PHYMI (费莫一科技) raises near-$100 million seed round led by IDG Capital
The AMW Read
A sizable seed round for a new embodied-AI entrant founded by a China autonomous-driving veteran reinforces the segment's talent-migration pattern without resolving a debate or introducing a top-tier lab.
PHYMI (费莫一科技) raises near-$100 million seed round led by IDG Capital
Embodied AI startup PHYMI, founded in March 2026, closed a seed round of nearly $100 million led by IDG Capital, joined by Yunqi Capital, DiDi Chuxing, Fosun RZ Capital, Yaotu Capital, and lidar maker Hesai Technology. Founder and CEO Liu Nianqiu previously served as a technical partner and board member at autonomous-driving company DeepRoute.ai (元戎启行), working on mapless driving systems and the shift toward end-to-end and vision-language-action models; he was earlier a senior engineer at Intel. The funds will go toward core technology R&D, a real-world dynamic data pipeline, product engineering, and team expansion.
PHYMI is building toward long-horizon, cross-scene mobility and manipulation rather than adopting the industry's default "embodied robot" label, via a four-part stack: BRAIN for multimodal perception and planning, BODY for mechanical, drive, sensing and power systems, DATA for real dynamic task data, and LINK for human-robot-environment interaction and multi-robot coordination. Liu argues that robot-specific motors and reducers, and a converged perception-to-action model stack, weren't mature enough during the 2024 embodied-AI startup wave; he says both gaps have since closed, and that autonomous-driving engineers who lived through China's shift from L4 prototypes to mass production are now moving into robotics, a path visible across several recent founding teams.
The round size signals that strategic investors with adjacent exposure — a mobility platform, a lidar supplier — are now backing embodied AI with conviction rather than caution. PHYMI's bet is that generalized task execution across open environments, not a fixed humanoid form factor, is the harder and more defensible problem to build toward; its team spans internet, humanoid-robotics, and autonomous-driving backgrounds consistent with that framing.