Physical AI rally lifts Unitree shares to double in intraday trading

On November 10, 2026 (Tuesday), the physical AI theme strengthened. Shiji Hengtong and Jingwei Hongrun rose more than 10%, while Ningbo Huaxiang and瑞芯微 hit their daily upper limits. Unitree Robotics, seen as the leader of the theme, touched CNY 809 intraday, double its intraday low of CNY 404 on October 15. Earlier on November 5, Wang Xingxing, founder and chairman of Unitree, said the company has been pushing a self-evolving loop for robot AI models, which he considers highly valuable; as more robots are deployed and the data they collect keeps growing, he expects the loop to keep optimizing. According to China Securities Journal, a representative of the Beijing Humanoid Robot Innovation Center said that matrix-based collection allows its data collection base to produce 180,000 hours of real-robot data per year, and that the center plans to complete one million hours of collection this year. A Guosen Securities research report notes that one million hours is becoming the main order of magnitude for model training and data collection plans across several physical AI vendors, and that projections for the long-term demand of general physical intelligence reach the tens of millions to the hundreds of millions of hours. Citing data from EqualOcean Intelligence, the report says embodied intelligence needs at least 10 million hours of real-scene interaction data to become genuinely usable, while the industry's total volume of high-quality real physical interaction data is only about 500,000 hours, leaving a large gap; the report attributes this structural scarcity to the physical nature of embodied data, which cannot be obtained directly from the internet and requires costly collection and simulation paths, making data the core bottleneck for industrial development. As of the report, there is no further information on how the market move evolved afterwards.