Scope definition:Consumer-facing + AI as the core interaction + users can "play" for an experience(not AI tools sold as B-end SaaS). Filtered from 1135 entries of 2026 H1 data on 36Kr Pitchhub 19 projects.
In one sentence: 6 tracks are being bet on at the same time
| Track | Number of projects | Leading players | Largest single round |
|---|---|---|---|
| AI short drama / comic drama / interactive drama | 6 | StoReel | 34 million USD |
| AI large model toC applications | 4 | Moonshot AI Kimi / MiniMax / StepFun | 2 billion USD |
| AI video generation toC | 3 | Aishi Technology PixVerse | 300 million USD |
| AI companionship / virtual partner / digital avatar | 3 | Luobo Intelligence / Mirro AI / Kanjian Gainian | Tens of millions of USD level |
| AI games / Vibe Gaming | 2 | Moonton Games / Feituo Xingchi | Over 6 billion USD (M&A) |
| Immersive entertainment / VR-AR | 1 | SoReal | Undisclosed |
Common characteristics:
- AI is not a gimmick, it is the core of the product(remove AI and the entire experience collapses)
- User behavior is not "using a tool", but "playing / watching / chatting / creating"(session duration ≥ several minutes, not open-and-close in 30 seconds)
- Monetization is mainly C-end IAP / subscription / platform revenue share, not dependent on B-end payment
Track 1: AI short drama / comic drama / interactive drama (6)
| Project | Round | Amount | Investors | Introduction |
|---|---|---|---|---|
| StoReel | 2026-03 Seed round | 34 million USD | Play Ventures + T-accelerate + TIRTA + VRF + PVX Partners | North American short drama App + AI drama tool Canvas |
| HoldOrange | 2026-01 Seed round | Undisclosed | Undisclosed | One-stop AI film/short drama creation platform, integrating text/image/video/3D |
| Pole Interactive | 2026-03 Angel round | Tens of millions of RMB | Xinglian Capital / Primavera Capital / Shengshu Technology | Xiamen-based, AI film/animation/comic drama creation tool |
| Chituma Film and Television | 2026-04 Angel round | Undisclosed | Undisclosed | "Watching drama becomes playing drama" AI + VR interactive short drama, multi-branch choices |
| AI comic drama full-link platform | 2026-05 Angel round | 5 million RMB startup funding | Undisclosed | Mass-produced comic drama via AIGC, distributed on Douyin / Video Accounts |
| Kanjian Gainian | 2026-03 Pre-A+ | Over 10 million USD | Ubiquant Investment + Baidu Ventures + Source Code Capital + MiraclePlus | AI subculture content co-creation community |
Innovation point breakdown
StoReel's "dual-wheel model" = the most imaginative one in this batch — most AI short drama companies only build platforms (like a short drama version of Netflix), while StoReel simultaneously built StoReel Canvas(AI drama tool community, creators use AI tools to produce dramas → distribute on its own platform).The production end and distribution end use the same technology stack, creators use AI to make dramas → directly enter the StoReel App to test user feedback → data feeds back into the AI model.
This is the "tool + content + platform" flywheel closed for the first time in China's AI short drama track. The 34 million USD Seed round = the largest single round for a domestic AI short drama company.
Chituma Film and Television's "watching drama becomes playing drama" = format innovation — upgrading from passive viewing to active participation:
- AI plot generation (foundation)
- Multi-branch interactive choices(user chooses → plot follows different branches)
- VR immersive hardware interface(multi-scenario experience open to VR devices)
The five elements of short drama + AI + VR + interaction + choice are superimposed at the same time, a first in China. The problem is that VR hardware penetration is too low; this bets that the "VR hardware + AI short drama" combination can scale in 3–5 years.
AI comic drama full-link platform's "assembly line" = industrialization innovation — not making individual premium works, but building an AIGC comic drama factory:
- In-house script reserve + core team of screenwriters + directors
- AI production → Douyin/Video Accounts → revenue share monetization
- Emphasizes "replicability + clear profit model"
Bets on scaling early in the "comic drama explosion period" to occupy information feed positions. Financing is only 5 million RMB, the playbook is "fast in fast out + mass production".
HoldOrange's "full-factor integration" = tool innovation — integrating world-leading AI models (text/image/video/3D) into one unified platform.Does not self-develop models, builds a "scheduling + integration" tool layer.
