Transparent AI Sustainability Data
How we calculate the Environmental Impact Score
Overview
GreenCodeAI ranks AI companies using the Environmental Impact Score (EIS), a comprehensive metric combining energy consumption, carbon emissions, water usage, transparency, and certifications. Our methodology is peer-reviewed and aligns with industry leaders like Google Cloud, Microsoft, and the GHG Protocol.
Externally Validated
Methodology validated against Google Cloud TTM approach, GHG Protocol Scope 2 Guidance, and Electricity Maps data.
Environmental Impact Score (EIS)
Final EIS Formula
EIS =
(15% × Per-Request Bonus) +
(25% × Renewable Energy Score) +
(15% × Efficiency Score) +
(15% × Carbon Score) +
(10% × Water Score) +
(10% × Transparency Score) +
(10% × Certification Score)
Why these weights? We prioritize metrics that companies actually disclose (renewable %, PUE) while reserving 15% for per-request metrics as an incentive for maximum transparency. No company can reach 100% without disclosing per-request data.
The 7 Metrics Explained
External Validation & Sources
Our methodology has been validated against industry-leading practices and scientific standards:
Google Cloud
Trailing-twelve-month (TTM) benchmarking and carbon-aware computing
2024 Environmental Report🇪🇺 EU Regulatory Pressure: Companies Must Report
As of 2025, major AI companies operating in Europe are legally required to disclose environmental metrics under the CSRD directive (~50,000 EU companies + ~10,000 non-EU companies including Microsoft, Google, Meta, Amazon).
Our methodology uses the same metrics mandated by EU law (PUE, energy consumption, carbon emissions, water usage) — making it easy to verify who's compliant and who's hiding data.
Note: GreenCodeAI is an independent transparency platform. We compile and publish public data — we do not audit or certify. For official certification, consult accredited auditors (Deloitte, PwC, KPMG, EY).
Comparison with Other Frameworks
| Framework | Regional CI | Dynamic Benchmarks | Prevents Double Counting |
|---|---|---|---|
| GreenCodeAI | ✓ Electricity Maps | ✓ TTM (6mo) | ✓ Adjusted weights |
| GHG Protocol | Recommends | Manual | Scopes separated |
| CDP Climate | Per facility | Annual | Guidance |
| MSCI ESG | Not specific | Annual | Partial |
Data Sources
We collect data from multiple verified sources to ensure accuracy and transparency:
Company ESG Reports
Annual sustainability reports from Google, Microsoft, Meta, Amazon, Anthropic, and other AI providers
Cloud Provider APIs
AWS Carbon Footprint Tool, GCP Carbon Footprint, Azure Sustainability Calculator
Scientific Datasets
ML CO₂ Impact, CodeCarbon, Electricity Maps API, World Resources Institute (Water Stress Index)
Public Registries
Datacenter locations (public records), grid carbon intensity (IEA, EIA), certification databases
Limitations & Disclaimers
Data availability varies
Not all companies publicly report environmental metrics. If data is not publicly disclosed, we mark it as NULL and assign a score of 0 for that metric. We do NOT estimate or invent missing data.
Independent platform
GreenCodeAI is not affiliated with the companies we rank. Scores reflect publicly available information and our transparent methodology.
Not financial advice
Rankings are for informational purposes only and should not be the sole basis for investment or purchasing decisions.
Subject to revision
As new data becomes available or methodologies improve (peer review), scores may be retroactively adjusted with full transparency.
Data Quality & Confidence Levels
All data collected by GreenCodeAI is classified according to source type and assigned a confidence level based on GreenCode AI's independent methodology.
Public Report
Official sustainability reports, environmental reports, or CSRD disclosures published by the company.
Examples: Annual Environmental Report, CSRD Sustainability Report, GHG Protocol Disclosure
API Data
Real-time or periodic data obtained through official company APIs.
Examples: Cloud provider usage APIs, Energy dashboards with API access
Blog Post
Technical blog posts or engineering posts discussing environmental metrics.
Examples: Company engineering blogs, sustainability update posts
Self-reported
Metrics provided directly by companies through forms or submissions.
Examples: Company-submitted data, voluntary disclosures
Transparency on Each Company Profile
Every company profile displays its data source type, confidence level according to GreenCode AI Methodology, original source link, report year (if applicable), and last update date. This ensures users can make informed decisions about data reliability.
Transparency Tiers - Data Disclosure Levels
GreenCodeAI classifies companies into three transparency tiers based on the level of environmental data disclosure. Higher transparency receives score bonuses and builds user trust.
Full Transparency
Reports per-request metrics (kWh/request or CO₂/request) publicly with quarterly or annual sustainability reports. Provides datacenter-level metrics (PUE, WUE, renewable %).
Criteria:
- • Reports kWh/request or CO₂/request publicly
- • Publishes quarterly or annual sustainability reports
- • Provides datacenter-level metrics (PUE, WUE, renewable %)
Examples: Google Cloud, Microsoft Azure
Partial Transparency
Publishes annual sustainability reports with aggregate metrics (total energy, total CO₂) but does NOT report per-request metrics.
Criteria:
- • Publishes annual sustainability reports
- • Reports aggregate metrics (total energy, total CO₂)
- • Does NOT report per-request metrics
Examples: Anthropic, Meta AI, AWS
Limited Transparency
Limited or no public sustainability data. Data collected from blog posts, press releases, or not available. No formal sustainability reports.
Criteria:
- • Limited or no public sustainability data
- • Data from blog posts or press releases only
- • No formal sustainability reports
Examples: Most AI startups
Transparency Policy
If data is not publicly disclosed, we do NOT invent it. Companies without disclosed metrics receive NULL values and score = 0 for missing metrics. This incentivizes transparency and builds user trust.
Company Clusters - Fair Peer Comparison
Companies are grouped into 7 clusters based on business model and infrastructure profile. This enables fair peer-to-peer comparison between companies with similar operational characteristics. Clustering is for comparison only — NOT for data estimation.
| Cluster | Classification | Classification Criteria | Business Model |
|---|---|---|---|
| C1 | Hyperscaler Cloud AWS, Azure, Google Cloud | ≥30 regions OR ≥100 AZs | Global cloud infrastructure provider |
| C2 | LLM Consumer API OpenAI, Anthropic | ≥10M monthly visitors | Direct-to-consumer AI API service |
| C3 | LLM Enterprise API Cohere, Mistral AI | ≥30% enterprise customers | B2B AI API for enterprises |
| C4 | Open-Source Platform Hugging Face | ≥100K models hosted | Model hosting and collaboration platform |
| C5 | AI Research Lab Meta AI, DeepMind | ≥10 papers/year | Research-first organization |
| C6 | Vertical AI SaaS Jasper, Copy.ai, Midjourney | Specialized AI application | Vertical-specific AI software |
| C7 | Emerging Startup xAI, others | Founded <3 years ago | Early-stage AI company |
Important: Clusters ≠ Estimation
Company clustering is used ONLY for fair comparison between similar companies. It does NOT involve estimating missing data. Companies without disclosed metrics still receive NULL values and score = 0 per our Transparency Policy.
Community Contributions
We welcome corrections, suggestions, and contributions from the community:
How to Contribute:
Report Inaccuracies
Email corrections to team@greencodeai.eu
Suggest Data Sources
Email your proposals to team@greencodeai.eu
Peer Review
Help validate our methodology and calculations via email