In recent years, many of the images that have popularized Chinese humanoid robots have had a spectacular element. We have seen them run, dance, practice martial arts, and even take part in marathons. Now, the industry wants to show that these machines can do something far more routine and economically relevant: work.
The 2026 World Robot Conference, held in Beijing, has showcased that transition. More than 300 exhibitors from 26 countries presented around 3,000 products, including more than 300 new releases, but much of the attention has shifted toward real-world applications. The goal is no longer simply to demonstrate how far a robot can go, but to identify specific tasks it can perform reliably and profitably enough.
One example presented by UBTECH replicates a complete logistics process. Around 10 humanoid robots work in coordination to lift and stack boxes, sort goods, and carry out various operations along a production line. The demonstration offers a glimpse of one of the markets manufacturers see as closest at hand for this technology: factories and warehouses where repetitive tasks still require human involvement.
From demonstrations to production lines
The shift is beginning to take place in the automotive industry. According to data presented during the conference, nearly 10 major manufacturers, including BYD and SAIC, are testing embodied intelligence systems in their factories for tasks such as material handling and component loading.
Unitree Robotics, one of the Chinese manufacturers to have gained the greatest international prominence, is also testing its robots on automotive assembly lines and at its own facilities. The Hangzhou-based company is simultaneously working to improve the machines’ ability to perform practical tasks in both industrial and domestic settings.
Unitree has also just made one of the sector’s most significant corporate moves by becoming the first humanoid robot manufacturer to list on mainland China’s stock market. Its market debut comes at a time when robotics is beginning to move beyond relying exclusively on future expectations and toward seeking markets capable of generating revenue at scale.
The first applications do not necessarily need to be especially sophisticated. Galaxea, another company at the conference, demonstrated a system through which a customer can request a bottle of water by scanning a QR code. The robot locates the product in a warehouse, retrieves it, and prepares it autonomously in just over a minute. The solution is already operating in Beijing and other cities, and the company believes it could be particularly useful during night shifts, when retaining workers to handle a low volume of orders may be less efficient.
The goal is to identify economically viable tasks
This point is crucial to understanding the next stage of robotics. The question is no longer simply whether a humanoid can lift a box, walk through a factory, or recognize an object. For the technology to scale, it will have to prove that it can perform certain tasks with sufficient accuracy, autonomy, and consistency to justify its cost.
Warehouses and factories provide a particularly suitable environment for beginning that process. They are relatively controlled spaces, where many operations are repeated and can be precisely defined. Goods handling, component transport, and certain logistics operations therefore offer a more immediate testing ground than the widespread use of robots in homes.
China is also exploring applications in environments where replacing human presence may be justified by safety considerations. According to information presented at the conference, systems are already operating in confined or hazardous spaces for emergency response, flood-related operations, and industrial inspections.
An industry worth more than $44 billion
The development of humanoids is part of a considerably larger Chinese robotics industry. According to the country’s Ministry of Industry and Information Technology, sector revenue exceeded 300 billion yuan — around $44.24 billion — in 2025, after recording average annual growth of more than 20% over the previous five years.
The expansion continued in 2026. During the first half of the year, revenue reached 165.5 billion yuan, approximately $24.4 billion, up 24.5% from a year earlier.
China also has an industrial advantage that could prove important in developing this market: an extensive network of manufacturers capable of producing electric motors, batteries, sensors, cameras, electronic components, and other elements needed to build robots. The World Robot Conference reflects precisely that structure, bringing together in one place developers of artificial intelligence algorithms, 3D vision specialists, battery manufacturers, robot producers, and companies interested in using them.
Collaboration among these players is intended to reduce the time needed to develop new models and subsequently transfer technological improvements to industrial production.
Training robots to deal with the real world
One of the major challenges remains getting robots to operate outside prearranged settings. A factory may contain obstacles, objects placed in unexpected locations, moving workers, or situations the system has not encountered before. Addressing that uncertainty requires combining sensors, computer vision, movement, and artificial intelligence models capable of interpreting the environment.
China is building facilities specifically dedicated to training these machines. According to a report by the China Academy of Information and Communications Technology cited by Xinhua, by the end of June the country had more than 70 embodied intelligence training centers, while another 46 were under construction or in planning.
So-called embodied intelligence —embodied intelligence— aims to bring advances in artificial intelligence into the physical world. A language model can interpret a written command and generate a response; the next challenge is to enable a machine to interpret an instruction and physically act on it.
The concept has also gained prominence within China’s industrial policy. Embodied intelligence appeared for the first time in the Government Work Report in 2025, and the 15th Five-Year Plan, covering the 2026-2030 period, includes it among the industries of the future eligible for financial support mechanisms and technological development.
The next leap: robots capable of learning new tasks
The industry’s ambition goes far beyond programming machines to repeat predetermined movements. The goal is to create robots capable of operating in places they do not already know and understanding instructions expressed naturally.
Wang Xingxing, Unitree’s founder, compares it to the leap ChatGPT represented for generative artificial intelligence. In his view, the major breakthrough will come when a robot can enter an unfamiliar environment, receive a command by voice or text, and independently complete most of the requested task.
The entrepreneur believes that moment could come within two
