PCIM深圳论坛 | 氮化镓与具身智能融合应用：探索 GaN 赋能 Physical AI 新机遇

1

2

聚焦氮化镓技术，探析 GaN 器件赋能具身智能系统高效化路径

Physical AI is widely regarded as the next wave since it was first introduced by Nvidia in 2025. Interaction with the physical world requires both sensing and actuation capabilities, as well as edge AI computing. This unified paradigm creates huge synergies between various applications, such as autonomous driving, humanoid robots, drones, eVTOLs, Robo-Vans, quadrupeds, etc. In this presentation, the focus is on the two leading applications: autonomous driving and humanoid robots. Almost all physical AI systems require compact and efficient power supplies or power conversion enabled by GaN, which is also the main topic of this seminar.

This presentation covers the market and technology trends of Automotive OBC, AI server power supplies, and other high-performance applications, where ultra-high efficiency and power density are becoming critical. As GaN technology continues to enable these next-generation systems, it also places higher demands on analog ICs in terms of noise immunity, gate driving capability, and protection performance. The session will discuss key analog IC technologies for GaN-based power systems, design considerations for maximizing GaN performance, and how NOVOSENSE's comprehensive product portfolio delivers high-performance, high-reliability application solutions for ultra-high power density designs.

Kathleen GAO

Head of Smart Power Stage

NOVOSENSE

The leap from industrial robots to humanoid machines requires a fundamental shift in actuator architecture. To achieve fluid movement across dozens of joints, engineers must solve for heat, space, and real-time communication simultaneously. Join TI to explore a holistic approach to humanoid motion control. We will examine a compact 48V/1kW actuator design that integrates GaN-based power stages, AM261x real-time MCUs, and EtherCAT communication. Discover how transitioning from Silicon to GaN allows for smaller DC-link capacitors, increased system efficiency, and the high-precision, low-ripple motor control required for sophisticated humanoid systems.

This presentation analyzes the GaN market set for a 55% CAGR to exceed $5bn by 2031, led by data centers, consumer electronics and EVs, with humanoid robotics as a fast emerging segment. It highlights GaN’s core merits: smaller size, lower power loss and higher efficiency for robotic joints, dexterous hands and main compute power systems. Infineon delivers full CoolGaNTM solutions including integrated half-bridge power stages, 3-phase GaN IPMs, paired with MCUs, magnetic sensors and safety ICs for robot functional blocks. The firm owns full in-house GaN 8-inch wafers, planning 12-inch GaN-on-Silicon ramp in 2027 to support mass production for humanoid power system demands.