As the countdown sounded, a robotic arm quickly located a curtain-revealing rope device with a radius of about 4mm, grasped the rope accurately, and the curtain slowly fell.
The opening curtain was pulled open by the dual-arm robot Vonnex - its multimodal perception system, similar to that of humans, can process visual, force, and voiceprint information simultaneously, and it has now become a "worker" in scenarios such as education.
This is a scene from the unveiling ceremony of the Hong Kong University of Science and Technology's von Neumann Institute of Artificial Intelligence today (May 9). As the first practical AI research institute in the Greater Bay Area, von Neumann tackles embodied intelligence and other directions, focusing on the core demands of the global industrial intelligent transformation. The appointment of its dean has attracted much attention - the helmsman is none other than Jia Jiaya, the world's top AI expert and founder and chairman of the unicorn company Simo Technology. This overlapping identity is a footnote to the institute's positioning of "deep integration of academia and industry".
At this time, 70 kilometers away in Bao'an, Shenzhen, in Simo Technology's intelligent manufacturing factory, a batch of non-humanoid industrial robots are undergoing final debugging before leaving the factory. They do not have dazzling anthropomorphic appearances, but with AI technology specially polished for production lines, they are about to rush to customer factories scattered around the world.
This scene reminds me of a phenomenon that many investors have talked about recently - when humanoid robots are still competing in marathons, the robots that first landed on the industrial track have already lined up to enter the factory to get work numbers and make money. In the final analysis, the truth is in the industry.
Say goodbye to dancing and somersaults:
robots are now working in factories
Since the beginning of this year, the popularity of embodied intelligence has been vividly remembered.
The latest scene is the humanoid robot taking to the sports field - the 2025 Beijing Yizhuang Humanoid Robot Half Marathon, 20 humanoid robot teams completed a 21-kilometer journey; Wuxi Embodied Intelligent Robot Games, more than 150 robots took to the sports field, racing back and forth, shooting basketballs, robot fighting, and continuous backflips.
These lively scenes attracted countless eyes, but investors had mixed feelings. There is no doubt that this is a microcosm of the outbreak of the domestic robot track and a series of attempts at robot intelligence. Whether in short videos or in reality, robots dancing street dance and continuous backflips are "more and more like people", but robots that cost tens of thousands or even hundreds of thousands of yuan are far from the successful commercialization of the 10 billion-level market?
The reality is in front of us: if it cannot be commercialized in large quantities, the popularity of humanoid robots will slowly dissipate over time. On the other hand, in the front line of industrial production, industrial intelligent bodies equipped with large models have replaced part of the labor, and have opened up a breakthrough for the commercialization of embodied intelligence.
These non-humanoid robots that have been put into use first are not mechanical arms or industrial robots in the traditional sense, or we should define them as "industrial intelligent bodies". In the past, traditional equipment relied heavily on preset programs to perform single tasks, lacked the ability to learn and judge independently, and had difficulty making real-time decisions in the face of complex environmental changes, and rarely entered the mainstream vision.
Today, the industrial intelligent bodies that have landed on industrial production lines are equipped with AI brains and multimodal perception systems, which have changed from passive execution to active decision-making, and have truly achieved a qualitative change from "mechanical arms" to "intelligent bodies". As mentioned above, Simou Technology has entered the assembly lines of 300 leading companies in the world and quietly realized scene applications, which is a place that venture capital circles often overlooked in the past.
Looking back, Jia Jiaya joined the Chinese University of Hong Kong after obtaining a doctorate degree from the Hong Kong University of Science and Technology in 2004, and later became a tenured professor. In many industry-university-research cooperation projects, Jia Jiaya keenly captured the complex needs of factory subdivision processes and began to realize that industrial intelligence requires full-stack integration from underlying algorithms to hardware equipment. This is a field with hard demands but extremely high barriers.
In December 2019, Jia Jiaya founded Simu Technology. Unlike other AI companies, Simu focused on intelligent manufacturing from the beginning, and then quickly obtained angel round financing from IDG Capital, and then completed three rounds of financing from leading institutions such as Sequoia China and Lenovo Capital in a year and a half, becoming one of the youngest unicorns at the time.
Correspondingly, Simu's road to industrial intelligence was paved. In the month of its establishment, the company undertook the first case of AI smart manufacturing. Now, in less than 5 years, Simu's non-humanoid robots and intelligent products have entered a number of leading companies such as Tesla, Carl Zeiss, BOE, CRRC, BYD, and Luxshare Precision.
This is just one of the representative cases. At the end of 2023, UBTECH, which was listed in Hong Kong as the "first stock of humanoid robots", has continued to face the dual torture of commercialization and profitability since it landed in the capital market. Recently, UBTECH issued a voluntary announcement, announcing that it had obtained a small-batch industrial manufacturing scenario procurement contract for embodied intelligent humanoid robots, and related products will soon be stationed in automobile factories to participate in production and manufacturing.
This progress is seen as a key breakthrough in the commercialization of the industrial track. When the market is skeptical about its profit prospects, UBTECH is pinning its hopes of reversing the situation on industrial manufacturing, the most mature and commercially promising track in the field of robots.
One year after UBTECH went public, Yuejiang Technology, which specializes in robotic arms, also landed on the Hong Kong Stock Exchange and became the "first Chinese collaborative robot stock." Previously, the investment community learned from founder Liu Peichao that Yuejiang Technology's first landing and largest share were also industrial scenes.
