#26 of 30 · AI lab + cloud
ByteDance (Doubao / Volcano Engine)
China's biggest Nvidia buyer; 50 trillion tokens a day; no emissions data.
In plain words
How scores workByteDance scores 34 out of 100 — laggard. Little or no disclosure, fossil-backed capacity, or active disputes with regulators and communities. Confidence in this score is little data (D).
What helps
- Carbon · Pledged operational carbon neutrality (≥90% real cuts) and 100% renewable electricity by 2030; Norway site to run on hydro.
- Commitments · Operational carbon neutrality by 2030 with at least 90% real reductions; 100% renewable electricity by 2030.
What hurts
- Carbon · Chinese grid plus Malaysia's gas-heavy grid; no renewable procurement disclosed.
- Water · Johor's 5.8 GW data-centre pipeline is straining local water and power enough that officials slowed approvals; ByteDance discloses nothing about its own use.
- Transparency · No environmental report, emissions, energy or footprint in model cards; Greenpeace ranked it third from bottom among Chinese cloud providers.
What nobody knows
- Per-prompt energy disclosed
- Per-prompt water disclosed
- Training emissions in model cards
- Annual environmental report
- Location-based Scope 2 reported
- +3 more
ByteDance's Doubao is China's most-used AI app (157 million monthly users) and Volcano Engine reported more than 50 trillion tokens a day in late 2025. ByteDance was the largest Chinese buyer of Nvidia chips in 2025 (≈RMB 85 billion) and planned a $14 billion H200 order for 2026, with data centres anchored in Johor, Malaysia — a region now slowing approvals over power and water strain. It pledged operational carbon neutrality and 100% renewable electricity by 2030 but has never published emissions or energy figures.
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
Doubao's models are mixture-of-experts designs with modest active parameters, comparable to DeepSeek-V3's class, served at enormous scale (50 trillion tokens a day). 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 and Malaysian 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
650 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
2026's chip budget exceeds 2025's RMB 85 billion; with no inventory to measure, a 50% rise in AI compute is the conservative reading.
How we got this number
No disclosure. Johor and 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 and Malaysia's about 20%; ByteDance discloses no renewable purchases despite its 2030 pledge.
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
ByteDance publishes no fleet size. It spent about RMB 85 billion (≈$12bn) on AI chips in 2025 alone — more than any other Chinese company — which at $25–40k per installed accelerator is 300,000–500,000 chips in a single year. Uncertainty ±2×.
Megawatts the AI fleet pulls. One megawatt runs roughly 750 US homes.
How we got this number
500,000 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
650 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
500,000 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
500,000 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
650 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
650 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
- Doubao models are MoE with modest active parameters, served at 50 trillion tokens a day; nothing measured or published. Analytics Insight
- Chinese grid plus Malaysia's gas-heavy grid; no renewable procurement disclosed. CO/AIData Center Dynamics
- Pledged operational carbon neutrality (≥90% real cuts) and 100% renewable electricity by 2030; Norway site to run on hydro. Data Center Dynamics
- Johor's 5.8 GW data-centre pipeline is straining local water and power enough that officials slowed approvals; ByteDance discloses nothing about its own use. CO/AI
- No environmental report, emissions, energy or footprint in model cards; Greenpeace ranked it third from bottom among Chinese cloud providers. Data Center DynamicsData Center Dynamics
- Operational carbon neutrality by 2030 with at least 90% real reductions; 100% renewable electricity by 2030. Data Center Dynamics
- A value-chain decarbonisation plan promised for end-2023 has not appeared. Data Center Dynamics
- No enforcement actions; expanding fastest in a region where regulators have paused approvals over grid and water strain. CO/AI
Models
What this company ships, and what each prompt costs
| Model | Weights | Energy / prompt | Water | Training CO₂e |
|---|---|---|---|---|
| Doubao 1.5 Pro MoE (≈20B active, undisclosed total) | Closed | 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. | 3,000 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
- Bridge MY06, SedenakJohor, Malaysia · 110 MW anchor tenant; region slowing approvals.
- ZhangjiakouHebei, China
- Green MountainNorway · Hydro-powered.
Timeline
Incidents and developments involving this company