The Major reshuffle in the Embodied Intelligence Industry: When Humanoid Robots Go from “Showy Toys” to “Capital Games”
Hello everyone, I’m your financial analyst. Recently, a significant event within the embodied intelligence (also known as humanoid robots) sector has completely exposed the industry’s underlying issues.
In simple terms, someone has publicly accused peers of manipulating financial records, inflating revenues through clever tactics just to make their financial statements look better before going public. It’s like students cheating on an exam by altering the test papers instead of studying diligently. Now, the regulatory authorities are taking action and tightening the requirements for company listings.
Today, I’ll break down this “embodied intelligence” storm into five key aspects, explaining how the money is made, the pitfalls created, and who will ultimately stand out in the long run.
---
1. How are these “false revenues” generated? – A game of money spinning within the same circle
Many people might wonder: Since the robots haven’t even been sold to factories for use, where do these companies get the money to claim billions in valuations?
Here’s a crucial concept: the data collection centers (also known as “data acquisition centers”).
The normal process is as follows:
Training large AI models requires massive amounts of data. Robot companies produce the robots and sell them to these centers, which then use the robots to perform tasks, record the processes, and convert the data into usable information, which is sold back to the robot companies or AI firms.
- Robot companies: Get orders and revenue.
- Data collection centers: Make money from both the robots and the data.
- The result: Everyone has a business, and data is generated. There’s nothing inherently wrong with this.
But here comes the problematic part:
Some companies resort to “related-party transactions” to quickly boost their revenue. In other words, they sell the robots to affiliated entities.
1. The robot company sells the robots to a related data collection center.
2. The center collects the data and sells it back to the robot company or to the investors behind it.
3. Through this cycle, the revenue figures on the financial statements look more impressive.
The problem is:
This type of revenue is unsustainable. There are no real third-party customers purchasing the robots; it’s just an internal transfer of funds. It’s like claiming your noodle shop is doing well by paying for meals there every day and then recording it as “today’s revenue of 100 yuan.” If you stop doing so or an auditor investigates, the “business” collapses. Shao Tianlan, the founder of Meikamand, pointed out this kind of artificial prosperity created solely for the purpose of going public.
---
2. Who are the actual buyers of these robots? – Besides related parties and spectators, no one is really buying
If we exclude related-party transactions, who are the real customers of these humanoid robot companies?
The answer might be disappointing: **Most robots aren’t being used in factories; they’re sold for “research and education” or “entertainment displays.”
- Research and education market: Universities, middle and high schools for science classes. Robots are bought to study their movements or for students to observe them.
- Entertainment display market: Shopping malls, Spring Festival Gala stages, tech exhibitions. Robots are bought for visual appeal and to attract attention.
Why don’t these markets support the industry’s growth?
In these scenarios, robots don’t need to perform actual tasks; they just need to move smoothly, not fall, and be able to pose. For example, Yuzhu Technology (the first A-share listed humanoid robot company) reported that 73.6% of its revenue came from research and education, with only 17.39% from commercial sales. Xinghaitu’s CEO also acknowledged that currently, only the “developer/research” and “entertainment” markets are mature, and the “real productivity market” hasn’t yet been tapped.
What does this mean?
It means that humanoid robots are still essentially “high-tech toys” rather than a “labor force.” Investor Zhu Xiaohu has long criticized: “The customers cited by founders are just products of their imagination.” If robots can’t perform practical tasks like tightening screws in factories or moving boxes in warehouses, they can’t replace human labor and thus can’t generate significant commercial value. The current revenue is largely sustained by mutual support from shareholders and partners.
---
3. Why the rush to go public? – Without a public listing, the company “dies”
If the products aren’t yet proven viable, why do so many companies strive to list on the Hong Kong or A-share markets?
There’s only one main reason: They burn through money too quickly and need a source of funding.
1. R&D is an endless pit: Humanoid robots involve hardware manufacturing, AI algorithms, and motion control, all of which require long development cycles and substantial investments.
- For comparison, AI model companies like Zhipu and MiniMax spend 1.7 to 3.2 billion yuan on R&D annually.
- Robot companies like Ubtech spend about 500 million yuan, while Yuzhu plans to spend 140 million yuan in 2025.
Although their R&D costs are lower than those of AI giants, they’re still substantial for startups. Additionally, they need to build factories and scale production, increasing costs further.
