Unique Research · Company / industry talk · May 22, 2023 · Full rewrite
Unique Capital partner Hu Xiaojing on Heima Talk (May 18): China AIGC map V1.0 (Feb, 70 firms) → V2.0 (May, 200+). Unique Capital: full-value-chain AIGC boutique bank (FancyTech, Chopstick Tech…); Unique Research’s China AIGC Industry Map widely cited.
Capital pulse
Invest early: 2020–22 China AIGC raised ¥5.41B / 38 deals; pre-B = ¥3.79B (typically ≤25% of volume — here ~70% of amount → early bias). Flagship: Microsoft Xiaoice ¥1B (Nov 2022) as symbolic China AIGC start. Nov 2022–Apr 2023: ¥2.24B = 41.4% of prior two-year total. Digital commerce outlook: generative AI China app scale → ¥207B by 2025 (still conservative). MSFT→OpenAI broke primary-market norms — capital freedom to ship.
Market size paradox: AI image replace designer — unit cost ¥500 → fen: value compresses (market “shrinks”) yet creators explode (market “grows”). Early AI markets defy single logic.
Stack + cases
Upstream foundation models / label / compute (big tech); mid industry models + services (finance, healthcare, retail, aerospace); downstream content produce/design/distribute. Hackathons = mid/downstream creator era.
- Chopstick Tech: enterprise commercial AIGC collab — short-video full pipeline; efficiency 5–30×; AI scripts/storyboards in 1s; mix-cut hundreds of variants; multi-platform publish; sample shop video 8k+ likes; MRR growing 2–3×/mo.
- Unigravity: apparel/jewelry redesign + test — shop owners AI-generate SKUs → private-domain order → C2B reverse custom.
- kua.ai: Amazon sellers — listing/SEO; ~2k-word EN copy; $19; thousands paying.
- Fancy Tech: 1–2 product shots → ad video rivaling 4A, far faster; Taobao/Tmall.
Q&A thesis
Mid/downstream undervalued — in 2–3 years “LLM” becomes infrastructure; scene + innovation where founders should invest. Window varies (engineering/healthcare/new energy slower). New industries (AI assistants, task bots, digital humans <¥10k) as infra matures. Ecosystem > model count (US: many unused models). To-B before to-C (China lacks top domestic base models; CAC higher; to-B monetizes faster). 10× factor growth only when AI hits a true bottleneck production factor (Chopstick 5–30×). Data + scene = moat; AI may be China’s industrial digitalization last mile.
— END —