Humanoid robots are helping to assemble parts for future humanoid robots at a new Schaeffler facility in China. It sounds like a science-fiction factory building its own workforce. The actual achievement is more specific and more useful to understand: robots are being tested on repeatable industrial component-assembly tasks alongside conventional machines and employees.
The German manufacturer announced on 20 October 2026 that it had opened what it calls a lighthouse factory in Taicang. The site brings component production, robot training, data collection and digital models of industrial processes together. The ambition is to learn which factory tasks humanoids can perform reliably, not to remove every human from the production line.
Humanoid robots making parts for the next generation
In its official factory announcement, Schaeffler says the new facility is intended to manufacture joint modules and other motion-related components used in humanoid systems. The company describes robots working at dedicated production stations and collecting information that can improve future deployments. Such components matter because a humanoid needs physical joints, actuators and controlled movement before software can make it useful.
The central image is striking: a robot’s mechanical hands working on the parts of another robot. Yet assembling a component on a defined workbench differs from recognising any object in a chaotic room, choosing a repair method or improvising when equipment fails. Industrial demonstrations should be judged by repeatable output, product quality and safety, not simply by how human-shaped the machine looks in photographs.
An important question is what happens when the workpiece arrives differently, a fastener is misaligned or an inspection finds a defect. A factory can use fixtures, predictable lighting, constrained motion and regular maintenance to make a robot’s task easier. Those are advantages of the controlled environment, but they also limit what the demonstration proves about general intelligence.
Why a factory might want a human-shaped machine
Traditional industrial robots can be exceptionally effective at specific repetitive tasks. A humanoid offers a different proposition: potentially using spaces, handling equipment and workflows designed around people. That might reduce the need to rebuild every station from scratch, although an individual deployment still needs safety planning, tooling and careful integration.
Schaeffler has an existing industrial relationship with Agility Robotics, the US developer of the Digit humanoid. Agility has previously described Schaeffler as both a customer and investor, with ambitions to introduce humanoids across its wider factory network. The new Taicang announcement should be understood within that longer programme, not as a sudden universal replacement for industrial automation.
Companies are interested in work that is strenuous, repetitive or awkward to recruit for. A robot that can move items between stations or carry out a defined assembly step might free employees to manage quality, maintenance, exceptions and logistics. Equally, it could change the number and types of jobs available. An announcement describing efficiency benefits is not enough to establish the employment outcome at a particular factory.
The gap between a robot demonstration and useful work
There is a difference between a machine completing a task once and completing it safely through an ordinary shift. Operators need to know whether the robot can recover when it makes a mistake, what happens when a person enters its working area and how maintenance affects throughput. A successful pilot may still prove uneconomic once power, supervision, repairs and integration are included.
LiveAIWire has previously looked at which physical jobs are actually economical for robots. The lesson is that the ability to perform a movement is only one step toward an attractive business case. Industrial buyers also need predictable output, reasonable costs and a clear advantage over existing machinery.
The new site offers Schaeffler a place to test these variables without pretending the open world has been solved. Engineers can change training procedures, inspect production quality and compare what the machines promise with what they do on the floor. This makes the factory an experiment with real industrial constraints, rather than merely an exhibition of futuristic hardware.
There is also a question about what the production data will teach. Repeating the same operation lets engineers see which movements are slow, where grasping fails and how much intervention is needed. A robot that looks accomplished in a short video may still struggle across changing batches. The value of a permanent factory site is that those less photogenic failures can be tracked and improved over time.
For buyers, safety is inseparable from output. A machine operating near people needs predictable stopping behaviour and clearly defined tasks. The eventual business case will depend not just on the price of the robot but on its use across shifts, maintenance arrangements and how easily a team can redeploy it.
Can robots really build more robots indefinitely?
Not on the evidence available here. The factory is described as integrating humanoid assistance into particular stages of component production. Making a joint module does not manufacture a complete autonomous robot, supply its electronic components, write its control software or independently run a whole supply chain. The catchy phrase that humanoids are making humanoids is grounded in one manufacturing step and should not be stretched beyond it.
The fact that people still plan, maintain and supervise the work is not a disappointment. Industrial technology has often improved through gradually automating practical steps. The question for Schaeffler is whether humanoid machines can add flexibility where fixed automation does not fit economically or physically. LiveAIWire’s earlier exploration of AI within industrial manufacturing provides the wider context.
For now, the most meaningful measure will be ordinary factory evidence: reliable cycles, consistent quality, safe working and commercially defensible operating costs. A robot assembling part of another robot is a memorable image. Whether that becomes a common industrial process will depend on less glamorous evidence accumulated after the cameras leave.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.