BEIJING — "China's robot market stands on the eve of explosive growth. The foundational technologies of each industry — large artificial intelligence models, sensors, data and computing power — are rapidly converging into one."
Wang Tianmiao, a professor at Beijing University of Aeronautics and Astronautics (Beihang University) and honorary director of its robotics institute, offered that assessment of China's robot industry in an interview with The Seoul Economic Daily on the 18th. A leading scholar in China's robotics field, Wang taught Yao Tengzhou, co-founder of Galbot, during Yao's time at Beihang. He also heads the Zhongguancun Zhiyou Institute, which identifies and supports early-stage companies in AI and robotics. The institute has incubated about 100 companies, with Galbot, electric scooter maker Ninebot and surgical robot specialist Tinavi among its flagship portfolio holdings.
"I believe technological convergence in robotics and other advanced industries — the share of technology that comes from the same root — has now passed 60%," Wang said. "Technologies that once developed separately within vertical industries are turning into general-purpose foundational technologies."
He meant that self-driving cars, robots and drones now share much of their technology — AI models, sensors, data, semiconductors and new materials — allowing know-how built up in one industry to be applied in another. Perception and sensor technology developed for autonomous driving becomes a robot's "eyes," while battery, motor and power electronics technology accumulated in the electric vehicle industry becomes its "joints." AI models have also begun serving as the "brain" that lets robots understand and judge their surroundings. As technical barriers between industries fall, Wang said, the time and cost required to develop robots can also fall quickly.
Wang pointed to the enormous automation demand created over the past six to seven years by the new energy vehicle and battery industries as the reason China's robot industry was able to grow so fast. Unlike conventional internal combustion engine cars, the electric vehicle and battery industries brought a flood of new manufacturing processes, including cell and pack assembly, inspection and logistics. As demand grew that could not be met by the standardized solutions offered by established global players, domestic Chinese firms found more room to enter the market for site-specific, non-standard automation.
"New demand emerged, innovative companies built non-standard solutions, and then they were able to scale up by modularizing and standardizing them," Wang said. Labor shortages at manufacturing sites during the COVID-19 pandemic added further momentum to robot adoption, he said.
He also cited China's distinctive "engineer culture" as a competitive strength. While Chinese scientists research new technologies such as AI algorithms, large models and graphics processing units (GPUs) while placing weight on commercialization, engineers show a strong tendency to optimize performance and cost at the same time in line with market demand.
"Chinese engineers do not stop at building a prototype," Wang said. "Along with the desire to stay technologically ahead, fierce market competition pushes them to drive costs down to the extreme while putting performance into the product that is hard to replace." As countless products competed simultaneously in a vast domestic market, the ability to raise performance relative to price became a strength of China's robot industry, he said.
As China's competitive edge in physical AI, he listed a wide range of application sites, a deep hardware supply chain and an open-source ecosystem. "China has many sites where the technology can be applied, which is an advantage for data training and real-world use," Wang said. "Physical AI ultimately requires physical hardware, and in new products that portion can account for more than 60%, so the strength of China's supply chain matters a great deal." He also said the emphasis on open-sourcing AI technology speeds up actual adoption across industries.
On the debate over "neijuan," or cutthroat competition, recently raised in China's humanoid robot sector, he said it is an unavoidable phenomenon in the course of an industry's growth. Noting that large AI models in the United States, home appliances in Korea and China's battery industry all went through fierce competition in the early stages of technological development, Wang called it "a normal phenomenon that appears during the rapid iteration of new technology."
Still, he said not every company will survive. "Companies must provide customers with proper service in terms of performance, safety and environment, while building an irreplaceable moat around their own technology," he said. "In the end, just as a big wave filters out the sand, the competitive companies will remain." Whether it takes five years or 10 for the market to settle on a stable product form and value is "ultimately for the market to decide," he added.
He also said Korea could become an important partner as Chinese robot companies expand overseas, noting that interest in the Korean market was high among the companies he has backed or incubated.
He cited Korea's open market, its semiconductor and new materials industries, and its ability to spread consumer products into global markets as the country's strengths. "Korea has world-class competitiveness in new materials, semiconductors and computing, and it also leads trends in medicine, health care and consumer goods," Wang said. "Because Korea is a window to the global market, more cooperation with Chinese robot companies is possible."
The following is an edited transcript of the interview with Wang.
— Where do you see China's robot industry standing now?
= I think it is on the eve of explosive growth. One important reason is that AI's foundational technologies are increasingly converging into one. Large models, sensors, data, computing power and new materials — technologies that once developed vertically within each industry — are turning into general-purpose foundational technologies that cut across industries. I believe the common base of these technologies has now passed 60%. Also, the core area where AI creates real value for industry and society is ultimately physical AI, and what connects it to the real world is embodied intelligence. In theory we are researching L4 and L5 levels, but the most important thing right now is to apply lower-level technology at scale and move from proof of concept (POC) to product-market fit (PMF).
— What drove the rapid growth of China's robot industry?
= Broadly three things. The first is new demand. As the new energy vehicle and battery industries grew over the past six to seven years, a great deal of automation demand emerged that did not exist in existing manufacturing. Labor shortages during the COVID-19 pandemic also increased demand for robots. When new demand appears, companies first build non-standard solutions tailored to the site, then scale up by modularizing and standardizing them.