Targets the pain points of "too many AI models + expensive access + difficult switching". But this track has too many competitors (Adobe Firefly / Runway / Pika, etc.), requiring localization moats.
Pole Interactive = Alibaba-affiliated — a digital intelligence service brand under the Alibaba ecosystem, with investment from Shengshu Technology (domestic Sora),indicating that big companies are starting dual-track internal + external layout in the AI comic drama track.
Kanjian Gainian's "subculture co-creation" = content format innovation — not a tool, but a community. Focuses on subculture (fan works / ACG / cosplay, etc.) users,generative AI solves the "content production relationship" — enabling users to co-create, resolving copyright issues, and unleashing creativity with AI.
Differentiation point:The "AI content engine" targets not the mass market, but vertical subculture groups(high LTV + strong willingness to pay + strong community stickiness).
Track 2: AI large model toC applications (4)
| Project | Round | Amount | C-end product |
|---|---|---|---|
| Moonshot AI | 2026-05 Series D | 2 billion USD | Kimi (conversation + long context) |
| MiniMax | 2026-01 Hong Kong stock IPO | 4.818 billion HKD | Hailuo / Xingye / Talkie / MiniMax Audio |
| StepFun | 2026-01 B+ | Over 5 billion RMB | Yuewen (StepChat) + Bubble Duck |
| ModelBest | 2026-04 D+ | Hundreds of millions of RMB | MiniCPM / on-device large models |
Innovation point breakdown
Moonshot AI's "long context moat" = Kimi is the only company in the domestic toC AI assistant track that has made 200K → 1M token long context a core C-end selling point:
- Upload an entire book / an entire webpage / a full set of financial reports → Kimi understands it all at once
- Overseas analog: ChatGPT 4 + Claude long context
- C-end PMF validation: In 2024–2025 Kimi was already a DAU leader among domestic toC AI assistants
Series D 2 billion USD (2026-05) + Series C+ 700 million USD (2026-02) = 2.7 billion USD in a single year; investors include Meituan Longzhu / China Mobile / CITIC Industrial Fund (state capital side), Alibaba / Tencent / 5Y Capital / Gaorong Capital / Cathay Capital (early-stage USD).
MiniMax's "product matrix + overseas expansion + IPO" three-piece set = the most complete model/business setup in this batch:
- Hailuo AI(domestic C-end, general assistant)
- Xingye(domestic C-end, AI character social, similar to Character.AI)
- Talkie Talkie (overseas C-end, English version of AI character social, large user base in Southeast Asia + North America + Latin America)
- MiniMax Audio MiniMax Audio (voice generation)
- MiniMax M1 / Hailuo-02 / Speech-02 / Music-01(self-developed multimodal model)
157 million individual users across 200+ countries globally + 50,000 enterprise developers across 90+ countries. 2026-01 Hong Kong stock IPO = the first IPO globally by a Chinese large model company(leading OpenAI's US listing by 2–3 quarters).
Why it matters: It proves that the path "Chinese large models → overseas toC applications → Hong Kong stock IPO" is feasible, and companies will replicate it later.
StepFun's "dual App trial and error" = StepFun built 2 C-end products simultaneously:
- Yuewen (StepChat): efficiency tool (conversation + documents + images) — ChatGPT-like
- Bubble Duck: role-play chat — Character.AI-like
StepFun bets on two segmented user groups,unlike Moonshot AI which only bets on the Kimi line. Over 5 billion RMB in B+ round, investors include Shanghai SDIC / China Life / Pudong / Xuhui / Wuxi / Xiamen ITG (6 state-owned institutions) + Tencent / Qiming / 5Y Capital (USD funds) = the most typical mixed state capital + USD syndicate.
ModelBest's "on-device" differentiation = not competing on large model size, but building "on-device large models" (running locally on phones / PCs / smart hardware) — three differentiation points: privacy + offline + low cost.MiniCPM ModelBest is the most vocal player in domestic on-device large models. D+ stage = pre-IPO preparation.
