Economic Observer New Technology
Economic Observer Network According to the Securities Times, recently, several leading embodied intelligent robot companies have disclosed large orders, with the buyers primarily being manufacturing enterprises, and the application scenarios mainly in the industrial field.
With humanoid robots frequently “entering factories to work,” can they really adapt to factory environments? What is the progress of their commercialization? Journalists from the Securities Times learned that the “entry into factories” is just the first step, aimed at connecting real scene data, and completing technological iteration and upgrades based on that data. Industry insiders believe that most current orders for humanoid robots are of a “pilot verification” nature, still far from achieving “commercial cost-effectiveness.”
Frequent Large Orders for Humanoid Robots
Recently, humanoid robots have seen frequent occurrences of orders worth hundreds of millions or thousands of units. According to incomplete statistics, seven humanoid robot companies have officially announced orders exceeding hundreds of millions or more than a thousand units, including UBTECH, Zhiyuan Robotics, Yushu Technology, Zhi Square, Stardust Intelligence, Songyan Power, and Yuanli Unlimited.
In October, UBTECH announced that its Walker series humanoid robots have received over 630 million yuan in orders for the year; Zhiyuan’s G2 robot secured several hundred million yuan in orders from leading smart product ODM (Original Design Manufacturer) Longqi Technology; Yuanli Unlimited, founded by a former Alibaba executive, announced a 260 million yuan order with Shihua Cultural Tourism Holding Group. Songyan Power also recently disclosed that it has secured over 2,000 commercial orders in the first half of this year, with a total contract amount expected to exceed 100 million yuan.
In September, Stardust Intelligence reached a strategic cooperation for a thousand-unit order of humanoid robots with Xian Gong Intelligent, with media estimates placing the order amount at about 500 million yuan; Zhi Square and Shenzhen Huizhi Internet of Things reached a strategic cooperation, planning to deploy over 1,000 embodied intelligent robots at Huike’s global production base over the next three years, with a cooperation order amount close to 500 million yuan.
The buyers behind these orders are mostly manufacturing enterprises, followed by universities and government units, with the products primarily landing in industrial scenarios, followed by academic research, cultural and entertainment guidance scenarios, and also including commercial sales, warehousing logistics, and traffic guidance scenarios.
For example, in the cooperation with Longqi Technology, Zhiyuan’s G2 robot will initially focus on application in the tablet production line, achieving large-scale deployment of embodied intelligent robots in consumer electronics assembly manufacturing scenarios. Zhi Square’s wheeled dual-arm AlphaBot series robots will enter Huike’s global production base, covering the entire process from warehousing logistics, material handling, component assembly to quality inspection testing. This year, UBTECH has received multiple orders from automotive companies; in March, Dongfeng Liuzhou Automobile announced the deployment of 20 UBTECH industrial humanoid robots Walker S1 in its automotive manufacturing plant.
Yushu Technology’s orders are also primarily B-end, including research universities, government, and enterprises, with enterprises in media, public safety, and supply chain sectors. According to third-party statistics, among the more than 100 winning projects of Yushu Technology from April 8, 2021, to March 17, 2025, research universities have become the main buyers, with nearly 30 schools procuring Yushu products through public bidding, covering the entire chain of robot research and development.
“Entering Factories” is Just the First Step
Industrial scenarios are seen as the most suitable for the commercialization of humanoid robots; however, there is always controversy about humanoid robots “entering factories”: their efficiency is far lower than that of humans, costs are hard to compete with human labor, and the turning point for general-purpose technology is still unclear… With wheeled robots already capable of handling most tasks, is there really a need for humanoid robots to “work” in factories?
Many first-tier investors and humanoid robot manufacturers believe that the greatest potential for humanoid robots lies in entering households, but B-end is also a necessary stage. Currently, humanoid robots are introduced into industrial scenarios mainly to obtain real scene data, completing technological iteration and upgrades based on that data.
Zhang Lei, a partner at Aixiang Investment, told reporters from the Securities Times that the development trend of robots is irreversible; both robot manufacturers and enterprise buyers are investing in the future of robots, meaning that implementation still requires time. Factories focus on the “solution” of robots rather than the “humanoid” form, such as adding mechanical arms to wheeled or tracked robots to reduce research and development and scene implementation costs with “mature forms + new functions.”
