#13 of 30 · Open-weight lab + cloud
Alibaba Cloud (Qwen)
Open models anyone can benchmark; own data centres now 74% clean electricity.
In plain words
How scores workAlibaba scores 52 out of 100 — middle of the pack. Some disclosure or clean-energy commitments, offset by rising emissions or blind spots. Confidence in this score is mostly estimates (C).
What helps
- Carbon · Alibaba Cloud's self-built data centres reached 73.6% clean electricity in FY2026 (+9.6 points); group-wide 62.2%.
- Commitments · Scope 1+2 carbon neutrality by 2030 and a 50% cut in Scope 3 intensity.
What hurts
- Carbon · Remaining load sits on China's coal-heavy grid, and a RMB 380bn AI build-out is under way.
- Water · No water intensity or consumption figures for cloud operations are published in English-language disclosures.
What nobody knows
- Per-prompt energy disclosed
- Per-prompt water disclosed
- Training emissions in model cards
- Water-use effectiveness reported
- Third-party verified AI figures
- +1 more
Qwen is the most widely used open-weight model family out of China and ships in sizes from sub-1B to very large mixture-of-experts. Alibaba's FY2026 ESG report puts group clean electricity at 62.2% and Alibaba Cloud's own data centres at 73.6%, with net emissions of 2.3 Mt down 4.6% — a real improvement — but it discloses nothing about the energy or water behind Qwen itself, and is spending RMB 380 billion on AI infrastructure over three years.
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
Qwen3's flagship is a 235B mixture-of-experts model with 22B active parameters, and most traffic goes to smaller sizes; comparable to DeepSeek-V3's class and well below dense frontier models. 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 hyperscale 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). Alibaba Cloud's own sites are 73.6% clean electricity, so its market-based figure would be far lower.
How we got this number
No fleet WUE. Zhangbei uses free air cooling; most Chinese hyperscale sites use evaporative cooling. 1.0 L/kWh sits between hyperscaler best practice (0.15) and the US average (1.8).
Group-wide 62.2%.
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
Alibaba publishes no chip count. A RMB 380 billion (≈$52bn) three-year AI and cloud budget at roughly $25–40k per installed accelerator implies several hundred thousand, and the company says it is rationing GPU access. Uncertainty ±2×.
Megawatts the AI fleet pulls. One megawatt runs roughly 750 US homes.
How we got this number
300,000 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
390 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
300,000 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
300,000 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
390 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
390 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
- Open weights across many sizes allow independent measurement; no first-party per-prompt data. Hugging Face, Salesforce, Cohere, Carnegie Mellon
- Alibaba Cloud's self-built data centres reached 73.6% clean electricity in FY2026 (+9.6 points); group-wide 62.2%. iTiger (Alibaba ESG report)
- Remaining load sits on China's coal-heavy grid, and a RMB 380bn AI build-out is under way. Data Center Dynamics
- No water intensity or consumption figures for cloud operations are published in English-language disclosures. Alibaba Group
- Annual ESG report with clean-electricity share and net emissions, plus open weights; no AI-specific energy, water or training-emissions data. Alibaba GroupiTiger (Alibaba ESG report)
- Scope 1+2 carbon neutrality by 2030 and a 50% cut in Scope 3 intensity. Alibaba Group
- No environmental enforcement actions on record in the sources tracked here. Alibaba Group
Models
What this company ships, and what each prompt costs
| Model | Weights | Energy / prompt | Water | Training CO₂e |
|---|---|---|---|---|
| Qwen 3 0.6B–235B (dense + 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. | 3,400 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
- HangzhouZhejiang, China
- ZhangbeiHebei, China · Wind-powered campus.
- SingaporeSingapore
Timeline
Incidents and developments involving this company
Alibaba Cloud's own data centres reach 73.6% clean electricity as net emissions fall 4.6%
Alibaba's FY2026 ESG report puts group clean electricity at 62.2%, Alibaba Cloud's self-built data centres at 73.6% (+9.6 points) and net Scope 1+2 emissions at 2.296 Mt, down 4.6% — while a RMB 380 billion AI build-out is under way.
AlibabaAI Energy Score launches star ratings for 166 models across 10 tasks
Hugging Face, Salesforce, Cohere and Carnegie Mellon publish an Energy Star-style leaderboard: GPU watt-hours per task, with five relative star bands. Open models are tested automatically; closed models through a sealed sandbox.
Salesforce, Hugging Face, Cohere, Meta, Mistral, Alibaba, DeepSeek, IBM