Silicon Valley vs Wudaokou Hype: Two Myth-Making Movements Surrounding Embodied Intelligence

In today's venture capital circle, Embodied AI is widely recognized as the next most certain trend after large language models. Investors generally believe that the ultimate form of AI will inevitably step out of the screen and into the physical world.

However, if you take a close look at the entrepreneurial ecosystems of this track in China and the United States, you will find a highly fragmented phenomenon: although they are telling the same "world model" story, they follow two completely different sets of business logic.

To put it simply, Silicon Valley is striving to hype "the future of robots", while Wudaokou (referring to the university-centered venture capital circle in China with Tsinghua University and Peking University as the core) is skillfully hyping "the myth of valuation".

The Path to Deification: Dropout Geeks vs. Elite School Prestige

In Silicon Valley, the founding story of embodied AI startups usually follows the narrative of "rebels". The background labels in official announcements are often: several Stanford dropouts, or former core engineers from OpenAI or Tesla. Their standard outfit is a Patagonia fleece jacket and Salomon trail running shoes, exuding a geek vibe of "I just finished fixing bugs in the garage". The words investors use to endorse are restrained yet grand: "It is an honor to support the future of physical artificial intelligence."

In Wudaokou, the narrative logic has turned into an extremely involuted elite selection contest. After investors make an investment, the signal they send to the outside world is usually: "We invested in a team of Tsinghua and Peking University alumni, one from the Yao Class of Tsinghua University and the other from the Turing Class of Peking University. They don't like empty talk and are extremely pure." In the shared office, these young founders wear faded hoodies printed with their elite school logos, place worn-out MacBook laptops on their desks, and there must be a folding bed in the corner — this is regarded as physical proof of "long-termism".

This difference reflects the different demands for "certainty" in the capital markets of the two regions. Venture capital in Silicon Valley bets on the dimensionality reduction strike of technical routes and practical experience; while domestic capital, after going through the traffic involution of the mobile Internet era, is taking "scientific research background from top universities" as the safest belief in the hard technology era. The prestige of elite schools and the label of top academic performers have become the most effective way to reduce the risks of early-stage investment.

Technology Showcase: Long-horizon Tasks vs. Basic Motions

When companies need to demonstrate demos to the outside world, the gap becomes extremely obvious.

Demos in Silicon Valley usually pursue a hardcore feel: for example, letting robots continuously challenge extremely complex consecutive long-horizon tasks (such as completing a full set of kitchen cooking processes), or emphasizing that there are hundreds of thousands of hours of real-world interaction data behind it. In their technical narrative, World Model and VLA (Vision-Language-Action Model) are standard hardcore terms.

In Wudaokou, some demos are full of exaggerated dramatic tension. A founder may only show a semi-finished robot that completes extremely basic translation or grasping motions (such as moving a water cup on the table to another position), which is enough to make the investors present stand up and applaud. If the robot accidentally knocks over a nearby object, the founder will smile and explain: "This is the world model exploring the boundary of physical laws."

This is not only a difference in technical maturity, but also a difference in hype environment. Demos in Silicon Valley need to face the scrutiny of the world's top peers; while in China, due to the extreme scarcity of pure embodied AI targets, capital tends to over-glorify the extremely small progress of early-stage projects. As long as the supervisor nods in terms of intellectual property ownership for the technical accumulation in the laboratory, a simple robotic arm motion is enough to support the grand vision of "the underlying infrastructure of the physical world".

Commercialization: Pay-per-task vs. Verbal Art

Business ultimately has to return to profitability. But there is no standard answer for the commercialization path of embodied AI around the world at present.

The hype in Silicon Valley is relatively pragmatic, with a clear logic: robots enter factories, and customers pay for each task. This is a typical B2B logic.

In Wudaokou, the expression of commercialization has turned into an advanced verbal art. When asked about revenue by investors, the standard answer is: "We have signed a strategic cooperation framework agreement with a large state-owned enterprise, and are promoting multi-scenario implementation." If investors press for specific financial data, the founder will answer in a tone of dimensionality reduction strike: "Embodied AI cannot be measured by the traditional SaaS logic."

Since it cannot be measured by SaaS logic, what can be used to measure it? Under the cover of "long-termism", the unimplementable business closed loop is packaged as a "grand narrative of the era".

Valuation Carnival: Benchmarking Against Giants vs. Resume Addition

What supports this narrative is the dazzling valuation game.

The valuation logic of Silicon Valley is benchmarking: if your technical route is similar to that of Figure AI and Physical Intelligence, you can obtain financing at the same level, and the over-subscription by top institutions speaks for everything.

The valuation in Wudaokou is more like doing addition: the background of the Yao Class of Tsinghua University is valued at 1 billion, and the background of the Turing Class of Peking University is valued at 1 billion. When two top geniuses join forces to generate synergy, the valuation of the Pre-A round is directly set at 3 billion. At the final stage of financing, what FAs (Financial Advisors) say to investors becomes typical hunger marketing: "Head funds are all pushing this project, there is still a little quota available on Saturday, and you will have to wait for the next round after Sunday."

This is a typical capital run driven by FOMO (Fear Of Missing Out). Venture capitalists look at the light in the eyes of these two young people and claim that they "seem to catch a glimpse of the Internet twenty years ago".

What Remains After the Bubble Bursts?

Whether it is the complex long-horizon tasks in Silicon Valley or the simple translation and grasping motions in Wudaokou, they all point to the same problem: there is massive hype and bubble in the current embodied AI industry.

The bubble itself is not scary. In the early stage of any technological revolution, irrational prosperity of capital is needed to promote the construction of infrastructure and the influx of talents. The burst of the dot-com bubble in the past left behind fiber optic cables and underlying protocols; the current AI frenzy has indeed purchased a huge amount of computing power infrastructure.

But what is worth guarding against is that when the smartest minds of the younger generation no longer focus their energy on breaking through the real bottlenecks in the physical world, but skillfully master the rhetoric of catering to capital's FOMO sentiment and creating valuation bubbles, the industry will fall into dangerous stagnation.

That robot that can only do a few basic motions in front of the camera may really represent a certain possibility for the future. But before it can truly walk out of the laboratory and create real cash flow in factories, it is nothing more than an expensive semi-finished product forcibly priced by the capital market. Hype can bring a short-term valuation of tens of billions, but only real engineering delivery and business closed loop can allow the company to survive into the next era.

This article is from the WeChat official account "Tech Business" (ID: InnoBusiness), author: Tech Business, published with authorization from 36Kr.