FOR 12 years, China has been the world’s biggest market for industrial robots, outpacing Germany and Japan to capture more than 50 percent of the market last year. Robots that can perform simple tasks like cleaning floors, waiting on tables, running a race and even dancing are now capturing the public imagination.

But can they also be endowed with the kind of “human brain power” that will take them to a higher plane that generates huge profits across all sectors of the economy? That’s the powerful tension lurking beneath the surface — a tug-of-war between the froth of investment and the faith it will someday pay off. Investors are plowing endless sums into robotics, but the road from entrepreneurial aspiration to bottom-line profits is rocky.

China’s robotics industry didn’t rise from a top-down master plan; it was a revolution from the ground up, powered by a growing supply chain.

A chief scientist at a national humanoid robot innovation center noted one important reason for lift-off. In 2018, a single robotic joint module cost between 50,000 yuan (US$7,000) and 60,000 yuan. Today, the same component costs 500-600 yuan. This advance opened the industrial floodgates. Latest data shows the localization rate for core humanoid robot components in China has now surpassed 70 percent.

Much of China’s “robotic muscle” was developed during another technological revolution: electric vehicles.

“You can think of a car as a type of robot,” explained the co-founder of a tactile sensor firm. “The battery, electronic controls, vision systems, algorithms, models, motors — all are essential elements for developing humanoid robots.”

Witness the recent entry of automakers like Tesla, XPeng and Xiaomi into humanoid robotics. It’s not just a coincidence; it’s natural extension of their core technologies. The auto industry gave robotics a ready-made foundation of high-quality, mass-produced and cost-effective components, turning what was once science fiction into reality.

Giants from the home appliance and consumer electronics sectors are equally committed to the new science. They are both the biggest customers for robots to-date and ambitious players themselves.

Symbolic of this transition was Chinese home appliances giant Midea’s acquisition of KUKA, one of Germany’s “big four” robotics firms, in 2015. At first, the takeover was regarded by skeptics as audacious, but opinion changed after Midea deployed its resources and its huge domestic market to KUKA technologies. It built a new production base in Guangdong Province, and by 2024, the robotics and automation division was generating over 30 billion yuan in revenue.

Midea’s story is not unique. Nanjing-based Estun, an industrial robotics leader, reported that domestic shipments in the first quarter of this year surpassed those of foreign brands.

The more Chinese companies applied software and algorithms to the field of automation, the more products appeared — robots that vacuum floors, cars that can technically drive themselves, unmanned drones used in agriculture, transport and the military. As the science progressed, Chinese companies accumulated technology and data that enabled development of more complex humanoid robots.

A classic business rule is proving true in this robotics gold rush: When everyone is digging for gold, it’s the people selling shovels who get rich first.

Tesla’s Optimus project is the perfect illustration. Founder Elon Musk paints a grand vision of a future with “billions of humanoid robots, but the product itself is still far from mass commercialization. Yet, Chinese companies in Tesla’s supply chain are already profiting from the development boom.

A source close to the company said the strategy of tying oneself to Tesla’s slipstream has been the key to many a successful company. Small companies like Sanhua Holding, a company in heating, ventilation and air conditioning, and auto-parts supplier Top Group have risen into industrial giants on the back of that credo and have now collectively invested over 5 billion yuan in development of humanoid robots.

The race to the top is intense. A source close to a key supplier revealed that its engineers work 16-hour shifts, creating an eight-hour overlap that retains continuity. It’s not just a matter of orders; it’s a matter of receiving technological support from Tesla and being able to say, with pride, that you work with Tesla.

“The plastic for the robot’s faceplate is not high-tech,” said one stock market investor. “But if you can say you’re doing business with Tesla, that’s a whole different story.”

This year has marked a breakthrough. Robotics startups like Ubtech, Zhiyuan and Unitree all secured landmark multimillion-yuan contracts. The chief business officer at Ubtech said orders this year have transition from being primarily for experimental research to actual industrial applications — “a huge step forward,” he called it.

So much for the froth. What about the faith? Beneath the elation, anxiety lurks. Even a 100-million-yuan contract is not truly a “large order” for a robotics firm. Profit margins remain thin. Before the era of mass commercialization arrives, the existential question for all these pioneers is brutally simple: How do we survive?

