Hesai used WAIC 2026 to make a point that sometimes gets lost in the humanoid hype: physical AI still needs extremely good eyes.
According to Hesai Technology, the company showcased its Kosmo spatial intelligence platform and robotic lidar at the World Artificial Intelligence Conference in Shanghai. The announcement sits inside a crowded WAIC news cycle full of humanoids, robot dogs, factory demonstrations, and embodied AI platforms. Even so, the lidar story deserves attention because it gets at one of robotics' most durable bottlenecks.
Robots do not fail only because they lack clever language models. They fail because they misunderstand the geometry around them, lose track of object boundaries, misjudge distance, or cannot build reliable maps in changing environments. Before a robot can plan, grasp, deliver, inspect, or avoid a person, it has to perceive the physical world with enough precision to act safely.
That is where lidar remains important.
Why Lidar Keeps Mattering
The robotics industry has spent years debating cameras versus lidar, especially in autonomous vehicles. The debate can get simplistic. Cameras are cheap, dense, and semantically rich. Lidar gives direct distance measurements, strong geometry, and often better reliability in conditions where visual models struggle.
Most serious robotics systems are moving toward sensor fusion rather than a single winner. A warehouse robot may use cameras for classification, lidar for navigation, IMUs for motion, encoders for wheel state, and ultrasonic or bumper sensors as fallbacks. A humanoid or mobile manipulator can add depth cameras, tactile sensors, force-torque sensing, and wrist cameras. The stack is not elegant, but real-world robotics rarely is.
Hesai's pitch around spatial intelligence is therefore timely. As AI models become more capable, the value of high-quality spatial input rises. A powerful model with poor perception still makes poor decisions. It may know what a pallet is, but if it cannot reliably tell where the pallet edge begins or how close a person is standing, it cannot be trusted in a busy factory.
The WAIC Signal
WAIC 2026 has become a showcase for China's embodied AI ecosystem. The visible layer is full of stage-friendly machines: humanoids posing for crowds, quadrupeds navigating obstacles, compact service robots moving through demonstrations. Underneath that spectacle is a supply chain of sensors, chips, simulation tools, software platforms, actuators, batteries, and integrators.
Hesai sits in one of the most strategic layers of that stack. Lidar is not just an automotive component anymore. It is becoming part of the perception backbone for robots operating in warehouses, industrial yards, ports, mines, hospitals, agriculture, mapping, inspection, and delivery.
That shift matters because robotics adoption is moving from carefully fenced cells toward more open environments. The less structured the environment, the more perception quality matters. A robot arm bolted beside a conveyor can survive with simple vision and fixtures. A mobile robot navigating a shared human workspace needs stronger spatial awareness, faster recovery, and better safety margins.
Kosmo's exact commercial shape will need to prove itself through deployments, not conference language. But the category is right: robotic companies need more than sensors sold as parts. They need spatial intelligence platforms that help convert raw sensor data into maps, scene understanding, localization, and actionable robot context.
The Broader Robotics Context
The timing also fits a larger market split. On one side, frontier AI labs are pushing foundation models into robotics, promising generalization and instruction following. On the other side, hardware suppliers are improving the physical sensing layer those models depend on. The winners will likely combine both.
This is why sensor companies are increasingly talking like AI infrastructure companies. A lidar unit is no longer just a spinning or solid-state device mounted on a robot. It becomes part of a data flywheel: collect geometry, build maps, label edge cases, train perception models, validate safety, and redeploy improvements across fleets.
For investors and operators, that changes how robotics suppliers should be evaluated. The useful question is not only "Who sells the cheapest lidar?" It is "Who helps robots perceive reliably enough to create economic value?" Price matters, especially for fleet deployment, but reliability, calibration, software integration, and support matter just as much.
For readers building small systems, the same lesson scales down. A simple mobile robot kit can teach navigation basics, but adding depth sensing or lidar quickly shows why perception is the heart of autonomy. Starter components like robot lidar sensors and robotics mapping kits are useful ways to understand the tradeoffs before moving into industrial hardware.
Hesai's WAIC announcement is not as flashy as a humanoid doing a backflip. It may be more commercially important. The future of robotics depends on machines that can see the world well enough to work in it.
---
Source: Hesai Technology via Google News, "Hesai Showcases Kosmo Spatial Intelligence Platform and Robotic Lidar at WAIC 2026, Underscoring Its Physical AI Vision", July 20, 2026. Related reading: our guide to robot foundation models and China's robotics strategy.