Second are scientists and engineers. Scientists learn new AI algorithms, models and GPU architectures while placing weight on turning research results into industry. Engineers combine market demand with products to raise price-performance competitiveness. Third is that the government is building relevant platforms and supporting entrepreneurial talent.
— What has been the core approach of the Zhongguancun Zhiyou Institute in incubating robot companies?
= Turning basic research into actual products and companies is not simply a scientific problem. It requires technological innovation, engineering innovation and business innovation together. The traditional approach in which a single scientist handles both technology and management does not have a high success rate.
So I think a kind of relay approach is needed. First, industry experts identify promising fields and scientists solve the technical problems. Then entrepreneurs with commercial leadership join to grow the business, and as it grows further, it moves into global markets through mergers and acquisitions with large companies. In the case of embodied intelligence, a structure that allows continuous data collection at actual application sites is also important.
— What is the appetite among the Chinese robot companies you have incubated for entering the Korean market?
= The roughly 100 companies we have incubated can be divided into those whose products are already mature, those refining their products in the Chinese market, and those still in a growth stage. When we talk with them, there is a strong desire to cooperate more deeply with Korean companies and the Korean market.
One reason is that the Korean market is relatively open. I think a wide range of robots can be applied there, including logistics, cleaning, coffee making, and care and conversation services. Korea also has strengths in new materials, semiconductors and computing, and it has a unique competitiveness in spreading consumer products into global markets.
— What support are you providing for Chinese robot companies going abroad?
= Support for overseas expansion has only just begun. First, I think associations or institutions that understand both the Korean and Chinese markets well should be used as important connecting channels. It is also necessary to bring companies to Korea to tour sites and take part in exhibitions.
For example, they need to understand firsthand how Korean semiconductors can be applied to Chinese robots and what kind of cooperation is possible with Korean new materials companies. Conversely, when Chinese robots enter the Korean market, they have to consider not only function and cost but also how much they need to adapt to the local environment. In the early stages, cooperating with large companies that have verification systems and capital is also effective. It is important to localize products, functions, design and even maintenance.
— Which industries do you expect to adopt robots in earnest first?
= New technologies and products tend to be accepted first in new environments rather than in existing industries. And the higher the value added in an industry, the easier it is to accept new technology.
I expect the order of robot adoption to begin with commerce and services, logistics, and retail and large supermarkets, then spread to industrial sites and specialized fields, and finally move into the home. In the home, elderly care and housework are representative areas of application.
At industrial sites, rather than putting robots into an entire production line from the start, they should be thoroughly verified at the level of a single product or workstation before being expanded to the whole line. That is because for an entire line you have to calculate how little existing equipment can be changed and how much investment and time are required.
— Some argue that the pace of AI development should be slowed.
= At the national level, it is necessary to set certain standards for the negative effects AI models and algorithms could bring, and to build third-party testing and norms.
But some room also has to be left for technology to develop quickly. For technology to advance rapidly, standardization and scale must be possible, the frequency of iterative development must be high, and openness is also needed. At a macro level, boundaries should be set that do not violate the basic principles of society and civilization, while within companies, technology development should be allowed to accelerate in combination with actual industrial sites.
— What are China's core strengths in physical AI?
= Three things. The first is that there are many sites where it can be applied. Because there are diverse sites, it is also advantageous for data training and real-world use.
Second is the supply chain. Physical AI ultimately requires physical hardware, and in new products hardware can account for more than 60%. China's supply chain has a major strength in this area.
Third is open source. China places relatively high value on opening up AI technology and letting more people use it. This helps speed up the actual adoption of AI in each industry.
— What differentiates China's engineers and talent from those in other countries?
= Chinese education has placed relatively high value on basic disciplines such as mathematics, physics and chemistry, and that became the foundation for training engineers. As the internet industry grew rapidly, young engineers also had many opportunities to encounter software and AI and to be trained in a large market.
There is also a large pool of talent that returns after studying abroad, and many who gain experience at large companies before going independent to start their own. I think this movement and combination of talent created the distinctive character of Chinese engineers.
— What do you see as China's distinctive engineer culture?
= Chinese engineers have a strong desire to stay technologically ahead. At the same time, because the Chinese market itself is so large and competition so fierce, once a product reaches the market it has to cut costs as much as possible while securing performance that sets it apart from other products.
So it does not end with simply building a prototype. There is heavy pressure to secure performance beyond what the times demand while at the same time achieving price competitiveness and differentiation. I see this as an important feature of China's engineer culture.
— How do you view the debate over cutthroat competition, or neijuan, in China's humanoid robot industry?
= When a new technology appears, it is a natural process for rapid iterative development and competition among companies to emerge. Large AI models in the United States, home appliances in Korea and China's battery industry went through a similar process.
Two things matter. The first is providing customers with proper service in terms of performance, safety and environment, and the second is building irreplaceable competitiveness in one's own technology. In the end, only the strongest companies will survive as the market weeds out the rest.
As technology and markets keep developing and customer demands accumulate, an appropriate product form and value will eventually take hold. Whether that process takes five years or 10 will be decided by the market.