“In fact, whether robots are ‘humanoid’ in factory scenarios is not the key. Currently, there is little dispute about the upper body of robots performing core operational terminal functions; the main debate centers on the lower body, whether the mode of movement is bipedal or wheeled,” an executive from a first-tier investment institution analyzed to reporters from the Securities Times.
The executive pointed out that the advantages of wheeled robots are low cost and high stability, which is superior to bipedal robots, but bipedal humanoid robots have seen a rapid decrease in costs due to increased funding and a maturing industrial chain in recent years. Currently, the costs of some bipedal robots are already close to those of wheeled robots, and the stability of bipedal robots is improving rapidly, leading to more industrial scenarios willing to introduce bipedal robots.
Yao Maoqing, partner, senior vice president, and president of the embodied business department at Zhiyuan Robotics, recently stated in interviews with the Securities Times and other media that wheeled robots and bipedal robots have different positioning in industrial scenarios. Wheeled robots have advantages in factories with flat ground and strong passability, with long endurance, high stability, and low cost; while bipedal humanoid robots are more suitable for environments with poor passability. Currently, the core goal of humanoid robots entering real scenarios is to obtain high-quality real machine data, which is also the key to Zhiyuan’s technological path.
Guo Yandong, founder and CEO of Zhi Square, also publicly stated that the competition in the next three years will be a competition of “real scene closed loops”; the company whose robots can be deployed and work in more real commercial scenarios, continuously feeding operational data back to models and hardware for iterative optimization, will be able to build an insurmountable moat.
Obstacles Still to Overcome
Industry insiders believe that with the current capabilities of humanoid robots, they are still insufficient to handle various general tasks, and there are still many obstacles to overcome in mass production, technology, and scenarios.
Tian Feng, founder and director of the Fast Thinking and Slow Thinking Research Institute, pointed out in an interview with reporters from the Securities Times that there are four major bottlenecks currently faced by domestic humanoid robot manufacturers: quality control gaps, mass production bottlenecks, talent shortages, and ethical gaps.
“In terms of quality control, the large-scale and consistent supply of high-performance core components requires extremely high precision, torque, and lifespan, resulting in a huge quality control gap between small batch trials and large-scale mass production. In terms of mass production, the complexity of system integration and testing poses challenges; the dynamic system composed of thousands of components has not yet matured in terms of calibration, debugging, overall aging testing, and fault diagnosis systems, leading to long delivery cycles and high post-maintenance costs,” Tian Feng pointed out.
The lack of necessary application scenarios is also a common problem facing the current industry. Zhang Lei stated that the common bottleneck in the humanoid robot industry is the insufficient necessity of scenarios, which leads to the inability to achieve a virtuous cycle of “scenarios-sales-cost.” “The insufficient necessity of landing scenarios leads to low product sales, and low sales result in high production costs, which further suppresses purchasing demand; at the same time, the lack of scenarios will lead to a lack of data iteration support in the industry, resulting in slow product iteration speed and an inability to complete the commercial closed loop.”
On the technical level, issues such as insufficient reliability need to be addressed urgently. Recently, Tesla’s Optimus may have to delay the release of the design of the third-generation humanoid robot Optimus until next year due to bottlenecks in hand technology. When UBTECH released the new full-size industrial humanoid robot Walker S2 in July this year, it stated that data, model adaptation, and production line efficiency are bottlenecks for humanoid robots to land in factories.
“Especially in the field of precision assembly, taking the automotive final assembly workshop as an example, the assembly of components and the tightening of screws pose two major challenges for robots: first, the dexterous hand needs to enhance its fine operation capabilities, covering basic skills such as grasping and recognition positioning; second, the production line has strict efficiency requirements, with each workstation on the assembly line having strict time limits, making it extremely difficult for robots to meet speed standards,” UBTECH’s vice president and head of the research institute, Jiao Jichao, admitted at a media communication meeting.
In Tian Feng’s view, there is indeed a “bubble” in the humanoid robot industry, but this is a “normal bubble” on the eve of a technological breakthrough. The manufacturing industry is willing to bet on humanoid robots, which reflects the hard technology capital’s early layout for the future new labor market. “The real risk is not overheating, but that some robot enterprise players may not be able to cross the ‘engineering death valley,'” Tian Feng admitted.
“Current orders mostly belong to ‘pilot verification’ nature, still far from ‘commercial cost-effectiveness.’ We need to understand that the core goal of pilot projects is not to calculate ROI (return on investment), but to jointly define the technical specifications and scenario boundaries of future scalable products,” Tian Feng emphasized.