Category 1 (Purchased goods and services)

Export Format

Generate CDP-compliant export:

openclaw green export --format cdp --period 2025

Output Structure:

{
  "reportingYear": 2025,
  "scope3": {
    "category1": {
      "emissions_tCO2eq": 0.01245,
      "methodology": "hybrid",
      "methodologyDescription": "Per-token emission factors estimated from academic research (Lacoste et al. 2019, Patterson et al. 2022) with conservative fallbacks. Factors account for GPU power consumption during inference with 3:1 output-to-input energy ratio.",
      "dataQuality": "calculated",
      "percentageCalculatedUsingPrimaryData": 0,
      "emissionFactorSources": [
        "ML CO2 Impact Calculator (Lacoste et al., 2019)",
        "Cloud Carbon Footprint methodology",
        "CodeCarbon hardware measurements"
      ]
    }
  },
  "intensity": [
    {
      "metric": "CO2 per million tokens",
      "value": 142.29,
      "unit": "gCO2eq/1M tokens"
    },
    {
      "metric": "CO2 per API call",
      "value": 6.74,
      "unit": "gCO2eq/call"
    }
  ]
}

CDP Questions Mapping

C6.5 - Scope 3 Emissions by Category

Field Response
Category Category 1: Purchased goods and services
Scope 3 emissions (metric tons CO2e) [from export]
Percentage calculated using primary data 0%
Explanation AI inference API services

C6.5a - Scope 3 Category 1 Details

Field Response
Description of activity AI inference API calls to cloud providers
Emission factor used Per-token factors from academic research
Emission factor source Lacoste et al. 2019, Patterson et al. 2022
Methodology Hybrid method

C7.9 - Intensity Metrics

Field Response
Intensity figure [intensityPerMillionTokens]
Metric numerator Metric tons CO2e
Metric denominator Million tokens processed
Scope(s) Scope 3

Data Quality Mapping

CDP uses different terminology than GHG Protocol:

CDP Term Description OpenClaw Confidence
Measured Direct measurement N/A (not available)
Calculated Calculated from activity data ≥50%
Estimated Estimated from proxies <50%

Most AI emissions are calculated (activity data = tokens × factors).

Intensity Metrics

CDP requests emissions intensity metrics. OpenClaw provides:

  1. Per million tokens — Standard for AI workloads
  2. Per API call — Alternative activity metric

These enable year-over-year comparison even as usage grows.

Methodology Description

Use this template for CDP methodology disclosure:

AI inference emissions are calculated using the hybrid method
combining secondary emission factors with activity data (token counts).

Per-model emission factors are derived from:
- Academic research on ML energy consumption (Lacoste et al. 2019)
- Published hardware power measurements (CodeCarbon project)
- Model size heuristics based on published architectures

Factors account for:
- GPU power consumption during inference
- 3:1 output-to-input energy ratio (iterative generation)
- Data center PUE of 1.2 (cloud provider average)
- Grid carbon intensity of [X] gCO2/kWh

Limitations:
- No supplier-specific emission data available
- Model architectures not fully disclosed by providers
- Data center locations and energy sources unknown

Scoring Considerations

CDP scores responses based on:

  1. Completeness — Report all material categories
  2. Transparency — Disclose methodology and limitations
  3. Ambition — Set and track reduction targets
  4. Leadership — Engage suppliers, improve data quality

To improve score: - Report every year consistently - Set SBTi-aligned targets (see SBTi Guide) - Document efforts to obtain supplier-specific data - Show year-over-year progress

Supplier Engagement

CDP encourages engaging suppliers for better data. For AI providers:

  1. Request emissions data through official channels
  2. Document attempts and responses
  3. Collaborate on methodology improvements
  4. Advocate for provider transparency

Example Disclosure

## Scope 3, Category 1: AI Services

We use third-party AI inference APIs for [use case].
Emissions are calculated using per-token factors from academic
research, as providers do not yet disclose per-request emissions.

| Metric | 2024 | 2025 | Change |
|--------|------|------|--------|
| Emissions (tCO2e) | 0.010 | 0.012 | +20% |
| API calls | 1,500 | 1,847 | +23% |
| Intensity (g/call) | 6.67 | 6.74 | +1% |

Intensity remained stable despite usage growth, indicating
efficiency improvements from model selection.

**Data quality**: Calculated (hybrid method)
**Uncertainty**: ±30%

We are engaging AI providers to obtain supplier-specific
emission factors and improve data quality.
  • GHG Protocol — Foundation for CDP reporting
  • TCFD — CDP aligns with TCFD recommendations
  • SBTi — Targets recognized by CDP