Is the valuation of embodied intelligence all a bubble?

As the "Billion - Valuation Club" in the embodied intelligence track continues to expand, established pioneers like Zhipu and Unitree have firmly secured their positions, while more new players are eager to break into the core of the track. The focus of this industrial competition has shifted from "how to enter the billion - valuation threshold" to "how to take root and grow after reaching a billion - dollar valuation."

Embodied intelligence is the real wealth - creator.

While other industries may take years to stand out from fierce competition and even need a record of consecutive profits to reach the billion - valuation threshold, embodied intelligence simply states that a billion - dollar valuation is just the entry requirement.

Zhipu Robotics is valued at 15 billion yuan; after its latest round of financing, Unitree is valued at 12 billion yuan, and it is reported that its valuation could reach 50 billion yuan after going public.

Neither Unitree nor Zhipu has yet presented a solid performance record. With their young and somewhat vague business results, it's hard to believe that these are billion - valuation enterprises. However, in the context of the embodied intelligence concept, all the logic makes sense.

This wealth - creation ability has inevitably sparked controversy, and the debate about the "bubble" has never ceased. Indeed, with the business model still in the exploratory stage, the rapid short - term expansion makes it difficult to distinguish whether the internal growth is real capability or just a bubble. Most people firmly believe that the current prosperity is just a bubble feast created by profit - seeking capital.

But is that really the case?

Looking globally, after completing its Series C financing, Figure is valued at $39 billion, equivalent to nearly 23 Unitree Technologies. In comparison, domestic embodied intelligence companies are not only free from the so - called "bubble" but are even a bit conservative. No wonder many domestic industry insiders have joked that we are far from the bubble stage when discussing the bubble controversy.

Li Qing, the founder of Zhuoyide, also once said that from a valuation perspective, with humanoid robot enterprises valued only between 10 billion and 20 billion yuan, the industry is far from the "bubble" stage.

From a business logic perspective, embodied intelligence is a typical "industry of the next decade." All current actions are just the "foundation." Therefore, the valuation cannot be measured by existing assets. Judging the future value of a high - rise building based solely on the value of its "foundation" clearly does not conform to business logic.

Whether it's the soaring valuation or the enthusiastic capital sentiment, what we don't understand now will be traceable in the future. A billion - dollar valuation is not just a "number in a capital game" but a cornerstone supporting the industry's progress towards a "trillion - dollar market in a decade." Instead of debating whether there is a bubble now, it's better to explore what enables enterprises to support a billion - dollar valuation. This is not only the answer urgently needed now but also the cornerstone for future development.

The Core Assets of a Billion - Dollar Valuation

The players in the billion - valuation club, such as Zhipu and Unitree, have distinct and diverse characteristics.

Let's first look at Zhipu. Zhipu's basic approach can be summarized as an "ecosystem strategy." Many industry insiders comment that Zhipu's style is "unlike that of a startup." After all, startups usually prefer a "single - point breakthrough" strategy, using a slightly advantageous strength to drive the expansion of financing and then gradually make up for their weaknesses. However, Zhipu seems to have no obvious weaknesses from the start.

The core elements of embodied intelligence are: a large and high - quality dataset, a hardware - strong and product - diverse ontology line, and a smarter brain that can understand the physical world. Zhipu has already achieved all three.

By the end of 2024, Zhipu had open - sourced the million - scale real - machine dataset AgiBot World. This year, Zhipu also released the one - stop development platform for embodied intelligence, Genie Studio, which has four core modules: "data collection, model training, simulation evaluation, and model inference." Simply put, in terms of datasets, Zhipu not only has a high - quality real - machine dataset (the "fish") but also masters the process from data collection to deployment (the "fishing method").

On the ontology side, Zhipu Robotics has formed three product series: the Expedition series, the Elf series, and the Lingxi series. These three product lines correspond to the current best - fit application scenarios for embodied intelligence: education, commercial use, and industry. They drive product implementation through their respective adaptations.

Image source: Zhipu Robotics

On the brain side, Zhipu has both the Qiyuan general embodied large - scale model and has launched the world model GenieEnvisioner. In addition, Zhipu has released the operating system Lingqu OS.

Beyond addressing the core pain points in the industry, Zhipu has also extended its reach in multiple directions, such as implementing a partner program, investing in upstream and downstream enterprises, and acquiring listed companies. A company that has been established for less than three years has completed the layout of an entire industrial chain. This may be Zhipu's core asset and the key to supporting its 15 - billion - yuan valuation.

In contrast, Unitree Technology is more focused, concentrating on its hardware advantages. This is related to its founder, Wang Xingxing. Having gained popularity by building a robot dog with just 20,000 yuan, Wang Xingxing's strict control over hardware performance and extreme cost - cutting have led to Unitree Technology, which continuously showcases its skills and breaks the industry's low - price record.

On the hardware advantage side, Unitree Technology's greatest highlight lies not in the dazzling hardware parameters but in its almost 100% self - developed industrial chain, which gives it absolute control.

In a mature industrial chain, full self - development of hardware has already given the enterprise a cost advantage and bargaining power. Moreover, with no standards in the upstream of embodied intelligence and most competitors still struggling with low yield rates and frequent returns, Unitree Technology's self - development not only provides a cost advantage but also a large - scale delivery advantage, which brings real short - term financial improvement. This may be one of the secrets of Unitree Technology's consecutive years of profitability.

