Hesai CEO: Physical AI Infrastructure Needs Stronger Bottom Layer

At Hesai's technology conference on October 17, 2026, co-founder and CEO Li Yifan said the company is positioning itself as a physical AI infrastructure provider. He described physical AI as having three layers—applications, models, and an underlying layer of hardware, computing, and data—and said the bottom layer is currently the weakest, even though it is the foundation for the sector.

Hesai is focusing on two areas. The first is robot motion engines, or joint actuation. Li said a robot can have more than one hundred actuated joints, which account for over 50% of total BOM cost, with power density as the key metric. The latest generation of Sharpa, an AI company founded by Hesai's three founders, uses Hesai joints and achieves a weight reduction of more than 50% while improving speed and precision. The second is data: Hesai developed the Kosmo spatial camera, which integrates LiDAR, cameras, and inertial navigation to provide high-quality 3D spatial data. It was originally designed for robot training and is now also used in gaming and construction.

Li also said Hesai has been GAAP-profitable for five consecutive quarters, with a gross margin of 39.7%, down 2.5 percentage points year over year. He noted that Hesai's decade of experience in optics, chips, precision manufacturing, and mass production supports its shift, and that the close relationship with Sharpa provides a fast validation loop for its products.