Embodied Data Startup Yuandian Technology Secures Tens of Millions in Funding, Targeting Physical AI Infrastructure

Yuandian Technology, founded by Guo Jiangliang, has completed a financing round of tens of millions of RMB, backed by top domestic embodied intelligence industrial investors, Hengxu Capital, and Kailian Capital. The funds will accelerate the development of MatrixOS, the company's physical AI operating system, and expand its data production network. Guo, a former founding member of Baidu Intelligent Cloud, emphasizes that competition in embodied intelligence is fundamentally about data. The industry faces a severe data gap, with only about 500,000 hours of high-quality physical interaction data available globally by early 2026, far short of the tens of millions of hours needed for universal autonomous capabilities. MatrixOS includes three key components: the ADA data generalization engine, which uses JEPA and latent action representation to improve cross-hardware data reuse, boosting success rates from 65% to 92%; the GDP data quality engine, which leverages information entropy and Bayesian active learning to achieve an 80% data usability conversion rate; and a scalable data production network supporting hundreds of thousands of collection nodes with over 95% automation. The company has already secured a major order from Itshizhi Aviation, helping its A-series robots enter Aptiv's factory, and has partnered with world model companies like Zhiwuwujie. Yuandian has joined Huawei's Ascend computing ecosystem and the Xinhua Net 'Robot+' initiative, and has started international expansion in Southeast Asia. The founding team includes experts from Baidu, Meta AI, and Huawei, with backgrounds in AI infrastructure and industrial applications. The company sees 2026 as the 'year of embodied data' and 2027-2029 as a critical window for humanoid robot commercialization.