Track 3: AI video generation toC (3)
| Project | Round | Amount | Remarks |
|---|---|---|---|
| Aishi Technology PixVerse | 2026-03 Series C | 300 million USD | AI video generation, large overseas user base |
| Jijia Shijie | 2026-04 B / 2026-03 Pre-B | Nearly 1.5 billion + nearly 1 billion RMB | World model / video generation |
| Shengshu Technology Vidu | 2026-02 A+ | Over 600 million RMB | Domestic Sora, Tsinghua-affiliated |
Innovation point breakdown
Aishi PixVerse's "overseas tool" = the most "product company" in this track:
- C-end AI video generation tool(users upload images / input prompts → generate short videos)
- Large overseas user base(36Kr reported it is "building an AI Native video product")
- Did not compete on models, competed on product experience and distribution
Series C 300 million USD investors: CDH / China Ruyi / 37 Interactive Entertainment Ventures / Guotai Junan Ventures — a mix of cultural entertainment industry + state capital + financial investors.CDH is exit-oriented, while CDH37 Interactive Entertainment and other content companies have strategic investment intent.
Jijia Shijie's "world model" route = more foundational than PixVerse:
- Not just "text→video", but a "world model" (simulating the physical laws of entire 3D scenes)
- Applications target:driving simulation / robot training / video content
Harder technology, higher valuation, but slower toC landing (mainly B-end payment). In 2026 H1 this single company raised nearly Series B nearly 1.5 billion RMB + Pre-B nearly 1 billion RMB = nearly 2.5 billion RMB, setting a single-round record in the world model track.
Shengshu Technology Vidu = Tsinghua-affiliated,the earliest team in China working on diffusion probabilistic models. Open-sourced UniDiffuser (China's first Transformer multimodal diffusion large model).
Vidu is one of the three domestic Sora players (Shengshu / Aishi / StepFun),Shengshu is the academic school, Aishi is the product school, each with different paths. A+ round over 600 million RMB investors include Baidu / Qiming / Beijing AI Fund — Baidu's strategic investment entry means Baidu's own video generation model has not broken through yet.
Track 4: AI companionship / virtual partner / digital avatar (3)
| Project | Round | Amount | Format |
|---|---|---|---|
| Luobo Intelligence | 2026-04 Pre-A | Undisclosed (led by Fibocom) | AI companion robot (physical hardware) |
| Mirro AI | 2026-03 Seed+ | Tens of millions of USD level | AI semi-acquaintance social (virtual digital avatar) |
| Kanjian Gainian | Already listed in Track 1 | — | — |
Innovation point breakdown
Luobo Intelligence's "physical + desktop + wearable" three forms = the only company that has made AI companionship into physical hardware:
- Team comes from ByteDance / Huawei / DJI / SenseTime / MiniMax / Pop Mart — Pop Mart background is key, meaning the product form integrates "designer toys + AI"
- First product covers both desktop and wearable scenarios(can be placed on a desk, worn on the body)
- Multiple degrees of freedom of movement, strong dynamic expressiveness, interacts with user body movements
Sequoia China + GSR Ventures + 01VC invested across consecutive rounds (Angel / Angel+ / Pre-A, 2025-06 / 2025-09 / 2026-04),Sequoia's tight investment cadence = top VCs are firmly betting on the AI companion hardware line.
Why it matters: All "AI toys" in the Chinese market over the past 5 years have failed (Qiqi robot / Xiaoai Classmate, etc.). Luobo's "designer toy + AI" DNA may be the key difference — it is not a tool, but a collectible.
Mirro AI's "semi-acquaintance social" entry = uses AI to solve the real pain point of "maintaining social networks takes energy":
- Not stranger social (Momo / Tinder)
- Not close-acquaintance social networking (WeChat)
- It is "semi-acquaintance" (colleagues / alumni / group chat members, etc. — relationships exist but contact is infrequent)
- AI helps you maintain / interact / discover potential value
Sequoia China Seed Fund + Jiangyuan Investment, tens of millions of US dollars; from 2025-05 to 2026-03, it took 10 months to go from Seed to Seed+,the fast pace indicates Sequoia is following closely.
Why it is difficult: Momo / Party Island / Soul and others have all tried semi-acquaintance social networking, none succeeded. Whether Mirro AI's AI intervention can break the deadlock remains to be observed.
Track 5: AI Games / Vibe Gaming (2)
| Project | Round | Amount | Notes |
|---|---|---|---|
| Moonton Games | 2026-03 Acquisition | Over 6 billion USD | Chinese mobile game overseas publisher acquired by Savvy Games Group |
| Feituo Xingchi | 2026-04 A | Undisclosed (Guangdian Capital + Richu Capital) | Mobile AI real-time motion capture games going overseas |
Innovation breakdown
Moonton Games acquired for over 6 billion US dollars = not an "AI company", but it was the largest exit across the entire interactive play track(consumer entertainment)across the entire interactive entertainment track in 2026 H1.
