PHYMI has closed a seed round of nearly $100 million for a robot that can do more than perform an isolated movement. The startup wants people to give a machine a practical goal—such as collecting an item—and have it work out the sequence of movements and object handling needed to finish, even when its surroundings change.
In a ChinaVenture interview republished by 36Kr, founder Liu Nianqiu used collecting a takeaway order to explain the problem. The item may have moved, the delivery arrangement may change, or a person may add another request. PHYMI is developing machines that can revise their plan through those interruptions. The intended user is someone delegating a physical errand, rather than an operator programming a single repetitive movement at a fixed workstation.
The company's technical plan combines four systems. Its model handles perception and planning; the robot supplies the movement and manipulation; interaction data informs subsequent improvement; and a human-facing layer carries permission, progress and feedback. Investor Yunqi describes those pieces as one integrated system. This is a development roadmap, not evidence that an autonomous assistant can already be purchased and trusted with arbitrary household tasks.
ChinaVenture reports that Liu established PHYMI in March after working as a technology partner at autonomous-driving developer DeepRoute.ai. In the interview, he argues that robotics can learn from the process of collecting driving data, training models and testing new versions. He also cautions that the methods cannot simply be copied into a different physical environment.
IDG Capital led the financing, with Yunqi, DiDi, Fosun RZ Capital, Yaotu Capital and Hesai Technology among the participants, according to the investor announcement and TechNode's report. The disclosed amount is approximate: nearly $100 million, not an exact $100 million balance. Public accounts identify this as a seed round, despite a conflicting Series A label attached to one syndicated page.
The practical test is whether planning survives contact with the real world. An impressive demonstration of one successful task does not establish reliability across many homes, objects and interruptions. PHYMI's stated ambition puts repeatable completion—not just graceful movement—at the center of the product investors are financing.