---
title: "Industry Map | China AIGC Industry Map V2.0 (2023)"
author: "Unique Research"
sourcePublication: "Unique Research (FFCAP English library)"
originalPublishedAt: "2023-04-26T09:34:16+08:00"
canonical: "https://ffcap.cn/en/research/20230426a01rzy00"
source: "https://ffcap.cn/en/research/20230426a01rzy00"
language: "en"
cnSourceUrl: "https://view.inews.qq.com/a/20230426A01RZY00"
---

# Industry Map | China AIGC Industry Map V2.0 (2023)

_Unique Research · Insight / industry map · April 26, 2023 · Full rewrite — not a stub_

After ChatGPT-4, AI markets flipped in ~2 months — March alone saw Google, Microsoft, Baidu, Huawei and Alibaba ship foundation models _and_ applications. Unique Research refreshes February’s China AIGC Map V1.0 into a finer three-layer **V2.0**.

**Breakout moment.** Midjourney’s _Théâtre D’opéra Spatial_ Colorado win → ChatGPT / Midjourney / Stable Diffusion / Firefly / Bard reach consumers; GPT-4 + MJ V5 remake content production. Applied first in games, film, news, literature, music and ads. AIGC is framed as a new production mode after PGC/UGC. a16z’s four stages (PGC → UGC → AI-assisted → full AIGC) place the industry in early AI-assisted. Cited outlook: China generative-AI apps ~¥207B by 2025 (CAGR ~84% 2020–25). Global AIGC financing 2020–22: ¥15.62B across 50 deals; B+ ~44.4%.

**Tech stack.** Drivers = generative algorithms + pretrained models + multimodal. GAN → Transformer / flow / diffusion; pretrained models as qualitative leap. Modalities: text, image, audio, video, strategy, and cross-modal (VL = NLP+CV). Five path examples: text→text (ChatGPT), text→image (Imagen), text→3D (DreamFusion), audio (Murf), image→image (Crypko). Text/image/voice already usable; video/3D still early.

**Three-layer China stack.** (1) _Foundation_ — chips/cloud/data; OpenAI / Stability-class pretrained NLP/CV/VL bases; GPT-4 as VL product. (2) _MaaS mid-layer_ — vertical fine-tunes (NovelAI-style on SD) for industrial deploy. (3) _Apps_ — B- and C-end tools atop both.

**Named China model players.** BAAI / Zhipu AI WuDao 2.0 (1.75T, Tang Jie); Baidu ERNIE Bot (knowledge/retrieval/dialogue enhanced on Paddle); Huawei Pangu (NLP/CV/science/multimodal/speech — AI-for-industry: coal, cement, power, finance, agri); Alibaba Tongyi / ModelScope / Tongyi-M6 (3e8 → 10T → M6-OFA; 200+ ecommerce/design/medical/legal/finance scenes); NetEase Fuxi (since 2017; Chinese datasets → understanding → image-text gen → RLHF; ~200 papers / 400+ patents); Tonghuashun AI Institute; 360 search+algorithm+data+compute bet (Apr 20 earnings).

**Named app players (sample).** Alimama creative center; Juzi.bot WeChat sales/CS robots; Chopstick FusionX / “Jiakuai” creative SaaS (director–shoot–edit–ads–ops); DeepOxygen O3.xyz text→3D video + Director GPT; Domo “D/大画家” AI paint community; WPS Writer bots; Yixie media NLP; AI-Guai AIGC+UGC social; Qiyuan World decision RL; Ruiqi Tech cognitive NLP for public security/gov/finance/aerospace; Baidu Intelligent Cloud TianSuan/TianXiang/TianGong.

_7 CN map images carry the full roster — attach on ingest. Methodology/disclaimer in CN §09._

— END —

---

Editorial English edition published on UniqueCapital: https://ffcap.cn/en/research/20230426a01rzy00
On-site reading page: https://ffcap.cn/en/research/20230426a01rzy00