China is deploying more than 100 humanoid robots in real factory environments to learn manufacturing tasks alongside human supervisors, signaling a major shift from laboratory experiments to practical industrial automation. At Dongfeng Liuzhou Motor in Guangxi province, 11 UBTech Walker S1 robots are undergoing hands-on training to handle materials handling, parts sorting, and bin movement . The robots practice in a dedicated 200-square-meter training zone, guided by instructors who teach them to navigate using visual recognition and perform precise tasks like picking up screws smaller than a fingernail .

Why Are Robots Training Like Human Interns?

Unlike pre-programmed machines that follow fixed scripts, these humanoid robots are learning to adapt to real-world factory conditions. The training approach reflects a fundamental challenge in robotics: machines cannot rely on human instinct or experience. Instead, they depend entirely on algorithms and step-by-step programming . Trainers must understand how robot joints move and how their vision systems interpret their surroundings, then refine control methods to improve task execution. This hands-on learning in actual production environments is far more effective than simulation alone, because factories present unpredictable variables like changing lighting, humidity, and layout shifts .

The city of Liuzhou has established an embodied AI data collection and testing center that replicates production environments across multiple industries. This facility generates vast amounts of training data daily, enabling continuous learning and refinement . Around 120 humanoid robots are undergoing training at the center alongside human supervisors, steadily developing the skills needed for industrial operations .

How Is China Scaling Robot Manufacturing to Meet Demand?

• Production Capacity: China launched a high-capacity humanoid robot production line in Guangdong capable of producing up to 10,000 robots annually, marking a shift from experimental development to mass production .
• Assembly Efficiency: The facility completes a new robot every 30 minutes, representing roughly a 50 percent improvement over traditional assembly methods, with 24 precision assembly stages and 77 inspection checkpoints .
• Flexible Manufacturing: The production system uses automated guided vehicles and digital control systems to switch between different robot models without major reconfiguration, enabling rapid adaptation to diverse industry demands including automotive and home appliance manufacturing .

UBTech, the company behind the Walker S1 robots, partnered with Siemens Digital Industries Software in March to scale humanoid robot manufacturing, targeting annual production of 10,000 units by 2026 . This partnership signals confidence that the technology is moving beyond prototype stage into commercial viability.

What Does Real-World Robot Deployment Look Like Beyond Factories?

While factories represent one frontier, embodied AI is also entering homes. X Square Robot, a Chinese company specializing in embodied artificial intelligence, has launched China's first home-cleaning robot service available to consumers in Shenzhen, partnering with 58.com, one of China's largest household service platforms . This deployment marks a significant milestone in bringing AI out of the laboratory and into everyday life .

The service pairs a professional human cleaner with an AI-powered robot that independently performs structured work such as wiping surfaces, organizing items, and collecting debris . Unlike conventional robots that rely on pre-programmed scripts or remote human control, X Square Robot's systems are designed to understand tasks, break them into steps, and execute them autonomously while adapting in real time to the unpredictability of home environments .

"We're bringing AI into people's homes in a way that's practical and helpful, not theoretical. This is about making everyday tasks easier and showing what intelligent machines can assist with in daily life," said Wang Qian, founder and CEO of X Square Robot.

Wang Qian, Founder and CEO at X Square Robot

The collaboration with 58.com, which operates across over 200 cities and serves tens of millions of households, provides a powerful real-world testing ground at scale . This enables continuous real-world learning, helping X Square Robot refine performance, improve reliability, and accelerate commercialization .

Chris Paxton, a respected researcher and engineer in robotics and embodied AI, highlighted the strategic advantage of this hybrid human-robot approach. He noted that starting with a robot handling some tasks while humans handle others allows companies to scale from 70 percent autonomy to 90 percent to 99 percent naturally over time . This gradual scaling reduces risk while building real-world performance data.

What Makes Embodied AI Different From Other AI Systems?

Most AI today operates through screens and digital interfaces. Embodied AI, by contrast, enables machines to perceive, reason, and act in real-world physical environments . This requires not just intelligent software, but hardware that can navigate, manipulate objects, and respond to sensory feedback. The robots in Liuzhou and Shenzhen represent this transition from digital intelligence to physical capability, transforming how AI interacts with the world beyond data centers and software applications.

The significance of these deployments extends beyond individual companies. As global competition intensifies around the future of AI, AI-driven robotics is emerging as the next frontier . By combining advanced foundation models with large-scale real-world deployment, companies like UBTech and X Square Robot are helping define how AI transitions from theoretical capability to practical utility, transforming not just how factories operate or homes are cleaned, but how people live and interact with machines.