Chinese AI data startups hit $2.5B valuations in under a year: inside the secretive gold rush
elsewhere别处发生 · wechat · 2026-09-30
A deep-dive into China's hottest and most secretive AI sector: data companies. UniPat reached a $2.5B valuation within a year of founding, with several peers at $300-800M — arguably the fastest valuation growth in primary markets.
The business: building RL environments and expert solution trajectories for model post-training — described as "AI-era tutoring schools." As compute and algorithm gains plateau, data has become the key differentiator, mirroring the rise of ScaleAI and SurgeAI.
Industry dynamics:
- Publishing benchmarks has become a marketing tactic aimed at model labs
- Data types go stale in 3 months; ByteDance is the top spender (reportedly $1B+), followed by Alibaba and Tencent
- Trust runs on relationships: researchers cluster around Tsinghua/PKU/HKUST, and data founders are often their classmates or ex-colleagues
Under-the-table deals: the author suggests some data firms serve as labs' off-balance-sheet proxies — distilling OpenAI/Anthropic models or scraping user data to launder compliance risks. Some now pitch themselves as future model labs ("NeoLab" narratives).
More from Venture
- Bank of England governor warns AI boom could trigger market shocks — nordicinst · 2026-10-01
- Meta claimed Zuck as a "researcher" to snag a $355M tax break on his $4.1B stock compensation — haydenfield · 2026-10-01
- Bootstrapper with 22K users and $1M+ revenue shares his first-100-customers playbook — kylegawley · 2026-10-01
- When Everyone Has the Same AI Tools, Moats Shift Back to Distribution and Trust — sebpaquet · 2026-10-01
- From $20K to $10M+ monthly revenue in 13 months: founder's 500x growth story — adityaag · 2026-10-01
- Memory stocks rally overnight in Asia as AI demand frenzy reignites — firstadopter · 2026-10-01