Summary of Key Points
Recently, the embodied intelligence company Zhi Ping Fang has been rumored to be preparing for an IPO on the Hong Kong stock market, attracting market attention. This company, which was established just three years ago, has become a focal point in the industry due to its "brain-like" technology, rapid financing pace (12 rounds of funding in one year with a valuation of 20 billion RMB), and some implemented use cases (such as industrial PCB loading and unloading, retail coffee shops). However, it also faces many questions: converging technical routes, insufficient ability to commercialize its products, a disconnect between valuation and performance, and the impending reshuffle in the industry. Behind this IPO is both a rational choice by the company to seize the capital opportunity and a reflection of the current state of the embodied intelligence sector, where capital determines who survives, but ultimately, success will depend on the actual ability to deliver real-world solutions.
I. Zhi Ping Fang's "Brain-Like" Technology: Making Robots More Human-like, But Not Exclusive
Zhi Ping Fang's core technology is the NeuroVLA brain-like architecture, which mimics the three-layer division of the human brain:
- Cortex: Responsible for understanding instructions (e.g., "make a cup of coffee") and breaking down tasks (fetching the cup → adding milk → serving the drink).
- Cerebellum: Adjusts the details of movements (e.g., the force used when picking up the cup to avoid breaking it).
- Spinal Cord: Handles unexpected situations (e.g., immediately avoiding obstacles without waiting for higher-level commands).
The advantage of this technology is that it makes robots more "intelligent" and capable of handling complex scenarios (such as working with fragile material trays in factories). However, the technology route is no longer exclusive; global competitors (such as Figure, NVIDIA, Zhi Yuan) are also adopting a "layered architecture," with differences being mainly in details (e.g., data quality, engineering adaptations). Moreover, the smooth operations demonstrated in trials (like the coffee demonstration at WAIC) do not necessarily reflect real-world performance, as changes in lighting and object placement can affect robot behavior, and many demonstrations are pre-adjusted.
II. The 20 Billion RMB Valuation: Stories in the Primary Market, Real Money in the Secondary Market
Zhi Ping Fang is valued at 20 billion RMB, but this figure is perceived differently in the primary and secondary markets:
- Primary Market: Focuses on "future potential." Embodied intelligence is seen as a trillion-dollar market, so investing in a leading company for 20 billion RMB seems worthwhile.
- Secondary Market: Looks at "current performance." Zhi Ping Fang's largest order is from Hikvision for 1,000 units over three years, generating an annual revenue of about 160 million RMB. Supporting such a valuation with this amount of revenue is extremely challenging.
The essence of this discrepancy in valuation is that there are no unified standards in the embodied intelligence industry: how large the market is, whether the technology will be disrupted, and how the market share will be distributed remain unknowns. Therefore, companies rely on their financing pace to maintain interest; Zhi Ping Fang has raised funds seven times in half a year from various investors, including internet giants, traditional manufacturing firms (such as Moutai and CP Group), and even securities companies (CICC and CITIC Construction Investment). However, this model is risky, as each round of funding requires new stories that may force the technology narrative to serve capital rather than truly solve problems.
III. Can Robots Really Get Work Done? Real Performance in Factories and Coffee Shops
Zhi Ping Fang's use cases are mainly in two areas:
1. Industrial Settings: For example, PCB loading and unloading for Hikvision. This task requires precision and stability (fragile material trays, no downtime on the production line). Zhi Ping Fang claims to have achieved repeat business from multiple customers in this scenario, but its delivery volume is still far behind leading players like Zhi Yuan (8,400 units) and Yu Shu (5,900 units).
2. Retail Settings: The "Zhi Mo Fang" coffee shops, where robots make hundreds of cups per day, with plans to expand to 1,000 outlets in three years. The question is whether the robots can cover their costs; for instance, a robot may cost several hundred thousand RMB, so how long will it take to recoup that investment by selling coffee?
In summary, robots can already perform tasks in relatively fixed scenarios, but they are still far from widespread adoption due to the complexity of real-world situations (e.g., customers knocking over cups, sudden changes in factory materials).
IV. Companies Accelerated by the Industry Chain: Is a Reshuffle Imminent?
Zhi Ping Fang's rapid growth is supported by Shenzhen's industrial chain. Within 10 kilometers of Nanshan Liuxian Avenue, companies like iRobot and Podu are located, providing a complete supply chain that allows for quick iteration. However, this also means Zhi Ping Fang is more of a product of the industry chain model rather than an independent innovator.
More importantly, the industry is entering a period of reshuffle: a CICC report indicates that robot valuations have overall decreased in the first half of the year, and capital is beginning to concentrate. Looking at the history of autonomous driving (Mobileye's peak upon listing followed by a sharp drop in Waymo's valuation), the reshuffle in embodied intelligence could last for three years, during which technical systems and revenue models will be restructured. Companies are rushing to go public to secure resources, but this only buys them time; ultimately, they must convert that time into value through real-world performance.
Conclusion
The rumors about Zhi Ping Fang's IPO reflect the "capital frenzy" in the embodied intelligence industry. Going public before a reshuffle is a rational choice, but both investors and the industry need to be aware of the challenges: converging technical routes, insufficient commercialization, and valuation bubbles. The companies that will survive are not those that raise the most funds but those that can solve real-world problems. After all, robots ultimately need to earn money by performing tasks, not just by creating compelling narratives and valuations.