Humanoid robots that can dance are no longer news. Getting them to actually work in a factory is a different matter. On October 20, 2026, Schaeffler opened its embodied-intelligence robot digital factory in Taicang, Jiangsu province, where six humanoid robots have been placed across five workstations on a joint-module assembly line, handling tasks such as material handling and parts sorting, according to Chinese automotive outlet D1EV.
The parts they help produce are the very core components needed by humanoid robots themselves. Upstairs, industrial scenarios including visual inspection, parts assembly and material replenishment have been moved into a training area where the robots learn and validate different operational skills. The roughly 7,500-square-metre facility connects joint-module manufacturing, robot training, data collection and digital-twin systems, with the real production line serving not only as a manufacturing site but also as a validation ground for robot applications.
From car parts to robot parts
Schaeffler, a company with decades of manufacturing experience in the automotive industry, is attempting to link parts production, robot training and real-world industrial applications. Its ambitions extend beyond simply selling robot core components. Once robots enter factories, the evaluation criteria change: walking and grasping are just the basics, while manufacturing customers care more about continuous operation stability, working precision and cost - engineering challenges that automotive parts suppliers have long grappled with.
The company is not starting from scratch. Its long-standing expertise in bearings, ball screws, precision reducers, motors, sensors and drive systems overlaps considerably with robot motion systems. In April, Schaeffler formed a strategic partnership with Chinese robot maker Luoshi Robotics to jointly develop integrated joint modules for humanoid robots. David Kehr, Schaeffler's global head of humanoid robotics, believes product design should already consider how to achieve large-scale manufacturing in the future - an approach that mirrors the automotive industry's experience of synchronising product design with manufacturing processes.
For harmonic reducers, a key transmission component in humanoid robots that demands high machining precision, Schaeffler is applying a precision forming process it has used in the automotive sector for about a decade to reduce material consumption and processing steps. According to Kehr, precision forming can shorten some processing steps from minutes to seconds. Schaeffler's validation data also shows the process can cut manufacturing costs by more than 25 percent and reduce material consumption by over 75 percent, with mass production planned to start in Germany in 2027.
Building an industrial data asset
The Taicang plant does more than produce joint modules - it also serves as a validation ground for robot applications on real production lines. "We not only focus on whether robots can move, but also whether they can meet production rhythm, quality, safety and cost requirements, and work stably over the long term," said Sun Haitian, general manager of Schaeffler's embodied-intelligence robot (Taicang) company. A training and data-collection centre on the second floor hosts six types of industrial scenarios selected from Schaeffler's global manufacturing network, with digital-twin systems connecting virtual environments to actual production lines.
One example is a packaging-box folding validation project with Qianxun Intelligent. Robots must not only fold boxes but also adapt to different sizes and operating conditions. To improve model generalisation, the training centre rotates different operators and materials, then cleans, calibrates and filters the collected data. "We don't just produce products, we also produce data. Data can also be an asset," a site presenter noted, adding that such data could eventually be commercialised through data trading platforms. Schaeffler has already included data services in its embodied-intelligence business plan.
However, the industry still faces hurdles in moving from single-scenario validation to large-scale deployment. Feng Tianlin, general manager of JD Industrial's electrical automation division, pointed out that current embodied-intelligence solutions lack standardisation and technical routes are highly fragmented. Chen Runze, R&D vice president at Sudo Technology, said the industry needs effective benchmarks to prevent companies from chasing individual technical metrics while ignoring overall system performance in real tasks.
A key question is whether robot applications validated in Taicang can be transferred to other factories. Sun acknowledged that cross-scenario migration remains a challenge the industry is exploring. He hopes to shorten deployment cycles from the current several months to weeks, ultimately reaching one to two weeks. Schaeffler's special equipment business, with over 70 years of history in production equipment development, automation integration and customised production solutions, is now extending those capabilities to robot testing, system integration and scenario deployment.
On the day the digital factory opened, Schaeffler disclosed that it received new orders from two Chinese customers in the third quarter, bringing its cumulative humanoid robotics order value to 400 million euros. Globally, Schaeffler has partnered with companies including Humanoid and Hexagon Robotics; in China, it works with Leju, Luoshi, Qianxun Intelligent and JD Industrial on component development, industrial applications and supply-chain collaboration. Notably, Schaeffler holds a relatively unique position: it not only supplies robot components but also purchases and uses robots itself.
