Li Auto to Open Up Its Proprietary Mach M100 AI Chip for External Use
Li Auto CEO Li Xiang announced that the company will gradually open up its proprietary Mach M100 5nm AI chip to external robotics and embodied AI enterprises. This follows strong interest from external robotics companies seeking high-performance computing foundations based on the chip's dataflow architecture. This development represents a major cross-industry shift, allowing automotive-grade high-performance AI hardware to power external robotics and embodied AI applications. By expanding beyond cars, Li Auto is positioning its proprietary silicon as a foundational computing base for broader AI ecosystems. The Mach M100 is a 5nm automotive-grade chip featuring a dataflow architecture with a single-chip computing power of 1280 TOPS and an operating efficiency exceeding 82%. It supports advanced features such as 3D ViT perception and a forward perception range of 300 meters.
## BACKGROUND
Embodied AI refers to intelligent systems with physical bodies, such as mobile robots, that can interact dynamically with the real world, distinguishing them from purely software-based AI like ChatGPT. Dataflow architecture is a specialized hardware design approach that optimizes how data moves through processing units, significantly improving execution efficiency for deep learning workloads.