Why are robots an important vehicle for physical AI?

  • Physical AI aims to enable artificial intelligence to understand and act on the physical world. Its world models must predict how physical states change after an intervention.
  • Robots combine perception, decision-making, execution and feedback capabilities. They can generate observable, annotatable interaction data in real environments and enable models to create economic value through physical action.

What changes are driving industry development?

  • The report groups the changes into four areas: tools, models, demand and mass production. Toolchains are gradually being standardized, models are placing greater emphasis on data organization and closed-loop feedback, the entry of leading companies is bringing external demand for paid services, and mass production could increase real-world interaction data.
  • Meng Pengfei of Huafu Securities believes that, as general-purpose model capabilities spread and toolchains become standardized, embodied AI could reach a breakthrough milestone in the next one to two years.

What are the two key areas of focus in the supply chain?

  • The first is data collection. In addition to robots themselves and vertical applications, it includes tactile and force sensing, machine vision, inertial measurement units, data collection gloves, and data collection services and platforms. These areas center on real-world interaction data.
  • The second is the robot component supply chain centered on Tesla, including lead screws, reducers and bearings. Huafu Securities sees both chains as relatively certain beneficiaries of physical AI's development.

What stage have the relevant companies reached?

  • Fosai Technology's robot components remain at the early research and development, small-batch trial production and sample submission stages.
  • Silin Intelligent Drive is actively advancing mass production of robot components, including harmonic reducers, actuator modules and bearings designed specifically for robot reducers.