Getting a robot’s body right is relatively easy, given a mature supply chain. But developing its intelligence — its “soul” — is still in the formative stages. This year’s World Robotics Conference in Beijing highlighted divergent opinions on the path forward. The entire industry is a cacophony of debate and “non-consensus.”

“The robotics industry is paying too much attention to data and not enough to the model itself,” said Unitree founder Wang Xingxing in a comment that reverberated through the conference.

He argued the core problem is that underlying model architectures are neither good enough nor standardized. This means that even with massive amounts of data, the resulting models fail to perform effectively in the real world.

Another camp vehemently disagreed. Its adherents argued that data is the only basis for the engine of intelligence and that even the combined data from all companies doesn’t amount to much yet.

“Without data, a model’s capability will never get from, say, 59 percent to 90 percent,” said the co-founder of another robotics AI startup. “Data determines the floor and the foundation. It’s the key to making that critical leap.”

The source of data is another point of contention. Many companies are banking heavily on synthetic data to drive success, with one industry playing saying those who dismiss simulation data as “toxic” simply don’t know how to use it properly.

The founder of one robotics startup said his firm, after working on the problem for over a decade, came to the conclusion that complex robotic hand movements, for example, cannot evolve through simulation data. Perhaps simple tasks, like pick-up-and-place motions, can be created that way but for more intricate interactions, the most crucial data is that collected from the real world.

The technical roadmap for the robot’s “brain” is even more fragmented. So-called vision-language-action models are currently in vogue. But Unitree’s Wang expressed doubts about them. He called it a “relatively simplistic architecture” but a necessary step forward.

This lack of consensus is a clear sign that the industry is still in its infancy. There is no single truth. Every path could lead to the bright future or to a dead end. The good news is that Chinese robotics platforms like Unitree have already become fundamental tools for global research and development, thanks to their powerful ecosystems and cost advantages.

Behind all the technical debates, one fact remains: companies have to make money to survive. But how to make it has literally become a billion-dollar question.

“We believe the first applications for humanoid robots will be in low-speed, high-frequency, essential-need scenarios, like pharmacy logistics, retail strategy or campus patrols,” said the marketing director of a collaborative robot company. The idea that robots will “join the workforce” before they “enter the home” is a widely held view.

Hong Kong-listed Ubtech did the math. In many advanced factories, 70 percent of tasks are already automated. The remaining 30 percent more non-standard jobs are filled by humans.

The allure of the consumer market is a powerful one that offers endless opportunities for automaton “family” members who can not only perform chores but also provide a sort of emotional companionship. Similar opportunities exist in entertainment and tourism markets.

In the past two years, one of the most profitable applications for humanoid robots has been in rental services for performances at tourist sites, as non-player characters or even taking photos with visitors.

The “Megatron” character at Universal Studios may be the most successful humanoid robot application to date. The hyper-realistic android model has even been extended to running a popular robot restaurant where everything from bartenders to waiters is automated.

Performing robots may not have been the industry’s original vision, but they have proven beneficial in raising public awareness and enthusiasm toward robots, while generating some cash.

The current mix of hype and hope for robotics signals the end of an era and the start of another. With the official entry of tech giants like JD.com, Huawei and Ant Group into the mix, the “Wild West” days of humanoid robotic development are over.

The arrival of these behemoths is a double-edged sword. On one hand, they bring in immense capital and wider application scenarios, like in JD’s logistics or Meituan’s food delivery. On the other, they elevate the competitive landscape, squeezing the window of opportunity for startups

Some analysts predict an industry shake-out that will spell the demise of as many as 80 percent of today’s humanoid robotics companies.

“The humanoid robot industry is on the eve of its ‘ChatGPT moment,’ which could arrive in the next one to two years,” predicts Unitree’s Wang.

A professor from Peking University offers a more measured view: “The value of the humanoid robot market will likely multiply by 10 every three years. In the next decade, it will become a hundred-billion-dollar market. But it won’t happen as fast as people think.”

(The author is founder of WisePromise, a boutique advisory agency specializing in the international expansion of Chinese tech companies in the advanced hardware and energy sectors. He also serves as a geo-economic expert for several think tanks in Beijing.)