Behind XPeng's display of technological prowess, a number of humanoid robot companies have already gone bankrupt.

As the tides of capital ebb and flow more rapidly, we may need to re - examine the essence of the robotics track.

On the eve of 2026, robots are pushing open the door to the "anthropomorphic era".

In November, the Zhiyuan Yuanzheng A2 walked 106.286 kilometers and won the Guinness certification, becoming the first in the "endurance running" of humanoid robots. Almost at the same time, the Xiaopeng IRON robot walked in a catwalk, with its posture approaching that of a real person. They are no longer just a stack of mechanical arms and sensors, but are becoming more and more similar to the form of "humans" in terms of movements and behavior patterns.

The expansion of technological boundaries has brought collective excitement to the industrial chain. However, on one hand, cross - border players are pouring in and the financing amount is soaring; on the other hand, mass production is delayed and technology implementation is weak. Is this wave of robotics, ignited by both capital and technology, the main line of the next industrial revolution or just another speculative bubble that starts high and ends low?

Part.1 The Third Pole of Industrial Products Enters the 2.0 Stage

Elon Musk once asserted in a conversation with Jensen Huang: "AI will keep you busier, and humanoid robots will become the largest industry in history." Although truly general and mature products have not yet emerged, it has not affected the judgment of the world's richest man on the market prospects of the robotics industry.

Actually, since the Turing test in the 1950s and 1960s, which raised the question of whether robots can think, robots have gradually stepped out of the imagination of science - fiction movies and become the third standardized industrial product with the potential for a trillion - level scale after mobile phones and cars.

A practitioner who has long been deeply involved in the consumer electronics industry told "TrueView" that the uniqueness of robots lies in breaking the barriers between C - end consumption and B - end industries, achieving two - way empowerment and a value closed - loop for both ends.

Put simply, different from mobile phones focusing on personal communication and smart cars emphasizing travel scenarios, robots can either be life assistants in households or production laborers in factories. This dual - track attribute enables it to meet the dual needs of consumption upgrading and industrial upgrading.

The person further pointed out that it is this dual - track attribute, combined with the breakthrough explosion of AI technology around 2020, that has promoted the robotics industry to quickly enter the 2.0 stage of large - scale development from the 1.0 stage of conceptual exploration. The industrial portrait of the 2.0 stage has three prominent characteristics: rapid market growth, deep integration into scenario practice, and a two - sided situation of enthusiasm and bubbles.

In terms of industry scale, the growth potential of the robotics track is clearly visible. A report released by the Development Research Center of the State Council predicts that the market scale of embodied intelligence (with robots as its main industrial carrier) in China is expected to reach 40 billion yuan in 2030 and exceed one trillion yuan in 2035. TrendForce believes that the shipment volume of robots will maintain a compound annual growth rate of over 20% in the next five years, and it is also expected that its revenue scale will be nearly 128 billion US dollars in 2029.

Among them, the market for industrial robots is relatively broader. The "World Robotics Report 2025" recently released by the International Federation of Robotics shows that in 2024, 542,000 new industrial robots were installed globally, more than double the number 10 years ago. Among them, 295,000 were newly installed in China, accounting for more than half, and it will maintain an average annual growth of at least 20% in the future.

The core sign of entering the 2.0 stage is the shift from single - function implementation to multi - scenario in - depth adaptation. In the early years, most robots were "functional tools". Industrial robots could only repeat simple welding and moving actions, and household robots could only perform basic operations such as sweeping and mopping the floor, with extremely poor scenario adaptability.

Now, with the iteration of large AI models and sensor technology, robots are becoming smarter and more flexible. The collaborative robot of Midea KUKA can accurately identify the actions of workers and achieve human - machine collaborative operations, with the error controlled within 0.1 millimeters. In the medical scenario, surgical robots can perform minimally invasive surgeries with far higher precision than manual operations.

Goldman Sachs previously predicted that humanoid robots would be first applied to factories from 2024 to 2027 and to the consumer market from 2028 to 2031. Judging from the current industrial progress rhythm, Goldman Sachs' above prediction is obviously too conservative.

The law of moving from being usable to being easy to use not only magnifies the industrial value of robots but also allows more players to see the opportunities, laying the groundwork for the congestion of the track.

Part.2 The Two Sides of the Industry: Bubbles in the Fiery Carnival

The other side of the robotics 2.0 era is the "coexistence of ice and fire" in the industry. On one hand, there is a fiery carnival of capital and technology; on the other hand, there is a splash of cold water in the form of bubble disputes. This contrast constitutes a typical feature of the development of emerging industries.

The fiery side is manifested in three aspects: financing, players, and technology iteration.

In most cases, capital is the forerunner of trending and emerging industries. Their profit - seeking nature gives them an innate sense of smell for such things. Public data shows that according to IT Juzi, in the first three quarters of this year, the financing amount in the domestic robotics field alone has reached 38.624 billion yuan, 1.8 times the total financing amount of 21.254 billion yuan in 2024.

As "embodied intelligence" is included in the national strategic emerging industries, a new round of investment boom is on the way. At the same time, different from the angel investment methods in previous years, investors may focus more on the commercialization efficiency and monetization speed of manufacturers.

The layout of capital directly drives the other fiery side, that is, the congestion of the track, especially in the terminal product interface of the downstream of the robotics industrial chain.

Automobile companies, mobile phone manufacturers, Internet companies, and home appliance giants have all crossed the boundaries and entered the field. Robotics seems to have become an industry where "anyone can give it a try", which is a business scene that most emerging industries have hardly ever had before.

A senior executive of Zhiyuan once pointed out, "We expect it to undertake repetitive and boring tasks such as delivery, sorting, and reception in scenarios such as retail, catering, corporate front desks, and even household services, and deeply integrate into our daily lives as a reliable partner of humans."

