The second half of the humanoid robot development in 2025: Listing, bubble clearing, and value reevaluation

“It feels like the market for embodied intelligence has cooled down recently. After launching several projects in a row, due to the soaring valuations and disputes over commercialization, investors are in a wait - and - see state,” sighed Xiaobei, an FA in China who focuses on the embodied intelligence track.

However, the industry's enthusiasm has not dissipated. At the beginning of this month, the first acquisition case of embodied intelligence on the Science and Technology Innovation Board triggered public discussion. ZHIYUAN ROBOTICS plans to “reverse acquire” the Science and Technology Innovation Board - listed company Shangwei New Materials for 2.1 billion yuan. Previously, the industry generally believed that Unitree Technology would become the “first stock in humanoid robots.” ZHIYUAN ROBOTICS' explosive move has shifted the focus of embodied intelligence from technological breakthroughs and practical applications back to capital operations.

In fact, the financing activities in the embodied intelligence track remain active. Recently, Xinghai Map completed strategic financings in Series A4 and A5 in succession, with a total financing amount of over $100 million in the two rounds. Xingdong Jiyuan received nearly 500 million yuan in Series A financing. Yun Shenchu completed financing of hundreds of millions of yuan. Xiaoyu Zhizao received over 100 million yuan in Series A+ financing. Tashi Zhihang completed $122 million in Angel + round financing... But the most sensational one was that Yinhe Tongyong completed a new round of financing of over 1 billion yuan, setting a record for the largest single - financing amount in the embodied intelligence track this year.

Currently, the embodied intelligence track is undoubtedly in a “polarized” situation. While Zhu Xiaohu shouted “clear the position and be bearish” on humanoid robots, state - owned assets, along with JD, Meituan, Tencent, and Huawei, are investing 1.1 billion yuan in Yinhe Tongyong. Tashi Zhihang even obtained $120 million in angel - round financing, setting a record for the highest angel - round financing amount in China's embodied intelligence field.

However, behind these large - scale financings, it seems that only leading institutions and enterprises are actively involved, and small and medium - sized players are rarely seen. This may reflect that the embodied intelligence industry is at a critical turning point from “wild growth” to “bubble clearing.” After all, no matter how appealing the financing stories are, they ultimately need to be verified by real money in the market and sustainable profitability.

Part.1 Disparity: Investors Clear Their Positions vs. Massive Financing and Listing

The embodied intelligence track in the first half of 2025 can be described as “disparate.” On one hand, investors are shouting that they will not invest. Some investment institutions have become more cautious due to the insufficient technological maturity and the long commercial return cycle, publicly questioning the “bubble risk.” For example, Zhu Xiaohu withdrew from Xinghai Map and Songyan Power in batches, stating bluntly that “the main customers are only research institutions, and the commercialization path is unclear.” On the other hand, large - scale financing records are constantly being set.

According to IT Juzi data, there were 286 financing events in the primary market of the domestic robotics field in the first half of the year, an increase of 166 compared with the same period last year, with a year - on - year growth rate of 138%. It is estimated that the total financing amount of robot startups in the first half of the year reached 27 billion yuan.

Let's first look at ZHIYUAN ROBOTICS, which triggered public discussion. It plans to obtain control of the Science and Technology Innovation Board - listed company Shangwei New Materials through agreement transfer and tender offer, acquiring at least 63.62% of the latter's shares. Since the announcement was released, Shangwei New Materials has had 11 consecutive daily limit up trading days.

Although the relevant person in charge of ZHIYUAN ROBOTICS denied that this move constitutes backdoor listing or restructuring listing, the fact that a startup company established only two years ago can so quickly initiate a controlling acquisition of a listed company is already shocking enough. Behind this rare “reverse acquisition” operation, it reflects the profound changes in the capital structure of AI and robotics startup projects.

ZHIYUAN ROBOTICS currently has three robot families: Expedition, Elf, and Lingxi. Its products cover various commercial scenarios such as interactive services, industrial manufacturing, commercial logistics, and scientific research and education. It is expected to ship thousands of units in 2025. Its shareholder lineup is impressive. In addition to industrial giants such as Tencent, JD, BYD, SAIC, BAIC, and TCL, it recently received strategic investment from Zhengda Robotics under CP Group. Strong capital support has not only skyrocketed its valuation but also given it the opportunity to “dominate” resource integration.

Another protagonist in the financing boom is the equally young Yinhe Tongyong Robotics. This company, also established in 2023, completed a new round of financing of over 1 billion yuan, setting a record for the largest single - financing amount in the embodied intelligence track this year. Its founder, Wang He, was born in 1992. He graduated from Tsinghua University in 2014 and obtained a doctorate from Stanford University in 2021. Then he joined Peking University as an assistant professor and doctoral supervisor at the Frontier Computing Research Center, focusing on the research of embodied intelligence.

