The "Fever Reduction" Moment for Embodied Intelligence: From Capital Exuberance to Industry Implementation – The Bubble is Bursting
Hello everyone, I'm your financial analyst. Recently, something significant has happened within the robotics community, something that could even be described as an "earthquake."
If you follow tech news, you might have heard of "embodied intelligence" – essentially, intelligent robots that can perform tasks and move around. Over the past two years, this field has been incredibly popular, with numerous startups emerging and their valuations reaching absurd levels. But recently, the tide has turned.
The core question we're going to discuss in this in-depth article is: Is the feast of embodied intelligence coming to an end? Or, in other words, is the bubble finally bursting?
My answer is: The bubble is indeed bursting, but the industry isn't dying; it's going through a painful process of weeding out the false from the true.
Below, I'll break down this news into five key points in plain language to help you understand what's really happening behind this robotics boom.
---
1. The Trigger: Big Names Fighting and Stock Prices Plummeting, Leading to Industry Anxiety
The incident that sparked this wave was quite dramatic.
Background:
Shao Tianlan, the founder of Meikamand, publicly criticized several embodied intelligence companies on his social media, calling them "fabrication-based" players. He said these companies were well-known, had appeared on the Spring Festival Gala, and had high valuations, but their success relied on "related-party transactions" with local governments or state-owned enterprises (for example, setting up data collection centers, selling robots to them, and then buying back their data) to create fake revenue, all in order to rush to go public and raise money.
Interesting Detail:
Shao Tianlan's own company went public on the Hong Kong stock market but fell below its issue price on the first day. Meanwhile, the companies he criticized were busy preparing for their IPOs (Initial Public Offerings). This behavior of criticizing others just after one's own failure reflects the extreme anxiety within the industry – everyone realizes that the current valuations can no longer be sustained.
The Most Indirect Signal: The Stock Price of Yushu Technology
Just a month ago, Yushu Technology, the "first stock of humanoid robots," was listed on the STAR Market. On the first day of trading, its stock price soared by 629%, with a market value reaching 444.9 billion yuan, causing widespread excitement. However, the celebration lasted only one day. By September 2nd, the 11th day of trading, its market value had evaporated by more than 220 billion yuan, almost halving.
Interpretation:
The fact that industry veterans are openly criticizing each other, along with the collapse of leading company stocks, indicates that the market no longer believes in the hype. People are no longer asking "how cool your robot is" but "is the money you're making real? Can it be sustained?"
---
2. Problem One: Revenue Growth Without Profit Growth, an "Emaciated" Revenue Structure
Why did the stock prices fall? Because the financial reports revealed the truth.
Data Says It All:
Looking at Yushu Technology's financial report:
- Slowing Revenue Growth: Revenue in the first half of 2026 was 1.152 billion yuan, still increasing, but the growth rate dropped from 332% last year to 48%.
- Declining Profit: Non-recurring net profit was 244 million yuan, a 19% decrease year-on-year.
- Conclusion: The more they sell, the less profit they make. This suggests that the quality of growth is deteriorating; costs are out of control, or the products they sell are not profitable.
A More Critical Question: To Whom Are They Selling?
According to disclosures, in the first nine months of 2025, Yushu Technology's humanoid robot revenue came from:
- 73.6% from research and education: mainly to universities and research institutes.
- 17.4% from commercial consumption: such as mall displays and performances.
- Only 9% from industry and logistics: these are the areas where large-scale replication and stable cash flow can be generated.
Plain Language Interpretation:
It's like a car company that sells 90% of its cars to university laboratories for experiments and only 10% to taxi companies for transportation.
- Research Customers: They buy one robot for research and rarely repurchase them.
- Industrial Customers: If robots can work reliably in factories, companies will continue to buy them – that's where the real business lies.
Currently, most robots are still at the "laboratory toy" stage and have not yet entered the "factory tool" phase. So, although the revenue seems high, it's based on one-time transactions and is not sustainable.
---
3. Problem Two: Crazy Capital Inflow, Valuations Drastically Out of Touch with Reality
If the revenue structure is an internal issue, the madness of the capital market is the external factor.
Frightening Financing Data:
- In the first half of 2026, there were 288 financings in the domestic embodied intelligence sector, totaling over 46 billion yuan, more than the entire amount of financings in 2025.
- At least 13 unicorn companies with valuations over 10 billion yuan emerged.
- Absurd Comparison: The capital raised by leading companies in one year is nearly five times the industry's annual revenue.
- The Gap: There's so much money that it's hard to allocate, leading to mutual support among companies.
Exorbitant Valuations:
When Yushu Technology went public, its price-earnings ratio (PE) was as high as 750 times. In contrast, the average PE for ordinary manufacturing companies in the A-share market is only 38 times. This means the market expected Yushu's net profit to grow by more than 52% annually over the next five years to justify such a valuation, which is almost impossible in manufacturing.
The Logic Behind the Post-2000s Startup Boom:
The article mentions a group of founders born in the 2000s, such as Huang Yi and Qin Shentao, who received hundreds of millions in financing just a few months after starting their companies.
- On the Surface: It seems like a generation of talented young entrepreneurs is emerging.
- Deeper Look: Capital needs new stories to maintain the hype. When the old players' stories run out, capital pushes new ones to the forefront, using high valuations to attract more funds. This is a typical "passing the torch" game.
