Fangqi Technology
Category: Robotics / Embodied AI
Beijing-based embodied-intelligence startup building a universal 'robot brain' using a Semantic World Model and Turing Learning Paradigm for cross-embodiment skill transfer. Fangqi Technology was founded in 2026. The company is led by Wang Xinzhou (王心舟). Based in Haidian, Beijing, China. Team size: 11-50. Total funding raised: $1.5M. Latest round: Angel. Key investors include TusStar Venture Capital (启迪之星创投).
- Founded
- 2026
- Headquarters
- Haidian, Beijing, China
- Team size
- 11-50
- Total funding
- $1.5M
Value proposition
Decomposes robot 'brain' capabilities into reusable, embodiment-agnostic skill modules (cloud brain + edge cerebellum + real-time digital twin world model) that can be adapted across wheeled robots, humanoids, and robotic arms without teleoperation dependence.
Products and solutions
Semantic World Model (unifies vision, language, and physical state into 3D semantic space), Turing Learning Paradigm (curriculum/task/practice/reflective learning pipeline), cloud-brain + edge-cerebellum + digital-twin architecture, commercial pilot in '3D cleaning' (bathroom/irregular-object manipulation).
Unique value
Uses humans as the 'greatest common denominator' to unify actions across different robot morphologies, enabling ~2-week algorithmic adaptation to a new robot body (vs. ~6 months typical), and reduces reliance on costly teleoperation data.
Target customer
Robot hardware manufacturers and commercial service operators (e.g., cleaning, hospitality) seeking a general-purpose 'brain' layer rather than building proprietary AI stacks.
Industries served
Robotics, embodied AI, commercial service robotics (cleaning), industrial/hardware manufacturing partners
Technology advantage
Team composed predominantly of Tsinghua University alumni spanning world models, VLA algorithms, and cloud-native systems; ranked 2nd globally in the Stanford BEHAVIOR 2026 household-task challenge; multiple patent applications filed for Turing Learning and Semantic World Model techniques.
How they differentiate
Focuses on a horizontally reusable 'robot brain' decoupled from any single embodiment, versus competitors who typically pair brain and body tightly; explicitly minimizes teleoperation-data dependence in favor of open-source/self-cleaned data and human-action generalization.
Main competitors
Galbot/Yinhe Tongyong (银河通用), AgiBot/Zhiyuan Robotics (智元机器人), Skild AI (US)
Key partnerships
UDI Robotics (优地机器人, HK:03231) and Jiangsu Longhuan (江苏龙寰) — strategic pilot/commercialization partnerships with tens-of-millions-RMB in orders.
Notable customers
UDI Robotics (优地机器人), Jiangsu Longhuan (江苏龙寰) — pilot deployment partners with tens-of-millions-RMB order value
Major milestones
Founded Feb 13 2026 (Haidian, Beijing), completed technical validation of Turing Learning and Semantic World Model in commercial service scenarios, ranked 2nd globally at Stanford BEHAVIOR 2026 household-task challenge, secured angel funding Sept 2026, targeting commercial service prototype launch by end-2026 and scaled deployment in 2027.
Market positioning
Early-stage (angel round) Tsinghua-affiliated 'robot brain' infrastructure player in China's fast-growing embodied-intelligence/humanoid-robotics sector.
Geographic focus
China (Beijing embodied-intelligence cluster), competing globally against US players like Skild AI and Physical Intelligence on cross-embodiment robot-brain architecture.
Patents and IP
Multiple technology patents filed (unspecified count/titles) covering Turing Learning and Semantic World Model methods.
About Wang Xinzhou (王心舟)
PhD, Computer Science, Tsinghua University; deeply involved in Tencent Hunyuan 3D model development and Physical AI algorithm architecture design prior to founding Fangqi.