2. Risk of cash flow disruption: Most humanoid robot companies have no positive cash flow and rely on financing to survive. It’s becoming increasingly difficult to secure funding from the primary market (VC/PE), and valuations are already high (from 10 billion to 20 billion yuan). Without a public listing, subsequent financing would be extremely expensive or impossible.
3. The fear of being eliminated: Some companies openly say, “Without a public listing, we might be out.” A listing not only brings in a large amount of money but also boosts brand awareness and attracts more capital. To secure a spot on the market, companies spend heavily on marketing. For example, four robot companies spent nearly 100 million yuan each on the Spring Festival Gala this year, a amount exceeding Yuzhu’s annual R&D budget. This investment paid off, as the exposure significantly boosted their valuations and listing processes.
So, a public listing is not the goal; it’s a lifeline to stay in business.
---
4. The technology gap: Why can’t robots perform practical tasks yet?
If companies are in such financial distress, why can’t they quickly improve their technology and sell robots to factories for profit?
The reason is that the technology is still not ready:
1. Easy to demonstrate, hard to implement: Robots can perform impressive moves in labs or on stages, but real factory environments are complex with uneven floors, changing lighting, varying objects, and unexpected situations.
- The CTO of Agility Robotics put it well: “Building a robot is one thing; building a robot that can perform useful tasks is another.”
Current robots are more like “large toys” than “reliable employees.”
2. Lack of advanced AI: The core of humanoid robots is embodied intelligence, which requires powerful AI models. However, current AI models are strong in general capabilities but fall short in real-time interactions, precise operations, and long-term stability.
- Insufficient data, immature algorithms, and low hardware precision are major hurdles that prevent robots from working reliably in complex environments.
3. The dilemma of R&D investment: Poor technology means they can’t mass-produce; lack of mass production means no cost reductions; high costs mean poor sales; poor sales mean no funding for further R&D. It’s a vicious cycle. Only a few leading companies (like Ubtech and Yuzhu) can afford to sustain this investment, while most small and medium-sized firms can’t.
---
5. Tightening regulation: IPOs are no longer a place for “storytelling”
Faced with the industry’s chaos and bubbles, regulators have stepped in.
According to media reports, domestic authorities have provided “guidance” to investment banks and firms, raising the standards for approving IPOs for humanoid robot companies. New requirements include:
1. Proving sustainable revenue: Revenues can’t rely solely on related-party transactions or one-time orders.
2. Moving towards profitability: There must be a clear path to profitability, not just continuous spending.
3. Real technological innovation: It’s not enough to just assemble hardware; there must be core algorithms or hardware breakthroughs.
What does this mean?
1. The era of “storytelling” is over: In the past, a compelling PPT and a hot concept were enough for a company to go public. Now, auditors will closely examine related-party transactions and eliminate inflated revenues. If revenues shrink significantly after these adjustments, companies may not meet the listing criteria.
2. The elimination process has begun: Yu Wenchao predicts that only about 3 companies will successfully go public in the first wave of humanoid robot listings on the Hong Kong stock market.
This means most companies will be excluded, and those that rely on related-party transactions to boost revenue will face a critical test.
3. The industry is returning to reality: Regulator intervention is a good thing as it forces companies to focus on their core business rather than capital games.
- Real capabilities: Only those that can solve practical problems with robots will survive.
- Real demand: Only those that find real commercial applications (such as logistics, manufacturing, services) will generate sustainable revenue.
- True innovation: Only those with core technological advantages will gain recognition from investors and regulators.
---
Conclusion: Lessons for the public
1. Don’t be fooled by fancy displays: Don’t cheer just because robots can dance or fight. Consider whether they can work stably in factories for 1,000 hours without errors.
2. Be wary of high valuations: If a robot company’s revenue mainly comes from related parties and lacks clear commercial applications, its valuation may be inflated.
3. Focus on practical applications, not just concepts: Future survivors will be those that enter factories, warehouses, and homes to solve real problems and generate positive cash flow.
4. The industry will undergo a major reshuffle: In the next 1-2 years, many companies will fail, be acquired, or fail to secure funding. This is a necessary part of the industry’s maturation from a bubble to a more mature state.
Embodied intelligence is a trillion-dollar industry with great potential, but the current chaos and bubbles are being addressed by both regulation and market forces. For investors and practitioners, slowing down and focusing on practical results is more important than rushing to go public and creating impressive narratives.