Schaeffler opens 7,500 m² humanoid robotics factory in Taicang

By Sophie Lindqvist

What has Schaeffler started in Taicang?

Schaeffler has opened a factory covering approximately 7,500 m² of floor space in Taicang, China, where six humanoid robots work together at five assembly stations. The building houses an intelligent assembly line developed by Schaeffler, a training and data acquisition center, a platform for digital twins, and a data center. The motion-technology company calls the building a “lighthouse factory.”

The robots assemble key humanoid components, including linear (straight-line motion) and rotary (rotational motion) actuators. They stage and sort parts, join them, and inspect, label, and package them; the line produces key joint modules for the next generation of humanoid robots.

Schaeffler says the factory marks the start of mass production of these components for the company. It also gives customers and partners a place to train and test physical AI systems—robots whose behavior improves from linked training, simulation, and production data—in actual industrial conditions, with the stated aim of driving humanoid adoption in industry.

Dr. Jochen Schröder, chief operating officer of Schaeffler AG, said:

“Schaeffler wants to shape the production of the future and become a leading supplier and preferred technology partner for humanoid robotics. With the lighthouse factory in Taicang, we are creating something that has never before been seen in Schaeffler’s global network: a closed manufacturing loop where humanoids assemble components for humanoids.”

“In the lighthouse factory, we combine decades of manufacturing expertise with physical AI and digital twins. With it, we are expanding our capabilities for key robot components and taking humanoid robotics from theory to industrial practice.”

Schaeffler reports two new Greater China customers in Q3 2026, and global incoming orders for humanoid robotics rose to about €400 million. Those figures represent booked demand; the announcement gives no shipment volume, recognized revenue, factory capacity, or customer delivery schedule.

Dr. Yilin Zhang, regional CEO for Greater China at Schaeffler, said:

“China’s humanoid robotics industry is entering a fast growth period. As a motion technology company, Schaeffler is accelerating the development of China’s humanoid robotics ecosystem by leveraging its local capabilities in research and development, manufacturing and industrial collaboration.”

“The opening of the Taicang lighthouse represents another important move in Schaeffler’s long-term commitment to China and business development in humanoid robotics.”

For developers, the factory shifts evaluation from a robot demonstration toward an operating manufacturing workflow, but it does not yet supply acceptance data.

How does the digital twin measure performance?

A digital twin acts as a virtual replica that mirrors the actual line processes in real time. Its sensors record four process variables:

• Assembly accuracy
• Individual work-step duration
• Energy consumption
• Asset condition, meaning equipment health

Engineers can test and optimize changes in the twin before applying them to the physical line. The method links planned revisions to observed behavior, using the recorded variables to evaluate a process before and during physical execution.

Schaeffler attributes shorter development times and more stable, reliable production to this loop. The announcement states those gains for the actual line but gives no measured time reduction, assembly-accuracy tolerance, cycle-time distribution, energy figure, repeat count, robot model, or independent test result. That remains a vendor claim in the announcement, not a published benchmark.

For buyers, the method supports iteration, but numerical acceptance limits remain the missing evidence.

How are robots trained for factory work?

In the training and data acquisition center, Schaeffler trains humanoid robots for industrial tasks. The task set includes:

• Visual inspection of parts
• Packaging
• Material handling and transport
• Container moving
• Line-side material replenishment

The environments target actual conditions at Schaeffler’s approximately 100 manufacturing plants. Together with the assembly line, they cover testing individual skills through application in an operating manufacturing environment, where robots can test and fine-tune learned skills directly.

The company links data from training, simulation, and production to improve performance continuously. This linked-data loop gives physical AI feedback from both simulated and physical operation, so the robots refine task performance as more records accumulate.

For integrators, the factory supplies a test bed resembling operating plants rather than a staged demonstration alone.

What must developers validate before adoption?

Buyers should ask for numerical tolerances around the variables Schaeffler records: assembly accuracy, work-step duration, energy consumption, and asset condition. They should also require evidence that skills learned in training persist in customer operations, rather than only under conditions modeled on Schaeffler’s plants.

The release names no manufacturing standard, acceptance threshold, customer deployment result, or production capacity. Until those data exist, the factory shows an integrated production model, not a qualified performance benchmark.

The adoption and compliance test is practical: can buyers verify repeatable performance and shorter development times against their own acceptance criteria?