#21 of 30 · Open-weight lab
Moonshot AI (Kimi)
A trillion-parameter open model trained on sanctioned H800s; no footprint published.
In plain words
How scores workMoonshot scores 38 out of 100 — laggard. Little or no disclosure, fossil-backed capacity, or active disputes with regulators and communities. Confidence in this score is mostly estimates (C).
What helps
- Efficiency · 1T total / 32B active mixture of experts with open weights; the company says it trained 'at a much lower cost than American companies'.
- Transparency · Open weights and technical reports for K2.
What hurts
- Efficiency · No measurement published by the company.
- Carbon · Chinese grid; no renewable procurement disclosed.
- Water · No water data.
What nobody knows
- Per-prompt energy disclosed
- Per-prompt water disclosed
- Annual environmental report
- Location-based Scope 2 reported
- Data-center locations public
- +3 more
Moonshot's Kimi K2 and K2 Thinking are open-weight, trillion-parameter mixture-of-experts models (32B active) trained on NVIDIA H800s bought before export sanctions; a CNBC-sourced $4.6 million training cost circulated widely but the CEO says it 'is not an official number'. Backed by Alibaba and Tencent, Moonshot serves Kimi from Chinese data centres and publishes no energy, water or emissions data.
Key figures
Per-prompt and fleet metrics. Hover a title for what it means; open “How we got this number” for the working.
ecogpt.co estimate; no company figure.
How we got this number
Kimi K2 is a trillion-parameter mixture-of-experts model with 32B active parameters — DeepSeek-V3's class — so about two thirds of a dense frontier model's work. Uncertainty roughly ±3×.
On-site cooling only, so comparable with Google and OpenAI's figures.
How we got this number
0.2 Wh per prompt × 1 L/kWh (estimated WUE for Chinese colocation sites). Counting the water evaporated to generate the electricity adds ≈0.36 mL.
Location-based: what the grid actually emitted. Renewable certificates would bring the market-based figure near zero.
How we got this number
0.2 Wh × 580 gCO₂/kWh (Chinese grid, location-based).
Location-based estimate for the AI fleet only; not comparable with a full corporate inventory.
How we got this number
39 MW × 8,760 h × 80% utilisation × 580 gCO₂/kWh (Chinese grid). Compute only — nothing else the company does is counted.
How we got this number
No inventory exists; serving a trillion-parameter model to a fast-growing user base roughly doubles compute year on year.
How we got this number
No disclosure; mainland Chinese sites mostly use evaporative cooling. 1.0 L/kWh is a mid-range assumption.
How we got this number
China's grid is about 35% non-fossil; Moonshot discloses no renewable purchases.
Hardware & materials
The chips, power, copper and future e-waste behind the fleet — the part sustainability reports leave out
GPUs, TPUs or custom accelerators the company runs. Analyst figures are in H100-equivalents, company figures in physical chips.
How we got this number
Moonshot trained K2 on pre-sanction NVIDIA H800s and discloses no fleet size; a DeepSeek-scale lab (≈50,000 Hopper GPUs) is the closest analogue, and Moonshot is smaller. ±2×.
Megawatts the AI fleet pulls. One megawatt runs roughly 750 US homes.
How we got this number
30,000 accelerators × 1.3 kW each, all-in. Fleet power implied by the chip count.
Terawatt-hours per year. One TWh is a year of electricity for about 95,000 US homes.
How we got this number
39 MW × 8,760 h × 80% utilisation. Compute fleet only.
CO₂e emitted making the chips and servers before they answer a single prompt.
How we got this number
30,000 accelerators × 514 kg CO₂e each (chip plus share of the host server). Emitted in manufacturing, before the first prompt is served.
Tonnes of server hardware that will be retired in three to five years.
How we got this number
30,000 accelerators × 17 kg of server hardware each. Tomorrow's e-waste unless reused; generative AI could produce 2.5 Mt of it a year by 2030.
Tonnes of copper in cabling, busbars and transformers.
How we got this number
39 MW × 27 t of copper per MW (cabling, busbars, transformers).
Diesel backup units or gas turbines at the data centres, permitted or observed.
How we got this number
39 MW ÷ 2.6 MW per generator, the average size permitted at US data centres. Permit counts are not published fleet-wide; this is the number a fleet this size normally carries.
Share of retired servers and parts given a second life.
Kilograms of antimony-equivalent, the life-cycle unit for scarce minerals consumed.
Score breakdown
Each pillar is 0–100; weights are fixed across the index
- 1T total / 32B active mixture of experts with open weights; the company says it trained 'at a much lower cost than American companies'. ImplicatorYicai Global
- No measurement published by the company. Yicai Global
- Chinese grid; no renewable procurement disclosed. Implicator
- No water data. Implicator
- Open weights and technical reports for K2. Implicator
- No environmental data; the widely quoted training cost is unofficial and the CEO declines to quantify it. Yicai Global
- No targets. Yicai Global
- No enforcement actions on record. Yicai Global
Models
What this company ships, and what each prompt costs
| Model | Weights | Energy / prompt | Water | Training CO₂e |
|---|---|---|---|---|
| Kimi K2 Thinking 1T total / 32B active (MoE) | Open | 0.20 Wh Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement. | 0.200 mL Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement. | 1,500 t Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement. |
Where the compute lives
Named sites from company disclosures and press reporting
- Beijing HQBeijing, China
- Alibaba Cloud / Tencent Cloud regionsChina