Not Just Cars: Watch The Robotic Shift At Auto China 2026
Chery's Aimoga Humanoid Robot. (Photo from Chery)
May 5, 20261 hour
Raymond Tribdino
0 Comments
Support CleanTechnica's work through a Substack subscription or on Stripe.
The Beijing International Automotive Exhibition marked a clear turning point in how China’s auto industry defines intelligence. What began over the past decade as a race toward electrification and software-defined vehicles is now expanding into what many exhibitors describe as Physical AI — the extension of artificial intelligence beyond the vehicle and into machines that can act in the real world.
At the China International Exhibition Center, the emphasis this year was not just on smarter cars, but on shared intelligence platforms powering both vehicles and robots. The shift is subtle but significant: the car is no longer the endpoint of innovation, but one of several embodiments of a common AI system.
At XPeng, that idea is expressed through its humanoid robot Iron, presented not as a standalone experiment but as part of a unified technology stack. Roughly human-sized at 178 cm and 70 kg, Iron runs on the same Vision-Language-Action architecture that underpins XPeng’s advanced driver-assistance systems. This matters because it collapses the distinction between mobility AI and robotics AI.
The same model that interprets road conditions and navigates traffic can be adapted to understand indoor environments, manipulate objects, and interact with people. Iron’s perception system borrows heavily from automotive hardware — LiDAR, cameras, and sensor fusion — reconfigured for close-range spatial awareness. Its compute platform, powered by XPENG’s in-house Turing chips, delivers performance in the multi-thousand TOPS range, reflecting how quickly automotive-grade computing is being repurposed for robotics. The company has signaled that early deployments will focus on controlled service environments, where safety, repeatability, and human interaction can be tightly managed.
A more grounded, production-oriented approach is visible at Xiaomi, where humanoid robots are already being tested inside its EV manufacturing operations. Rather than positioning these machines as future-facing concepts, Xiaomi is treating them as incremental upgrades to factory automation. In trials at its Beijing facility, robots were tasked with repetitive fastening work in a die-casting workshop, achieving a success rate just above 90% over several hours of autonomous operation.
The significance lies less in the headline number and more in the environment: these are real production conditions, with moving conveyors, tight cycle times, and minimal tolerance for error. The robots are powered by a large-scale multimodal model trained through reinforcement learning, enabling continuous improvement in balance, coordination, and task execution. Xiaomi’s next phase involves expanding capabilities into more variable tasks such as bin-picking and component placement, areas that have historically challenged traditional industrial robots.
Chery is taking a different route, pushing robotics into the commercial and consumer-facing domain through its division AiMOGA. At Auto China 2026, the company demonstrated a full lineup that included its Mornine humanoid, the Argos quadruped, and specialized service robots designed for public safety roles. What sets Chery apart is its attempt to industrialize and distribute these systems using its existing automotive infrastructure. By pricing its humanoid robot at around 285,800 yuan and integrating sales into dealership networks, the company is effectively testing whether robotics can follow the same scale-driven model as automobiles. Early deployments have focused on guided retail interaction, customer assistance, and controlled public environments, where predictable behavior and branding value are as important as technical capability.
On the factory floor, UBTECH Robotics continues to define the current limits of humanoid deployment. Its Walker S2 platform is already in use within the production lines of Nio, BYD, and Geely, performing logistics and assembly support tasks. The robot’s design emphasizes flexibility, with dozens of degrees of freedom and the ability to operate continuously through an autonomous battery-swapping system. Despite these advances, performance remains below that of human workers, with current efficiency estimates at roughly half of human output. However, the trajectory is improving, with targets approaching 80 percent productivity within the next few years and production volumes scaling into the thousands. The emphasis here is not replacement, but augmentation — robots handling repetitive or physically demanding tasks while humans oversee more complex operations.
Geely extended the Physical AI narrative into mobility itself with a purpose-built robotaxi platform. Unlike retrofitted autonomous vehicles, this prototype is designed from the outset as a fully driverless system, eliminating traditional controls and relying entirely on an integrated perception and planning architecture. Equipped with high-performance computing and next-generation LiDAR, the vehicle operates using a world-model approach, continuously interpreting its surroundings and predicting outcomes in real time. In practical terms, it behaves less like a car and more like a mobile robot designed to transport passengers.
Across all these examples, a consistent pattern emerges. Chinese automakers are no longer treating AI as a feature layered onto vehicles, but as a foundational capability that can be deployed across multiple physical forms. The result is an ecosystem where humanoid robots, factory systems, and autonomous vehicles share the same underlying intelligence. Auto China 2026 did not abandon the electric vehicle narrative — it expanded it. The future on display was not just electric and autonomous, but embodied, mobile, and increasingly human-like in how it interacts with the world.
CleanTechnica's Comment Policy
Share this story!
Схожі новини
