Summary of Key Points
Embodied intelligence (in simple terms, robots that can perceive and act like humans) has become a highly sought-after target for investors, with leading companies boasting a total valuation of over 600 billion yuan. The pace of financing is so rapid that the C-round of funding often begins before the B-round is even finalized. However, beneath the surface of this excitement lies a stark reality: the actual capabilities of these robots are still akin to those of "2-4-year-old child laborers"—low efficiency and high costs, making it difficult for them to meet the real needs of industries such as factories and hotel services. The industry's focus is shifting from the "physical body" of the robots to their "intelligence" (generalized models) and data collection, but there is no consensus on the technical approach, and there are even signs of a bubble in circular transactions. True commercialization is still a long way off; the key to measuring the value of these robots lies in their repurchase rate, as most of the current shipments are merely impressive on paper.
I. Capital Frenzy: Valuation Inflation and Rapid Financing
The capital enthusiasm for the embodied intelligence sector is skyrocketing. There are at least six companies with a valuation of over 20 billion yuan, and more than a dozen with a valuation of over 10 billion yuan. Including publicly traded companies like Yuzhu (270 billion yuan) and Ubot (42 billion yuan), the total valuation of the leading players exceeds 600 billion yuan. The speed of financing is even more astonishing: one company has completed four rounds of funding since last November, with the C-round starting before the B-round was concluded.
Why is capital so eager to invest? Because some investors see huge potential in this industry, which could combine the high sales volume of smartphones with the high profit margins of passenger vehicles. FOMO (Fear Of Missing Out) has shifted from the AI model space to embodied intelligence. Even state-owned financial institutions are seeking investment opportunities, with investment banks actively competing for deals. For example, one investment bank commented, "They're not good at this; everyone in the industry knows that."
However, capital also has practical demands: companies need to go public as soon as possible. After all, only after going public will they have to produce financial reports to prove their capabilities. No wonder Yuzhu's founder appeared indifferent during the IPO ceremony, and their estimate for the arrival of a "ChatGPT-like moment" has been revised from 1-3 years to 2-10 years.
II. The Harsh Reality: Robots Are Still Ineffective
Despite the hype, the actual capabilities of these robots are far from satisfactory. Yuzhu's newly released humanoid robot can run faster than humans, but the founder has repeatedly stated that it cannot be mass-produced for use in factories due to limited efficiency and generalization skills.
For example, a machine tool factory would need to invest several million yuan in a complete setup (data training, software, and robot body) and retrain the model for each new task, resulting in minimal productivity—far inferior to human workers. The hotel industry, with its goal of having 300,000 rooms by 2030, is also struggling to find robots that can handle luggage transportation and room service tasks.
A founder of a unicorn company with a valuation of over 20 billion yuan said, "The best robots are as capable as 4-year-olds, while the worst are like 2-year-olds; you can't expect child laborers to replace adults." As a result, these companies are only targeting large enterprises willing to invest in continuous product iteration, ignoring the needs of the general market.
III. Shifting Focus: From "Physical Body" to "Intelligence," with Data Becoming a Critical Asset
Previously, investors were focused on the physical design of robots. Now, they are turning their attention to the underlying models and data. The core of embodied intelligence is intelligence; without a powerful "brain" (a generalized model), a robot is just a sophisticated robotic arm.
There are two main technical approaches in the industry: VLA (Vision-Language-Action) and WMA (World-Action Model), but no consensus has been reached, and leading companies are investing in both. Both approaches require large amounts of real-world data for training. This has led to a boom in data collection services. For instance, Baidu Maps has established a team of 20 people dedicated to collecting data for embodied intelligence, and TouchSee Technology has developed tactile data gloves that can capture thousands of sensor points. Some companies, such as Zizheng, offer data collection solutions that do not rely on robots.
However, there are also risks associated with this trend. Some robots are sold to government data centers, which then resell the data, creating a circular transaction similar to NVIDIA's investment in AI companies followed by the sale of chips. Wall Street is wary of such practices.
IV. The Mystery of Commercialization: High Shipments Do Not Equal Real Success
In the first half of this year, over 40,000 humanoid robots were shipped, accounting for 97% of the global total. Does this sound impressive? However, less than 5% of these robots are actually being used stably in factories. For example, the news of a car company purchasing 100 robots does not necessarily indicate commercial success; it only proves that the company is testing the technology. True commercialization is demonstrated by customer repurchase rates—only when customers buy more robots year after year can we say that the technology has met its efficiency and cost targets.
The high shipment numbers are mainly due to research purposes (universities using them for experiments), display purposes (companies using them in exhibitions), or government procurement. The industry is still in its early stages, and no profitable business models have yet emerged.
Therefore, don't be misled by claims of a "trillion-dollar market" or comparisons to the smartphone and automotive industries. The real indicator of success is the repurchase rate—only when customers are willing to buy robots repeatedly can we say that the industry has truly taken off.
In Conclusion
Embodied intelligence is currently a hot topic for investors, but the technology is still in its infancy, and commercialization is slow. There are many bubbles in the market, but the long-term potential is significant. If you want to invest, focus on whether the robots can actually provide practical value, not just on their valuation and shipment numbers.