Acquired by Saudi Arabia's Savvy Games Group (backed by the Public Investment Fund sovereign fund); *Mobile Legends: Bang Bang (MLBB)* has hundreds of millions of global monthly active users, with Southeast Asia as its core market.
Why it is listed here:
- A mobile game overseas publisher with a single exit exceeding 6 billion US dollars resets the valuation anchor for the entire track
- A landmark event for global value recognition of Chinese overseas entertainment content companies
- It sets exit expectations for AI game / Vibe Gaming companies (over the next 3–5 years)
Feituo Xingchi's "mobile AI motion-sensing" entry point = uses mobile phone cameras + AI real-time motion capture to make motion-sensing games:
- No VR headset required (lowers hardware threshold)
- No dedicated sensors such as Kinect required
- Pure mobile phone + AI algorithms achieve body motion recognition → game interaction
- Going overseas (potentially a natural fit with TikTok algorithms + short video native format)
Why it matters: This is a typical case of AI causing "hardware thresholds to collapse". Over the past 5 years, VR / Kinect failed repeatedly due to high hardware costs. After AI real-time motion capture matured,the mobile phone becomes the game controller.
Track 6: Immersive Entertainment / VR-AR (1)
| Project | Round | Amount | Notes |
|---|---|---|---|
| SoReal | 2026-03 Series D | Undisclosed | VR/AR offline immersive entertainment, founded by Zhang Yimou |
Innovation breakdown
SoReal's "offline scenarios + IP + VR" combination = the only VR company with large financing in 2026 H1:
- Founding team:Qi Xiao / Zhang Yimou / Wang Lei — IP creation + offline scenarios + capital
- Model: does not sell VR devices, operates "VR theme parks" (similar to Meow Wolf)
- Revenue: offline tickets + IP derivatives
Series D (Shijingshan District Modern Innovation Industry Development Fund) = state capital lifeline — the VR industry as a whole has entered a downturn; only the "offline scenarios + IP content" model can still raise funding, while pure hardware / pure software can no longer secure financing.
Why it is listed here: It represents the "non-AI route for consumer immersive entertainment", forming a contrast with AI short dramas / AI games.Over the next 3 years, VR companies must integrate AI content generation to revive(to reduce content production costs).
6 Major Innovation Dimensions (Horizontal Comparison)
| Dimension | Description | Representative projects |
|---|---|---|
| Form innovation | From "watching / using" to "playing / creating" (more user actions) | Chituma turning drama-watching into drama-playing / Kanjian Gainian co-creation / StoReel Canvas |
| Monetization innovation | Overseas IAP / Hong Kong stock IPO / new Douyin revenue-sharing paths | MiniMax first Hong Kong stock IPO / StoReel North America IAP / comic drama Douyin revenue sharing |
| Production innovation | AI industrialized production replaces manual creation | Full-chain comic drama / HoldOrange / Vidu |
| Overseas expansion innovation | Models trained in China, products sold overseas | MiniMax Talkie / PixVerse (Aishi) / StoReel / Feituo |
| Platform innovation | Model companies do not only sell APIs to B2B, but build their own consumer-facing matrix | MiniMax multi-app matrix / StepFun Yuewen + Bubble Duck |
| Hardware innovation | AI causes traditional hardware thresholds to collapse or become feasible again | Feituo AI motion-sensing (mobile phone replaces Kinect) / Luobo AI companion robot (designer toys + AI) |
5 Judgments
1. The AI short drama track will collapse to 2–3 leading players within one year
The 6 projects are severely homogeneous (all claim "AI mass-produced dramas + multi-platform distribution"),StoReel alone raised 34 million USD = more than 5 times the combined total of the other 5 companies. Within one year, it is highly likely that only StoReel + Kanjian Gainian + 1–2 domestic players focused on Douyin/Video Channels will remain, and the rest will disappear.
Basis for judgment: Historically (analogous to 2020 short video + 2021 metaverse),in homogeneous tracks, the top 3 players take 80% of the capital.
