On 15 September, Unitree traded at RMB 469.88, down 57 per cent from its 19 August IPO high of RMB 1,100, yet its market value still stood near RMB 190 billion. More than ten embodied-AI firms including Yunxi, AgiBot, Zhongqing, Galbot and Leju have started IPO processes, and valuation has turned into a brawl. Mech-Mind’s Shao Tianlan publicly said some peers manufacture revenue through related-party deals and “want to list after only two or three years”.
Wang Chao, an executive at CAS Star who backed AgiBot early, says embodied AI is on its fourth startup wave: mechanical and control founders first, AI founders second, autonomous-driving founders third, and world-model founders this year. He sees a local bubble but argues the sector as a whole is under-funded, and he prefers “industry veterans” who can actually land products. Zhao Fuheng of Jiuhe Ventures splits the bubble into a value gap (expected versus achievable commercial worth) and a count gap (the market expects 100 champions, reality likely fewer than 10).
Yan Qianhang of FreeS compares today’s embodied AI to autonomous driving in 2012: too early to talk landing, because the tech has not hit its inflection. He expects no change to his early-stage strategy. Feng Tian of BAI argues bubbles and innovation move together; what tests investors is who survives after the bubble pops. She notes Anthropic at roughly 10 to 20 times price-to-sales on an assumed US$100 billion of annual recurring revenue is not unreasonable, and premiums are unavoidable when the winner is unknown.
Some investors stay cautious on lofty primary valuations, but the consensus is that capital is still flooding in. The debate is no longer whether a bubble exists, but which firms will still be standing when it deflates.
Editor’s note: This is an adapted translation of the original Sohu IT report. It has been trimmed and restructured for readability for an international business audience.
Translated and adapted from Sohu IT (https://www.sohu.com/a/1076370835_115565).