Teaching steel to fold laundry: Inside Wuxi’s School for Robots

WUXI, CHINA — At a brightly lit workstation in Huishan District, a young university student carefully guides a pair of mechanical arms through the quiet choreography of domestic life. Over and over, she demonstrates motions that a human performs without a second thought: lifting a dinner plate, plucking a piece of fruit from a bowl, creasing a cotton towel into a neat square, and drawing a broom across the floor.

She does not look like a traditional assembly-line worker, and the humanoid machine mirroring her movements is no ordinary piece of industrial hardware. Asked what it is like to spend hours tethered to a robotic apprentice, she smiles and describes an unexpected bond.

"We add everyday life activities: taking plates, grabbing fruit, folding towels and clothes, sweeping the floor, things like that," she explains, adjusting the posture of the machine beside her. "When you are in constant contact with it, it stops feeling like a machine. It feels like a colleague relationship."

Welcome to the Kaipuyun Embodied AI Training Ground, the Wuxi base operated by Wuxi Puying Data Technology Co., Ltd. Stepping onto the sprawling training floor last week felt like walking into a surreal miniature world. Rather than stamping sheet metal or soldering circuit boards, this facility operates as a finishing school for machines, designed to deploy 300 humanoid robots across simulated home, office, and consumer settings.

Across the floor, dozens of everyday scenarios play out simultaneously. In one zone, robotic arms hover over stovetops, learning the delicate wrist rotations required for cooking meals. A few paces away, machines practice pinching the corners of garments to master folding clothes, while others navigate stocked shelves set up like miniature grocery aisles, rehearsing to become supermarket assistants.

Behind every fluid robotic gesture is an immense investment of human patience. The facility employs more than 300 staff members who rotate through two eight-hour shifts each day, turning the building into a relentless engine of motion capture and behavioural demonstration. Under their guidance, it takes two to three months of repetitive training for a robot to master a single suite of tasks: learning how much pressure picks up an apple without bruising it or how to adjust its grip when a towel slips.

For Sun Weidong, General Manager of Wuxi Puying Data Technology Co., Ltd., this labor-intensive spectacle is the necessary foundation for a looming technological shift. While these machines are still learning the basics of human dexterity today, Sun estimates it will take three to five years for service and humanoid robots to achieve widespread commercial usage in homes and retail spaces.

Walking through the Huishan facility makes one thing unmistakably clear: China is playing a fundamentally different game in artificial intelligence. While much of the global tech industry remains fixated on digital screens—training chatbots and cloud-based software models on internet text—China is aggressively industrialising embodied intelligence, bringing AI out of the server room and into the physical world.

In humanoid robotics, the primary bottleneck is no longer mechanical hardware but the severe shortage of high-quality real-world data required to teach machines how to interact with unpredictable environments. Rather than leaving robot training to isolated university labs, China is solving the data bottleneck the same way it scaled electric vehicles and solar manufacturing: through municipal-private industrial clustering and sheer human scale.

Established in April 2026, Wuxi Puying is a joint venture between Kaipuyun—a STAR Market-listed, full-stack AGI service provider bringing two decades of data-governance expertise—and the state-backed Huishan District Science and Innovation Industry Group. Together, they built this base to anchor the local industrial chain around a unified formula: brain, data, and scenarios.

The true product leaving this Wuxi facility is not the hardware standing on the floor, but standardised foundational data—the algorithmic fuel supplied directly to leading Chinese robotics makers like AgiBot and its ecosystem partners. By turning physical human motion into a tradable industrial commodity, Wuxi Puying expects to generate roughly RMB 500 million in cumulative data-sales revenue over five years.

"In the future, once the robots have been trained for these motions, it will be easy to deploy this data," Sun says, looking out over the rows of trainers and machines. "We won't need this level of manual human training forever. We are providing the data that future algorithms will iterate from."

That raises an inevitable economic question: are the young trainers on the floor gradually teaching their own replacements? Around the world, advancements in robotics regularly spark anxiety over job displacement. Yet standing in a room buzzing with hundreds of young workers, Sun views the labor equation differently.

"There is no need to worry about AI replacing human jobs," Sun argues. "Entirely new jobs will be created, and this factory is the living example of that."

Before a machine can ever step into a supermarket aisle or a family kitchen on its own, someone has to teach it how to navigate the messy textures of the physical world. For now, in the workshops of Wuxi, China’s high-tech ambitions still depend on the very human touch of a university student showing her metallic colleague how to fold a shirt.