China’s humanoid robotics industry has been talking about mass production for years. This week, it delivered the first concrete evidence that the transition from prototype to production line is real.
A new automated manufacturing facility in Foshan, Guangdong province, began operations on March 29, marking what Chinese media and industry observers are calling the country’s first production line capable of manufacturing humanoid robots at a genuine industrial scale. The facility, a joint venture between Leju Robotics and Dongfang Precision Science and Technology, is rated at an annual capacity of more than 10,000 units, and it is designed to produce one completed humanoid robot every 30 minutes.
The technical specifications of the production line are impressive. According to Interesting Engineering, the facility integrates 24 precision assembly stages and 77 inspection checkpoints. The 30-minute cycle time represents approximately a 50% improvement in efficiency compared to traditional assembly methods. Before any robot leaves the line, it undergoes 41 simulated work-condition tests designed to verify reliability in real-world operating environments.
The factory’s design reflects a deliberate strategic choice about how to organize the emerging humanoid robotics industry. Leju Robotics focuses on the design and software side of the business; Dongfang Precision, a company that built its reputation in corrugated packaging equipment before pivoting aggressively into robotic contract manufacturing, handles large-scale production, system integration, and after-sales support. Dongfang’s 2.8% equity stake in Leju aligns the two companies’ incentives, creating a partnership that mirrors the contract manufacturing models that have proven successful in consumer electronics.
The flexible manufacturing design is a notable feature. Using automated guided vehicles and digital control systems, the facility can switch between different robot models without requiring a complete production overhaul. This matters because the humanoid robotics market is still in its early stages, with customer requirements varying significantly across sectors. A factory that can adapt its output mix without major retooling has a significant advantage over one locked into a single product configuration.
The Foshan facility is not the only evidence of China’s manufacturing ambitions in humanoid robotics. Agibot, the Shanghai-based company backed by prominent investors including Zhipu AI’s parent company, recently announced the rollout of its 10,000th humanoid robot. Unitree Robotics, which filed for a $610 million IPO last week, has set a five-year target of producing 75,000 humanoid robots annually. UBTECH Robotics is working toward 5,000 units per year, aiming to bring costs below $20,000 per robot.
The cumulative picture is of an industry that has moved, with remarkable speed, from a phase dominated by research demonstrations and limited pilot deployments to one where industrial-scale production is becoming a reality. The Spring Festival Gala in February, where dozens of humanoid robots performed live on one of China’s most-watched television broadcasts, was a cultural moment that signaled the technology’s arrival in the mainstream. The Foshan factory is the industrial infrastructure that follows.
The remaining challenge, and it is a significant one, is software. Hardware production capacity is no longer the binding constraint on the growth of the humanoid robotics industry. The systems needed to make humanoid robots genuinely useful in unpredictable, real-world environments are still developing. Current deployments are largely confined to structured settings, factory floors with well-defined tasks, research environments, and demonstrations, where the limitations of AI-driven motor control and perception are less exposed.
As Interesting Engineering noted, the pressure is now shifting to AI developers to create systems that can turn these machines into productive workers rather than sophisticated showpieces. The Foshan factory can produce 10,000 humanoid robots a year. The question the industry is now racing to answer is what those robots will actually be able to do.