Interpretation:
Capital is not empowering the industry; it's distorting it. With so much money available, people focus on inflating valuations and going public rather than on improving the quality of the robots.
---
4. The Truth: Practical Implementation Is Extremely Difficult, and the "Fabrication-Based" Model Is Unsustainable
Why is capital so eager? Because bringing technology to market is incredibly challenging.
Technical Barriers:
- Differences Between Demonstrations and Reality: Robots shown at conferences are often pre-adjusted and may require manual control. In real factory environments, changes in lighting, uneven surfaces, and varying object shapes can cause them to malfunction.
- Wang Xingxing (Yushu's Founder) Admits: Large-scale commercialization could take 2-3 years at the earliest, or 5-10 years at the latest.
Revealing the "Fabrication-Based" Model:
The "data collection center" model criticized by Shao Tianlan is essentially a financial trick:
1. Embodied companies partner with local governments to set up data collection centers.
2. They sell robots to these centers to generate revenue.
3. The centers use the robots to collect data and then sell it back to the companies, creating costs but also generating income.
4. Result: Companies show impressive revenue and profits on paper, which boosts their valuations and allows them to go public.
5. Reality: The money just circulates within the same company, without creating real social value. It's like paying oneself; any break in the funding chain or stricter regulation can lead to a collapse.
Regulatory Action:
The China Securities Regulatory Commission has warned investment banks that humanoid robot companies going public must provide evidence of sustainable revenue, narrowing losses, or genuine technological innovation. In other words, stop with the hype and show real results.
Interpretation:
Previously, it was about PPTs and financial projections; now, it's about actual data. Companies that rely on related-party transactions to maintain their valuations will face increasing difficulties.
---
5. The Way Forward: From "Climbing the Mountain" to "Filling the Holes," Returning to the Industry's Essence
Although the bubble is bursting, the embodied intelligence industry is not dying; it's moving from the virtual to the real.
The Right Approach: Don't Expect Instant Success
Zhang Peng, the founder of Geek Park, made an analogy:
- **Mobile internet startups were about "filling holes": the path was already laid out, and the fastest runner won.
- Embodied intelligence startups are about "climbing a mountain": no one knows where the summit is, and every step is uncertain.
Since the path is unclear, don't wait for a perfect plan; start by testing in real scenarios.
Who Is Really Making Progress? (Positive Examples):
The article lists companies that are actually making progress:
1. Logistics and Warehousing (the First to Succeed):
- Xingdong Jiyuan, Zhiyuan: They operate regularly in the warehouses of SF Express, JD.com, and the postal service.
- Figure 03 (USA): They sort packages live-streamedly, at a rate of 21 pieces per minute.
- Significance: Warehouse environments are relatively standardized, making it easier for robots to adapt. This is the closest we are to large-scale commercial use.
2. Industrial Manufacturing (The Next Front:**
- Qianxun Intelligence: Their robots work three times as fast as humans on the production lines of CATL.
- Youai Zhihe: They have implemented in over 800 industrial scenarios, serving more than 400 customers, with 2025 revenue of 340 million yuan.
- Characteristics: Each revenue stream corresponds to a specific process and a clear ROI (Return on Investment). Customers pay because the robots save them money and improve efficiency, not because of the AI concept.
3. Infrastructure Providers:**
- Wuyi Shijie: They develop simulation platforms for robot training in virtual environments.
- Momenta: They apply autonomous driving technology to robots, using real road data for training.
- Logic: Robots may not be widely used yet, but the tools and data for training them are already profitable.
Core Message:
- Don't Wait for AGI: Before AGI arrives, we need 100 specialized companies with unique capabilities in specific scenarios, not 100 homogenized players claiming to be "universal brains."
- Be Better Than Humans in a Specific Scenario: For example, in pharmacies or on production lines. As long as robots are more reliable, cheaper, and more durable than humans, they have value.
---
Summary: The Feast Hasn't Ended, but the Participants Have Changed
The article concludes with four key points:
1. A 50% Stock Price Drop Is a Good Sign: It indicates that the fundamentals are not collapsing; prices are returning to reality. Previously, it was about "market dreams," now it's about "profitability." This process is harsh and will eliminate companies that were pushed onto the stage before they were ready.
2. China's Biggest Advantage Is Being Wasted: China has the most complete manufacturing supply chain and the largest number of application scenarios. However, resources are being used for inflating valuations and going public rather than for improving products. If this continues, we will fall behind the long-term investments made by the US in foundational technologies.
3. AGI Is a Trap: Don't focus on creating "all-around robots"; instead, focus on specialized robots that solve real problems in specific industries.
4. Cooling Capital, Heating Up the Industry: The future winners won't be those that raise the most money but those that can show a clear ROI in specific scenarios.
In One Sentence:
The bubble of embodied intelligence is bursting, but this is the beginning of the industry's maturation. Companies that rely on PPTs and related-party transactions will fail, while those that solve real problems in factories, warehouses, and production lines will embrace a true breakthrough.
For Ordinary People or Investors:
- Avoid High Valuations: Robots with valuations in the hundreds of millions and no stable revenue are highly risky.
- Focus on Practical Implementation: Look for companies with real orders, recurring sales, and profits in industries like logistics and manufacturing.
- Think Long-Term: Embodied intelligence is a long-term industry that requires patience. Don't dismiss the entire sector due to short-term stock price fluctuations, but also don't blindly believe in capital myths.
The feast continues, but the "wine" is no longer champagne; it's now based on solid industry data.