Guangxiang Technology
Category: Robotics / Embodied AI
Tsinghua-incubated industrial embodied AI startup building wheeled robots for automotive manufacturing, centered on its proprietary GOPS platform and Phi-Bot X1 robot. Guangxiang Technology was founded in 2025. The company is led by Zhang Tao (张涛). Based in Beijing, China. Team size: 101-500. Total funding raised: $45.0M. Latest round: Angel. Key investors include IDG Capital, Oriental Fortune Capital (东方富海), 01VC (零一创投), Lighthouse Capital (光源资本), Eft (埃夫特), Data Capital (达泰资本), L2F Entrepreneurs Fund, Zhuhai Science and Technology Industry Group (珠海科技产业集团), Xingzheng Capital (兴证资本), Songhe Capital (松禾资本), Shunxi Fund (顺禧基金), Mua Kechuang (慕华科创).
- Founded
- 2025
- Headquarters
- Beijing, China
- Team size
- 101-500
- Total funding
- $45.0M
Value proposition
Industrial embodied AI for automotive manufacturing — wheeled robots using reinforcement learning (not imitation learning) to solve complex manipulation tasks in real factory environments, with the GOPS platform enabling rapid deployment (weeks vs months) and self-evolving capabilities.
Products and solutions
GOPS Platform (one-stop embodied intelligence system development platform), Phi-Bot X1 (industrial-grade self-evolving embodied intelligent robot, released Jun 2026), Physical-native foundation model (物理原生基座模型), Phi-RL Matrix (reinforcement learning algorithm suite), Phi-Space (physical data assets), Phi-Arch (general physical intelligence development platform)
Unique value
First embodied AI company to pass stringent POC verification at top automotive OEM assembly lines; wheeled form factor (not humanoid) optimized for industrial efficiency; physics-native foundation model trained via RL rather than VLA/imitation learning; GOPS platform enables modularized development and large-scale delivery.
Target customer
Automotive OEMs (NIO蔚来 and other global automakers); industrial manufacturing plants (3C electronics, heavy industry expansion planned)
Industries served
Automotive manufacturing (primary), warehousing/logistics, industrial manufacturing, engineering operations, 3C electronics
Technology advantage
DSAC series RL algorithms (SOTA in reinforcement learning); first full-neural-network end-to-end autonomous driving system in China (iDrive); Smonet neural network structure for industrial smooth control; RAD neural network training algorithm; Phi-Bot X1 with 27 DOF, 0.05mm repeatability, 1kHz force-controlled arms; GOPS platform modularizing entire embodied AI development pipeline.
How they differentiate
Rejects humanoid robot hype — uses wheeled robots with lower power consumption and more stable positioning; uses reinforcement learning (not imitation learning) to achieve near-100% accuracy vs 90% for imitation learning; "physical-native" foundation model approach distinct from mainstream VLA and video-prediction world models; simulation-first training strategy to overcome real robot data scarcity; GOPS platform for modular embodied AI development.
Main competitors
UBTECH (humanoid robots), Fourier Intelligence (humanoid robots), Unitree Robotics, other industrial embodied AI startups like Agibot (智元机器人), GalaxyBot (银河通用)
Key partnerships
NIO (蔚来) — POC completed in real production lines, Tsinghua University (School of Vehicle and Mobility, School of AI — official equity stake), Multiple international automotive OEMs — strategic partnerships, Lighthouse Capital (光源资本 — incubator and exclusive financial advisor)
Notable customers
NIO (蔚来), multiple leading international automotive OEMs (undisclosed)
Major milestones
2025-04: Company founded, 2025-07: Seed round (~¥数千万) + official Tsinghua equity stake, 2026-03: Angel + Angel+ rounds cumulatively >¥100M led by IDG Capital and Oriental Fortune Capital, 2026-06: Phi-Bot X1 officially launched and demonstrated 21.5hr continuous zero-failure operation at ATC expo, 2026-06: POC completed at NIO production lines, 2026-07: Cumulative angel round of hundreds of millions of yuan announced
Growth metrics
Launched Phi-Bot X1 in Jun 2026; deployed in real automotive production lines; completed 21.5hr continuous zero-failure welding operation; Phi-Bot X1 mobile quality inspection efficiency +51% vs non-cooperative, cycle time 25-45% faster vs manual labor; 5,600+ autonomous mobile robots global delivery experience (from team prior work)
Market positioning
Early-stage Tsinghua-incubated industrial embodied AI leader; chose wheeled over humanoid form factor; targeting the ¥100B ($14B+) automotive assembly line market before expanding horizontally to other manufacturing verticals over 3-5 years.
Geographic focus
China (domestic automotive manufacturing sector); plans to serve international automotive OEMs
Patents and IP
DSAC series reinforcement learning algorithms (SOTA); RAD neural network training algorithm; Smonet general neural network structure for industrial smooth control; GOPS platform IP
About Zhang Tao (张涛)
Ex-Alibaba (Amap) Head of Spatial Perception; PhD from Tsinghua University School of Vehicle and Mobility. Led spatial perception/positioning tech deployed in millions of automotive terminals.
Official website: https://www.qcc.com/firm/acb93218be2f6bfbd0917b78a7cfaa14.html