Looking back, in terms of the layout paths of relevant companies, the market has stepped out of the short - term thinking of chasing trends and turned to in - depth cultivation of niche markets. For example, traditional manufacturing enterprises extend across boundaries by leveraging technological homology, and technology companies focus on breakthroughs in niche scenarios. This differentiated layout reduces the industrial trial - and - error cost, accelerates the collaboration of the entire industrial chain, makes the industry development more resilient, and forms a dual - wheel - drive pattern of core component research and scenario application implementation.

This has also promoted the continuous acceleration of the technology iteration speed. For example, the combination of large AI models and robots is becoming closer. Natural language processing and computer vision technology make the interaction of robots smoother, and the precision of core components such as servo motors and reducers continues to improve.

However, the other side of the influx of hot money is the breeding of bubbles and the accumulation of risks.

Goldman Sachs clearly pointed out in the "Global Robotics Industry Outlook 2025" that there is an obvious over - valuation phenomenon in the current robotics track. About 60% of global robotics companies are valued at more than 100 times their revenue, and the over - capacity rate of the global robotics industry may reach 25% in 2025.

A large number of enterprises lack core technologies and only rely on assembly and branding to ride on the wave of hot topics. To some extent, the current bustling scene of the robotics track is quite similar to the O2O boom in 2015. It looks extremely lively, but many enterprises are engaged in pseudo - innovation and have not really solved the pain points of scenarios.

Over time, robots are falling into a dilemma. On one hand, there is a significant deviation between market expectations and actual development; on the other hand, the real difficulties they face in performing open - ended tasks such as fine hand operations are generally underestimated, and the challenges of technological breakthroughs far exceed the existing understanding. In addition, referring to the new energy vehicle industry, while facing involution and homogenization, robots also have prominent problems such as insufficient national standards, unclear safety boundaries, and ethical disputes.

Since 2025, the embodied robot track has entered a recession period. Many enterprises have successively announced shutdowns. For example, the humanoid robot startup K - Scale Labs, which was established only one year ago, announced its closure; Aldebaran, which had 20 years of industry experience and once launched the Pepper robot, could not escape the fate of bankruptcy; and the US AI companion robot enterprise Embodied officially terminated its operation. In October of the same year, there was also a setback in the domestic embodied intelligence startup force, and OneStar was reported to have disbanded.

Even CloudMinds Technology, which was once valued at up to 3 billion US dollars, has fallen into a capital chain break - down dilemma and had to launch a contraction strategy to save itself.

Part.3 Mass Production or Technology: Which Is the Core Constraint?

When the tide recedes, the core contradiction of enterprise shutdowns points to the commercialization bottleneck. Is mass production or technology the key factor restricting the industry's progress and enterprises' competitiveness?

Logically, mass production can drive down prices and promote market penetration. However, the prerequisite for mass production is not a lack of manufacturing capacity or an unworkable business model. The fundamental bottleneck still lies in the immaturity of technology.

In other words, robots may still lack a "dedicated large model" different from general large models. Especially in key fields such as motion control and autonomous navigation, the technology link from intelligent perception to decision - making and then to execution is not proficient enough, which is also the basis for Musk to emphasize that robots are not yet fully practical.

A dedicated large model can enable robots to receive end - to - end instructions and better perform tasks. Wang Xingxing, the founder of Unitree Robotics, has repeatedly emphasized that the bottleneck for the large - scale development of robots is the insufficiency of AI models and called on the industry to accelerate the arrival of the "ChatGPT moment" of embodied intelligence.

The next question is, how to break through in the model field?

Some industry insiders emphasize that this essentially tests the R & D strength of all players in the embodied intelligence track. Among them, the "redundancy of R & D resources" is the core factor affecting the R & D efficiency of the dedicated large model for robots.

Currently, the players can be roughly divided into two categories: cross - border players and startups focusing on the track. Although the former have the support of their main businesses, it is often difficult for them to allocate R & D resources to the dedicated large model for robots; the latter, although focused, are often restricted by a shortage of funds and have difficulty keeping up with the iteration rhythm of large models.

Teams with sufficient R & D resource reserves can plan the R & D path of the dedicated large model for robots in advance, finalize the product layout and R & D progress simultaneously, allocate core human resources in advance, and prepare sufficient R & D funds to seize the technological opportunity. They can take the lead in the model improvement of core scenarios such as motion control and autonomous navigation. As for the construction of the supply chain, material reserves, and marketing, these can be gradually promoted as secondary priorities.

Therefore, as the tides of capital ebb and flow more rapidly, we may need to re - examine the essence of the robotics track. This is not a short - lived bubble game but a long - term test of industrial patience and strategic determination.

The current "congestion" is precisely a typical feature on the eve of the industrial explosion. It does not mean that the direction is wrong but signals the end of the simple model. The era when companies could obtain valuations by assembly, branding, and hyping concepts is coming to an end. The real competition has just begun, and the ticket to the second half is the "brain" of robots.

The future industry pattern will also be divided accordingly. Teams with abundant resources that can painstakingly tackle the "perception - decision - execution" closed - loop will define the "intelligence" of machines; while players who only stay at the level of form imitation and lack core AI capabilities may run aground in the deep - water area of technology.

Therefore, for the robotics industry, we might as well have more long - term patience and less impetuosity in chasing trends. Its future does not depend on the number game of financing amounts but on whether intelligent emergence can be achieved at the key model level.

Whoever can first enable robots to truly understand the world and independently cope with complex environments will truly hold the key to the next era.

This article is from the WeChat official account "TrueView", author: Laolian, editor: Yong'e. Republished by 36Kr with permission.