Liqing Intelligence
Category: Robotics / Embodied AI
Physical AI startup building full-stack world model systems for embodied intelligence, enabling robots to perceive, reason, and act across diverse physical environments Liqing Intelligence was founded in 2026. The company is led by Li Yiming (李一鸣). Based in Beijing, China. Team size: 51-100. Total funding raised: $69.0M. Latest round: Seed. Key investors include Shunwei Capital (顺为资本), Sequoia Capital China / HongShan (红杉中国), Hillhouse Capital / GL Ventures (高瓴创投), Fengrui Capital (峰瑞资本), Xinglian Capital (星连资本), Tsinghua Alumni Seed Fund (水木清华校友种子基金), SEE FUND, Zhiyuan Robot (智元机器人), Lingxin Qiaoshou (灵心巧手), Century Golden Resources (世纪金源).
- Founded
- 2026
- Headquarters
- Beijing, China
- Team size
- 51-100
- Total funding
- $69.0M
Value proposition
A data-physics dual-driven Physical AI Infrastructure (data pipeline + world model + physics engine) that enables robots to learn complex physical interaction skills (cutting, screwing, inserting, stirring, pressing, pinching, threading) and deploy across different robot morphologies and scenarios with minimal real-world data
Products and solutions
Physical AI Infra: (1) Self-developed data pipeline scaling data collection to millions of hours, (2) Native world model (self-supervised pre-training + RL post-training with differentiable physics engine), (3) Real-to-Sim-to-Real closed-loop physics engine, (4) Full-stack hardware (self-designed tactile gloves, wheel-arm robot platforms with various manipulators/grippers)
Unique value
Only company globally building a complete, integrated system (data pipeline + world model + physics engine) for Physical AI that achieves cross-embodiment (different robot arms/grippers) and cross-scenario generalization — can train policy models with only 1% of real robot data vs competitors by leveraging physical constraints in the world model
Target customer
B2B industrial clients in manufacturing, logistics, hospitality, commercial kitchens, retail services; eventually expanding to consumer (ToC) market
Industries served
Manufacturing, Logistics, Hospitality, Commercial Kitchens, Retail Services, Medical Assistance
Technology advantage
(1) Full-stack self-developed: data collection hardware -> data pipeline -> world model -> differentiable physics engine, enabling system-wide co-evolution; (2) Native world model architecture (not VLA/JEPA/video generation) with physical-constraint-aware training; (3) Proprietary visual tokenizer outperforming Meta's DINOv3; (4) Founder's deep ties to NVIDIA world-class research ecosystem; (5) Talent moat: 50+ elite researchers from top universities, avg age 22-23
How they differentiate
Unlike Spirit AI (uses 'dirty data' approach, VLA-based) and Galaxea (Fast-WAM architecture focused on inference speed), Liqing differentiates by building a fully integrated native world model system from data collection hardware to physics engine — they reject VLA and video-generation approaches, focusing on physical-constraint-compliant world models that can train policies with only 1% of competitors' real robot data
Main competitors
Spirit AI / 千寻智能 (Qianxun Intelligence), Galaxea / 星海图 (Xinghaitu), Unitree Robotics (宇树科技)
Key partnerships
Zhiyuan Robot (智元机器人) — strategic investor & industry partner, Lingxin Qiaoshou (灵心巧手) — industry investor, Century Golden Resources (世纪金源) — industry investor, NVIDIA (via founder's prior research collaboration), Tsinghua University College of AI (incubation platform)
Notable customers
Partnerships secured with factories, hotels, property management, commercial kitchens, and shopping malls for data collection and pilot deployments (specific customer names not publicly disclosed)
Major milestones
April 2026: Company founded, June 2026: Completed multiple rounds of seed financing totaling several hundred million RMB led by Shunwei Capital, Sequoia China, Hillhouse, 2026 year-end target: Release cross-B2B-scenario world model, 2028 target: Achieve规模化 (large-scale) solution deployment
Growth metrics
Founded April 2026; achieved several hundred million RMB seed round within 2 months of founding; 50+ team members assembled from top universities in under 3 months
Market positioning
Emerging Physical AI / world model startup in China's embodied intelligence ecosystem; positioned as a full-stack 'Physical AI Infra' company rather than a robot hardware or model-only company; aims to be the 'iOS for physical world interaction'
Geographic focus
China (Beijing-based, Tsinghua-incubated); targeting global markets long-term
Patents and IP
4 US patents filed/granted (via founder Li Yiming at NVIDIA); self-developed visual tokenizer, differentiable physics engine, and tactile sensing hardware (gloves)
About Li Yiming (李一鸣)
Ex-Research Scientist NVIDIA; Assistant Professor Tsinghua University College of AI; PhD in AI & Robotics from NYU (2025); NVIDIA Graduate Fellow (2024, 10 globally)
