At the 2026 China Auto Forum, Horizon Robotics VP Cai Ning proposed that smart cars are the most mature "four-wheeled robots," and Horizon will build the "super platform" for the Physical AI era. The company has achieved mass production exceeding 10 million units, building a scaled data flywheel.
Physical AI Moves from Concept to Reality
On July 24, at the 2026 China Auto Forum, Cai Ning, VP of Horizon's intelligent vehicle division, delivered a speech titled "Building Super Platform for the Physical AI Era." He noted that Physical AI is moving from concept to reality—digital agents now possess autonomous planning and execution capabilities, while embodied intelligence is rapidly moving out of the lab.
Looking back at the PC and mobile internet eras, the winners were platform companies combining deep software-hardware integration with massive ecosystems. Horizon believes succeeding in Physical AI requires clearing four hurdles: underlying software-hardware synergy, true openness, cross-scenario optimization, and extending technology from smart vehicles into broader fields.
Smart Cars: The Most Mature "Four-Wheeled Robots"
Cai emphasized that smart cars are the most mature and large-scale "four-wheeled robots" available today—and the toughest testing ground for technology. Horizon has validated its tech through mass production deliveries exceeding 10 million units and scaled data loops involving more than 40 partners.
To keep pace with evolving vehicle architectures, Horizon unveiled its "Starry Series" central computing platform, covering from single-chip to multi-chip combinations for L2 to L4 autonomous driving needs. The platform adopts an open architecture supporting third-party algorithm integration.
From Technology to Ecosystem Leap
Horizon's strategic path is clear: starting with smart cars, gradually expanding to general robotics and broader Physical AI applications. Core capabilities accumulated in the smart vehicle domain—high-compute chips, perception algorithms, planning and control algorithms—have cross-scenario reusability applicable to industrial robots, service robots, drones, and more.