Where the spotlight cannot reach, a super track is quietly exerting its strength.
A case of a unicorn in the Greater Bay Area
Of course, it is not a simple matter to send robots from the laboratory to the assembly line.
After communicating with investors, although the industrial intelligent market has huge potential and is regarded as a trillion-level blue ocean, the professional barriers are high and the scenes are highly fragmented, which makes many participants discouraged. Everyone mentioned that most manufacturing companies generally have "three high anxieties" when they transform themselves into intelligent ones - high customization costs, high trial and error risks, and high integration difficulties.
To this end, we asked Professor Jia Jiaya about the key pain points of the current industry on the spot -
First, the algorithm technology is not mature enough. The core of the intelligent body is the algorithm. Although the AI large model has shown strong capabilities in handling complex tasks, when it is combined with the real physical world, especially industrial-level production scenarios, it is not professional and universal enough, and robots cannot make the same flexible judgments and decisions as humans.
Secondly, the development speed is slow. This includes the integration and compatibility of technology, and it also needs to improve maturity through large-scale applications.
"Robots and robotic arms have complex structures. There is a long industrial chain from power management, basic components to sensor systems, planning and control. Integrating intelligent body technology with existing hardware and software systems not only involves cross-industry and cross-disciplinary, but also has a strong correlation with the entire industrial chain, which has slowed down the development of related intelligent body products to a certain extent."
In addition, does an embodied intelligent robot have to have a human form? Faced with the overwhelming craze for humanoid robots, primary market investors have raised such doubts, and Simou Path may have answered this question.
To a certain extent, the degree of intelligence determines the upper limit of the industry value, and the adaptability of the form determines the speed of commercialization. Therefore, since the establishment of Simu, a strategic choice of "removing form and retaining intelligence" has been made, which is not limited to form. This idea has been recognized by many investors. In June last year, Hong Kong Investment Company took action, and the first project was Simu Technology. Hong Kong Investment Company has a steady style. The investment team conducted a strict due diligence on Simu, and the commercial application scenario has become a key consideration.
The investment community has obtained a set of data: in the five years since its establishment, Simu has entered multiple industries such as automotive manufacturing, new energy, 3C, and precision industry through industrial intelligent products such as AI machines and robotic arms, and has cumulatively tested nearly 10 billion industrial products.
Today, the unveiling of the von Neumann Artificial Intelligence Research Institute can be seen as an acceleration of Hong Kong's layout in the field of artificial intelligence, and it also re-emphasizes the importance of the application of robots.
The next storm: What kind of robots does China need?
Let's open our horizons.
At this moment, the primary market is arguing over whether to invest in embodied intelligence or not. But in summary, there are still two points of consensus: one is that valuation bubbles have emerged under the hot financing of embodied intelligence; the other is that commercial scenarios are still being explored.
Although most people firmly believe in the broad prospects of this track, in the short term, humanoid robots still have a long way to go in terms of technology and cost before they can enter thousands of households.
For startups, the first priority is to survive in the market. Looking overseas, American star companies Anki, home robot company Jibo, companion robot Embodied, etc. all fell rapidly after a burst of popularity; the domestic old robot unicorn Dahua also reached the edge of the cliff because of its lack of hematopoietic function.
These cruel realities all show the necessity of embodied intelligence companies to take the lead in laying out industrial scenarios. After experiencing this wave of humanoid robot craze, investors began to return to industrial embodied intelligence to find projects.
This is not difficult to understand. On the one hand, this year's humanoid robot financing boom has seen frequent valuation surges; however, industrial robots have not yet reached a high degree of consensus in the primary market, so financing valuations are relatively calm, and there is still huge room for explosion.
In terms of commercial implementation, which everyone is most concerned about, the industrial track can be said to be a scene direction with fewer "demand-first" scenarios in the industry. The demand for robots in industrial scenarios such as automobile manufacturing and precision instrument manufacturing is much higher than in other scenarios, so the technology development in this field is relatively long-term and stable.
An investor friend also added: Because the demand scenarios are relatively fixed, the costs of new industrial robot products such as intelligent bodies are relatively controllable in both hardware and AI empowerment, and the prices are gradually decreasing, which is one of the reasons why it is more suitable for commercialization at present.
In early 2025, Goldman Sachs pointed out in a report that (humanoid) robots are expected to be first used in factories between 2024 and 2027, and the time to enter the consumer market is expected to be between 2028 and 2031. This prediction once again proves that industrial embodied intelligence will be the first to explode.
"Industrial embodied intelligence is expected to become the next hottest track." Such voices have become more and more frequent recently.
In the words of investors, the embodied intelligence industry is still in a relatively early stage of development. At this stage, the deep integration of embodied intelligence and AI big models has been first applied in the industrial field. The next step is to achieve large-scale application in the industrial field and form industrial value.
There is no doubt that embodied intelligent robots, no matter what form they appear in, will be the best carrier of AI. Stimulated by new technologies, the era of Industry 4.0 - the comprehensive industrial intelligent revolution marked by industrial intelligent bodies, is accelerating.
This will reshape the global industrial landscape. In 2025, the term "embodied intelligence" was written into the government work report for the first time, and its strategic significance is self-evident. The river of history is vast and surging. Perhaps looking back many years later, this will be a technological competition that is related to the fate of the country.
Source: Investment Circle, Yang Wenjing
