China's humanoid robotics industry is rapidly transitioning from stage performances to real-world commercial applications, with Chinese companies now accounting for approximately 90 percent of the 13,000 to 16,000 humanoid units sold worldwide. While robots can dance and play piano at trade shows, the harder challenge lies in teaching them to handle unpredictable tasks like grasping a bag of potato chips or folding clothes in actual work environments.

Why Are Humanoid Robots Struggling to Move Beyond Demonstrations?

The gap between stage performance and real-world functionality remains significant. At the Beijing International High-Tech Expo in May, dozens of robots impressed crowds by dancing, playing instruments, and guiding visitors, yet engineers and market representatives acknowledged that widespread adoption in factories, stores, and homes remains years away. The fundamental difference is that performances require only a few predefined actions in controlled spaces, whereas real-world tasks demand understanding complex object relationships and physical properties.

A concrete example illustrates this challenge. At a FamilyMart in Beijing, Galbot G1, a 1.73-meter-tall humanoid robot, successfully fetches drinks from refrigerators and pulls sausages from warming ovens, yet struggles with a seemingly simple task: grasping a flimsy foil bag of potato chips. The robot's grippers, optimized for certain objects, cannot adapt easily to the variable properties of different materials.

"The applications are still limited, but the speed of development is very fast," said Yu Guotao, a field application engineer at Galbot.

Yu Guotao, Field Application Engineer at Galbot

What Technical Breakthroughs Are Helping Robots Learn Complex Tasks?

Researchers are shifting toward physics-based AI models that help robots understand how objects behave, rather than simply copying human movements. This approach allows robots to grasp the relationships between objects and their environment, which is essential for tasks like pouring water or folding clothes.

At Hong Kong University of Science and Technology, PhD student Liang Qiwei and his team are developing AI systems that teach robots to understand object properties such as fabric texture and shape. When a robot encounters a task it cannot complete, human trainers intervene to identify the problem and guide the robot toward the correct solution. Through repeated corrections, the system gradually improves its success rate.

"A performance is limited to a few predefined actions in a given space; the robots just need to go through the motions gracefully without falling over. Real-world manipulation is almost endlessly complex by comparison," explained Liang Qiwei, a robotics and autonomous systems student at Hong Kong University of Science and Technology.

Liang Qiwei, Robotics and Autonomous Systems Student at Hong Kong University of Science and Technology

How Are Companies Deploying Humanoids in Commercial Settings?

• Logistics Operations: Beijing-based startup Robot Era deployed hundreds of humanoid robots across logistics hubs nationwide by late May, reportedly achieving 90 percent of the efficiency of human sorters in parcel handling and sorting tasks.
• Warehouse Processing: Figure AI livestreamed its Figure 03 humanoid processing nearly 250,000 parcels in a 200-hour endurance test in the United States, a video that garnered more than one million YouTube views and sparked discussions about automation's impact on employment.
• Household Automation: In June, Wuhan-based GigaAI announced plans to deploy robots in 100 households for chores and companionship, signaling expansion beyond industrial settings into consumer markets.

These deployments represent a significant milestone for an industry that, until recently, remained largely confined to laboratories and research institutes. China's rapid advancement reflects substantial government backing and investor interest. Beijing aims to reach 10,000 humanoids and establish a 100-billion-yuan industrial cluster by 2027, while Shanghai is actively promoting adoption across logistics, manufacturing, and retail sectors.

Annual sales of humanoid robots are estimated to reach 50,000 units in 2026, according to a Morgan Stanley report cited in the source material. This growth trajectory underscores the industry's momentum, even as technical challenges persist.

What Obstacles Remain Before Humanoids Become Mainstream?

Despite progress, the technology still faces hurdles that prevent everyday adoption. A 32-year-old accountant in Shanghai who ordered a robot-made coffee in April described the experience as novel but stilted, noting that the robot required the cup to be placed in a specific spot before proceeding to the next step, lacking the fluidity of a skilled human barista.

The humanoid form factor itself offers advantages for researchers. Since humanoid robots resemble humans, it is easier to map human actions onto their systems, making training and deployment more efficient than with wheeled or quadrupedal alternatives. However, this advantage only extends so far when robots must navigate the unpredictability of real-world environments.

The industry's evolution from stage performances to marathon appearances, logistics experiments, and early commercial uses demonstrates genuine progress. Yet engineers and researchers agree that the transition from controlled demonstrations to reliable, autonomous operation in unstructured environments will take years of continued development, data collection, and refinement of AI systems that can understand and respond to physical complexity.