Image source: Unitree Technology

In the long run, when there are no standards in the upstream and Unitree can achieve large - scale delivery through self - development, it means that Unitree is likely to gain the right to set hardware standards in the industrial chain. This will bring not only financial returns but also an ecological position.

Besides these two companies, there is also Galaxy Universal, which has great potential at the threshold of a billion - dollar valuation.

Galaxy Universal has many characteristics, but its most prominent one is its practical productivity value.

Supported by Galaxy Universal's self - developed large - scale model and the flexibility of its ontology, the wheeled humanoid robot Galbot not only has strong generalization ability but also shows "emergent capabilities." When facing human interference, it can also adjust the "perception - decision - execution" closed - loop in a timely manner and provide real - time feedback. This enables Galbot to improve quality and efficiency in real - world applications as required by the scenarios.

From unmanned pharmacies, unmanned retail stores, and simulated factory environments at various exhibitions to the Galaxy Space Capsule and the "snack bar" at real - life exhibitions, Galbot has transitioned from retail - scene applications to factory applications. This application is no longer just a "check - in point" at exhibitions but a real - life retail store where people can purchase and experience the productivity of embodied intelligence.

According to the Embodied Intelligence Research Institute, at the periphery of the WRC exhibition, the booth where Galbot serves as a salesperson has a daily turnover of over 30,000 yuan.

It's worth noting that Galaxy Universal is not just creating "single productivity" but a replicable business model. Its Galaxy Space Capsule is staffed by Galaxy Universal's Galbot robots, which can handle retail clerk functions such as receiving customers, taking orders, processing payments, retrieving products, making on - the - spot deliveries, and attracting customers through multi - language interactions. It only requires a minimum operating space of 9 square meters and can be flexibly adapted to scenarios such as streets, business districts, scenic areas, and transportation hubs. It supports both the independent construction of space capsules and the transformation of traditional stores, with an estimated delivery time of as fast as 15 days.

Image source: Galaxy Universal

Based on the basic capabilities of embodied intelligence, Galaxy Universal has extended to a more mature business model and explored unique commercial value. This value may only be a glimpse of the future transformation of production and living paradigms.

It can be seen that Zhipu, Unitree, and Galaxy Universal all have "real capabilities," and their respective core assets silently testify to the rationality of their valuations.

The Next Step for a Billion - Dollar Valuation

As the "Billion - Valuation Club" in the embodied intelligence track continues to expand, established pioneers like Zhipu and Unitree have firmly secured their positions, while more new players are eager to break into the core of the track. The focus of this industrial competition has shifted from "how to enter the billion - valuation threshold" to "how to take root and grow after reaching a billion - dollar valuation."

Whether it's the old players in the club trying to maintain their valuation or new players aiming for a breakthrough, they all have to face two core issues. These are not only the growth paths for old players but also the ways for new players to break the deadlock.

Firstly: In - depth exploration of key scenarios

Focus on one or two high - value tracks. For example, in the industrial scenario, form standardized solutions for pain points such as precision assembly and dangerous operations. Replace the general narrative of "wide - ranging technology coverage" with quantifiable efficiency - improvement data to truly transform technological advantages into commercial revenue.

Secondly: Transform mass production into a standard advantage.

Currently, embodied intelligence has entered a stage of rapid growth in orders and deliveries. With orders often exceeding hundreds of millions, it is necessary to further deepen industry hardware standards. At present, the industry is often trapped in the dilemma of "repeated debugging and frequent returns" due to inconsistent specifications of upstream components and unstable yield rates. Even if new players develop prototype machines, they may find it difficult to achieve mass production because they "can't find suitable standardized components," getting stuck in the last mile from "laboratory technology" to "market products." This requires not only the efforts of ontology enterprises but also the in - depth coupling of upstream and downstream. Only standardized hardware can improve delivery quality and thus reduce costs.

For new players, if they want to get a share in the billion - dollar track, they should not follow the old path of "comprehensive benchmarking." Instead, they should seize the niche scenarios that old players have not fully explored and establish a differentiated advantage through "extreme single - point breakthrough." This path of "creating great value in a small scenario" allows new players to avoid direct competition with old players and reach the billion - valuation threshold at a lower cost.

In addition, new players should actively embrace and participate in standardization construction, which is a shortcut to quickly achieve mass production and reduce costs. It can avoid delivery risks caused by "non - standard components."

Essentially, a billion - dollar valuation is not the "finish line" of the industry but the starting point of the "elimination round." Whether it's old players defending their territory or new players seeking a breakthrough, they must ultimately return to the core of "creating quantifiable value for the industry."

Only when robots can truly replace human labor to reduce production costs and improve efficiency, and when embodied intelligence changes from an "exhibition novelty" to an "enterprise necessity," will a billion - dollar valuation change from "capital expectation" to "industry consensus." This is the answer that all embodied intelligence enterprises must provide in the second half of the game.

This article is from the WeChat official account "Embodied Intelligence Research Institute." Author: Lv Xinyi. Republished by 36Kr with permission.