2. Consumer-facing large AI models = respective survival paths for three types of companies
- Moonshot AI bets on scaling Kimi alone → exit via US / Hong Kong stock IPO
- MiniMax multi-App matrix + overseas IAP + Hong Kong stock IPO exit
- StepFun + ModelBest + second-tier players bet on "state capital + B-end + on-device"
- Zhipu AI / Baichuan / 01.AI partially shift to B-end / pivot / go silent
The landscape should stabilize over the next 12 months,second-tier players will no longer have a chance to "counterattack and build large models", and can only target niche segments (Mirro AI semi-acquaintance / Luobo AI companion hardware / Feituo AI motion-sensing).
3. "AI content tools" are already over-saturated; the only differentiation left is two paths: overseas expansion + vertical focus
The domestic "AI video generation tools" track plus overseas Pika / Runway / Sora,general-purpose tools have no room left for further competition.
Two differentiated paths that can succeed domestically:
- Compete in overseas markets by going global(PixVerse (Aishi) / StoReel / Talkie model)
- Deep vertical cultivation(Kanjian Gainian subculture / Pole Interactive comic drama / full-chain comic drama)
4. "AI companion hardware" = the next designer toy / IP cycle
If Luobo's "designer toys + AI" route succeeds,it will redefine consumer AI entry points. Originally, AI entry points were mobile apps (Kimi / Hailuo); hardware can "occupy physical space on users' desks / on their persons".Analogy: Switch relative to game apps.
Judgment: 2026 H2 – 2027 should see 3–5 AI companion hardware companies raise large financing rounds; Luobo is the forerunner.
5. The true breakout hit for AI Native interactive entertainment has not yet appeared
Among the 19 projects, none has reached a "nationwide-level" product(over 100 million monthly active users, used by everyone). Reasons:
- Consumer-facing large model products (Kimi / Hailuo) are primarily "tools", not "play experiences"
- AI short dramas / companions / interactive dramas are still in early form exploration
- AI Native games (Vibe Gaming platforms like Astro Kid) require leadership from overseas teams
Judgment:2026 H2 - 2027,An AI Native breakout hit will emerge, most likely from:
- Overseas teams launching Chinese products globally (StoReel / Talkie path)
- Internal incubation at domestic big tech companies (Tencent / ByteDance are possible)
- It will not be any of the existing 19 companies(they are all building version 1.0; a true breakout hit requires a version 2.0 form)
Data query
For the complete dataset, see sources/reference/36kr-pitchhub-2026/ (local reference materials):
cd ~/kb/sources/reference/36kr-pitchhub-2026
# 看具体项目深读
cat events/3174857436945411/summary.md # StoReel
cat events/2189424843266945/summary.md # MiniMax
cat events/2298265962431748/summary.md # 月之暗面
cat events/3092185436618760/summary.md # 阶跃星辰
cat events/3022850194524808/summary.md # 珞博智能
cat events/2288266890147970/summary.md # Mirro AI
# 看每个项目的 36 氪关联深度报道
ls events/<projectId>/articles/
Cross-wiki references
- AI venture capital panorama (including B-end + industrial AI):Deep Dive into China AI / Frontier Technology Financing 2026 H1
- Industry-wide overview:Panoramic Snapshot of China Venture Capital 2026 H1
- IPO + M&A exit side:Deep Dive into China IPO + M&A Exit Side 2026 H1(Detailed analysis of MiniMax Hong Kong IPO / Moonton 6+ billion USD acquisition)
- Crawling methodology:36Kr Pitchhub Crawling Methodology
- Lao Huang's newtype perspective on AI business judgments:newtype Observations on AI Business, Industry and Investment
- Wang Kai's perspective on AI entrepreneurship methodology:Wang Kai's AI-era Entrepreneurship Methodology and Business Judgment Framework
Data limitations
- Small companies not covered by 36Kr (seed rounds below 10 million RMB / undisclosed financing / Chinese teams registered overseas) will be missed
- Some industry labels are misaligned (MiniMax labeled "IoT/hardware", StepFun labeled "culture and entertainment") — manually corrected and reclassified
- Moonton's acquirer ("Savvy Games Group") needs to be confirmed via overseas media (not explicitly stated by 36Kr)
- "Undisclosed" amounts still account for ~45%; this wiki uses "round + investors" to indirectly estimate the magnitude
Compiled on 2026-05-19; an additional round of data capture should be conducted during the Q3 Hong Kong stock IPO peak season to see if leading companies emerge.