In June 2024, Yinhe Tongyong released its first embodied large - model robot, Galbot G1, which is also its only robot product. G1 adopts a body structure of “double arms + foldable body + wheeled chassis,” not a bipedal form, and focuses more on the operation ability of the upper limbs. Since its establishment, the company has developed several embodied intelligence models, including the end - to - end embodied grasping basic large model GraspVLA, the end - to - end embodied large model GroceryVLA, and the product - level end - to - end navigation large model TrackVLA.

Since Unitree Robotics became popular after performing the “Yangge Dance” at the Spring Festival Gala in the Year of the Snake, although the enthusiasm in the entire embodied intelligence market has been rising, the long commercial return cycle has always been a key obstacle in front of the industry, causing doubts among investors and the market. ZHIYUAN ROBOTICS' aggressive capital operation, Yinhe's record - breaking financing, and hundreds of financing events undoubtedly demonstrate the firm optimism of some capital about the future of this track. However, it is still unknown whether these real - money investments will ultimately yield sufficient returns.

Part.2 Delivery: Not Fully Cooled, but Entering a Value Re - evaluation Period

Currently, the competition focus in the humanoid robot industry is more on the delivery ability of enterprises. The order fulfillment, accelerated mass production, price reduction, and scene expansion of leading enterprises all indicate that the industry is entering a critical stage from concept disputes to value re - evaluation.

On July 11, China Mobile spent 124 million yuan to purchase robot products from ZHIYUAN ROBOTICS and Unitree Technology, becoming the largest single - order of humanoid robots known in China so far.

In addition, ZHIYUAN ROBOTICS has announced 8 winning bids since December 2024. Its new smart robot, Lingxi X2, started recruiting partners in May this year. The price range of this series of products is from hundreds of thousands to several million yuan. It is expected to achieve large - scale shipments in the second half of this year. ZHIYUAN ROBOTICS will adjust its production capacity and output according to market feedback, hoping to achieve a delivery scale of several thousand units by the end of 2026.

Unitree Technology, which was established earlier, has also made earlier progress in commercialization. Since 2025, the company's winning bid projects have mainly been concentrated in fields such as education and scientific research, and factory inspection. In March alone, it won more than a dozen projects, with application scenarios including universities, science and technology museums, and intelligent inspection. The purchasers include the School of Artificial Intelligence of Sun Yat - sen University, Shenzhen University, the permanent exhibition hall of Weihai Science and Technology Museum, Shanghai Lingang Water Supply and Drainage Development Co., Ltd., the humanoid robot laboratory system platform of Qilu University of Technology, China Mobile (Hangzhou) Information Technology Co., Ltd., the College of Mechanical and Electronic Engineering of Northwest A&F University, and the general humanoid robot training platform of Tongji University.

After the CES 2025 exhibition, Unitree also received several overseas orders, including a purchase agreement for hundreds of quadruped robots with a warehousing and logistics company in the United States for warehouse inspection.

In 2025, Ubtech also made progress in product orders and commercialization of humanoid robots. In the middle of this month, the Tianggong Xingzhe humanoid robot launched by Ubtech Technology has received orders for 100 units, and it is expected to deliver more than 300 units in 2025. Among its main products, the industrial humanoid robots WalkerS1 and WalkerS2 and the commercial humanoid robot WalkerC, the WalkerS series has become the humanoid robot that has entered the most automobile factory training programs globally. According to internal sources from Ubtech, it has received over 500 units of intention orders from automobile factories. Although the unit price has not been made public, the high - performance version is expected to be priced at up to 500,000 yuan in the high - end market.

In terms of market scale, the humanoid robot is a potential multibillion - dollar track. The High - tech Robotics Industry Research Institute (GGII) predicted in the “Blue Book on the Development of the Humanoid Robot Industry in 2025” that the global humanoid robot market scale will be 6.339 billion yuan in 2025 and will exceed 64 billion yuan in 2030.

In the domestic market, analysts from the Zhongshang Industry Research Institute predicted that the scale of the Chinese humanoid robot industry will reach 5.3 billion yuan in 2025. The “Research Report on the Humanoid Robot and Embodied Intelligence Industry in 2025” shows that the market scale of Chinese humanoid robots is expected to reach 8.239 billion yuan in 2025, accounting for about 50% of the global market.

