How to read an index score
The number
Every company's score is computed, not assigned. Six pillars are scored 0–100 against published rubric anchors on the methodology page, multiplied by fixed weights and summed. Carbon carries 25%, efficiency and transparency 20% each, water 15%, commitments and conduct 10% each. Mistral's 75 is 0.2 × 70 + 0.25 × 85 + 0.15 × 55 + 0.2 × 88 + 0.1 × 55 + 0.1 × 85, rounded. Changing a pillar input changes the score; there is no other lever.
The bands
Leader, 65 and above: discloses real numbers and runs mostly on clean power. Still has a footprint, but you can see it. Mistral, Google, Apple.
Middle of the pack, 45–64: some disclosure or clean-energy commitments, offset by rising emissions or blind spots. Most of the hyperscalers and the better Chinese clouds.
Laggard, below 45: little or no disclosure, fossil-backed capacity, or active disputes with regulators and communities. The pure labs, the image service, the neoclouds, and xAI.
The grade
Each pillar carries a confidence grade: A for company-disclosed figures with a method or independent measurement, B for company-level disclosure that is not AI-specific, C for figures derived from hosts, architecture or press, D for no disclosure at all. The company grade is the weighted average. Apple's 70 carries a B; Anthropic's 42 a D. A high score with a D grade means the company looks acceptable on the little it publishes, and the index says so on the profile: "In plain words".
What the score is not
A measure of absolute impact. Mistral's whole flagship model emitted 20 kt; Google's company emitted 11,500 kt. The score rewards disclosure, efficiency, clean supply, water design, dated targets and good conduct, all of which a small company can have and a large one can lack. The hardware block and the "sort by" options on the rankings show the scale separately.
How to use it
Read the band first, then the grade, then the pillar that surprises you. A leader with a low water pillar (Google) and a laggard with a high efficiency pillar (DeepSeek) are both telling you something specific, and the evidence under each pillar says what.
Companies in this post
Mistral AIThe only frontier lab with a peer-reviewed life-cycle assessment.75
Google (Gemini)The most detailed per-prompt disclosure — attached to the fastest-growing footprint.71
xAI (Grok)81 gas turbines across two sites, most of them beyond their permits.19
Anthropic (Claude)A safety-first brand with no environmental report — and gigawatts on order.42
Sources
- 1.Measuring the environmental impact of delivering AI at Google ScaleGoogle (arXiv 2508.15734) · Aug 21, 2025
- 2.Our contribution to a global environmental standard for AIMistral AI · Jul 22, 2025
- 3.In South Memphis, Elon Musk's Colossus operated gas turbines without appropriate permits, residents and activists claimInside Climate News · Jul 17, 2025
- 4.AWS Project Rainier: the AI compute cluster powered by Trainium chipsAmazon · Oct 29, 2025
Figures labelled as estimates on the index carry roughly ±3× uncertainty; disclosed figures are the company’s own. Corrections: team@ecogpt.com.