As leading enterprises accelerate mass production, price competition in the humanoid robot industry has begun to emerge. It is reported that Tesla has clearly set the target cost at $10,000, and the product prices of domestic enterprises have also fluctuated. The Unitree G1 humanoid robot is priced at 99,000 yuan, but now it can be bought for only 50,000 - 60,000 yuan on a second - hand platform. In the future, there may also be a price war in the humanoid robot market, just like the current new - energy vehicle market.

Meanwhile, the industry is actively exploring emerging scenarios. For example, Yinhe Tongyong has deployed robots in unmanned pharmacies, Unitree's robotic dogs have entered 30 university laboratories, and the feasibility of application scenarios such as logistics and power inspection in industrial scenarios is also being gradually verified.

Currently, the industry is going through a painful stage from bubble disputes to value re - evaluation, and “delivery” has become the core and key factor in the competition among embodied intelligence enterprises.

Part.3 Survival Game: Breaking the Mismatch between Technology and Market Demand

Although humanoid robot orders seem to be everywhere, currently about half of the orders in the market are concentrated in non - core application scenarios such as performance display and data collection.

The essence of the difficulty in the implementation of humanoid robots is the disconnection between technological ideals and commercial realities, and the contradictions are concentrated in dimensions such as technological bottlenecks, scenario mismatches, mass - production difficulties, and lack of standardization.

First of all, there are still few domestic companies focusing on the “brain” of humanoid robots. The error rate of language models in processing cross - modal tasks is still relatively high. Due to the lack of physical interaction training data, execution deviations often occur. For example, “fetch a glass of water” may be wrongly executed as “pour the kettle.” The control delay problem of the “cerebellum” is also prominent. The time it takes for a bipedal robot to pass through a stepped terrain is much longer than that of a human. The reliability of motion control is insufficient. For example, the high fall rate of robots at the Beijing Marathon previously caused extensive discussions. There are also disadvantages in the hardware cost structure. For example, dexterous hands still rely heavily on imports.

“The market's demand for humanoid robots has always existed, but because the technological capabilities of the products have not met the expectations of the market and the public, there has been an imbalance or mismatch between the market and demand. This is the core reason for the cooling of the entire market,” Li Jun, the technical director of a leading robot company, told TrueView.

Secondly, the problem of scenario mismatch also restricts development. The requirements for robots in different scenarios vary significantly. Nursing homes need “humanoid appearance” to reduce the sense of alienation, while industrial scenarios require “function first.” At the same time, due to the fragmented demand and relatively cost - sensitive nature of the family scenario, the penetration cycle of humanoid robots in this scenario will be longer.

Thirdly, the humanoid robot industry still faces mass - production difficulties. In 2025, Tesla plans to produce 10,000 units of Optimus, while the average delivery volume of domestic enterprises is less than 100 units, and the delivered products are mainly used for data collection. “The price of humanoid robot products is hundreds of thousands of yuan per unit, which makes companies and individual consumers hesitant to make a purchase,” Li Jun pointed out the restriction of high prices on market expansion.

Unitree Technology, which has strong cost - control ability in the industry, priced the standard version of Unitree G1 at only 99,000 yuan, and the EDU version supports secondary development (such as installing the NVIDIA Jetson Orin module). Compared with the pricing of similar products in the industry, which are mostly around 500,000 yuan, its products have an advantage in cost - performance. It can be seen that cost - control ability has become one of the key factors for its success.

Finally, the lack of standardization also brings many challenges. In addition to the urgent need to solve the joint compatibility problem of service/industrial robots, at the legal level, when negative events such as injury or violence occur during the application of humanoid robots, there are currently no clear laws, regulations, and safety standards for supervision.

Now, a tug - of - war between capital frenzy and real - world challenges has begun. Although policies and industrial capital continue to provide support, it is expected to promote the mass - production of humanoid robots in industrial scenarios to reach the scale of thousands of units in the next 2 - 3 years. However, in the long run, enterprises must face technological challenges directly.

The current embodied intelligence track is similar to the new - energy vehicle manufacturing boom in the past. Since embodied intelligence has received wide attention from the market since the second half of 2023, two years have passed. Capital has started to shift from “blind pursuit” to “rational screening,” and the market has gradually transitioned from “concept fever” to “scenario fever.”

The core of the competition among enterprises in the next stage lies in who can truly take root and succeed in niche scenarios, and who can effectively reduce costs, increase efficiency, and even create profits for customers. Those enterprises with self - hematopoietic ability naturally don't need to fear market fluctuations.

Note: Xiaobei and Li Jun are all pen names.

This article is from the WeChat public account “TrueView”. Author: A Wen, Editor: Yong E. Republished by 36Kr with permission.