The robot marathon and other events have been held one after another, and multiple companies are accelerating the layout of humanoid robots. Related products are constantly expanding their trial scenarios... Currently, a global transformation triggered by the rise of humanoid robots is taking place. According to research conducted by reporters, the global humanoid robot industry is currently in the early stages of development. China has certain advantages in scale manufacturing, supply chain capabilities, and application scenarios. In the future, it will need to break through bottlenecks in key technologies, data collection, software ecology, and other aspects to further promote the accelerated landing of the industry. Multiple parties are rushing into the scene to accelerate the expansion of new industries. Now is a period of capital frenzy, and everyone is rushing in this direction. ”Cao Zheng, the Managing Director of Guotou Investment Management Co., Ltd., which conducted intensive research on several leading enterprises, described the current situation of the humanoid robot race as follows. Building a machine that resembles humans has been a long-standing aspiration of humanity. Since this year, from the popular video of Yushu Technology Robot performing Yangko at the Spring Festival Gala of the Year of the Snake, to the popular video of humanoid robot showing front somersault, back somersault, bicycle riding and other stunts on social media, to the "man machine running together" at the world's first humanoid robot half marathon, Chinese humanoid robots are accelerating their iterative evolution, and their popularity is rising. At present, both local governments and enterprises are accelerating the layout of embodied intelligence and humanoid robot tracks. Local governments such as Beijing, Shanghai, Shenzhen, and Chongqing have successively issued relevant plans or established and prepared to establish industrial funds. According to the latest data from Tianyancha Professional Edition, as of May 2025, there are nearly 900000 "robot" related enterprises in China. Among them, from January to May 2025, there will be 100000 new related enterprises added. Compared to the same period in 2024, it has increased by 44%. In the investment industry, humanoid robot industry chain enterprises are also highly favored. According to the latest data from IT Orange, the global investment in humanoid robots has reached 18.651 billion yuan since 2025, surpassing the 12.952 billion yuan for the entire year of 2024. Since 2025, the investment amount related to humanoid robots in China has reached 6.106 billion yuan, also exceeding the 5.893 billion yuan for the whole year of 2024. State owned investment funds are an important way for China National Investment Corporation to fulfill its mission as a state-owned capital investment company, managing a national fund scale of over 230 billion yuan, which is nearly six times the amplification factor of the central government. ”Cao Zheng introduced that Guotou Investment Promotion is an important member of Guotou Fund, which has long been deeply involved in the robotics industry. It has systematically laid out core components and invested in key common products such as reducers, encoders, planetary roller screws, and artificial intelligence chips, breaking through bottleneck links and promoting the realization of independent and controllable upstream industrial chains and cost reduction and efficiency improvement of humanoid robots. With the support of multiple parties, the commercialization of humanoid robots is accelerating, and industrial scenes are regarded as the "first stop". Guangdong Tuosida Technology Co., Ltd. (hereinafter referred to as Tuosida) has launched a new generation of intelligent robot control platform that integrates "sensing computing control" with motion control as the entry point, and has achieved industrial application. Following the release of the intelligent robot universal technology base and the humanoid robot new product YobotW1 at the end of February this year, not long ago, Evertech Intelligent Equipment Co., Ltd. (hereinafter referred to as Evertech) signed a memorandum of cooperation with Huawei Cloud to deepen cooperation in the field of embodied intelligence and jointly explore a new ecosystem of embodied intelligence. Yu Wei, Chairman and General Manager of Evertech, stated that industrial scenarios are relatively structured and suitable for verifying the basic capabilities of robots. Especially with the changing market demand, the demand for customized large-scale products has become increasingly strong, and traditional industrial robots often find it difficult to adapt to a wider range of flexible production needs. With the support of large models, humanoid robots are expected to complete more complex and diverse factory production tasks. The "Research Report on the Development of Humanoid Robot Industry (2024)" released by the China Academy of Information and Communications Technology predicts that from 2040 to 2045, China's versatile humanoid robots will achieve large-scale applications in industrial and service scenarios, and the overall market size will reach about 500 billion to 1 trillion yuan. Industry insiders believe that although humanoid robots have broad prospects for application, they are still in the small-scale experimental stage in the industrial field, and it is still too early to truly enter households. The bottlenecks faced in promoting large-scale applications are quite prominent. Firstly, the technological roadmap has not yet converged, and the core components urgently need to be improved. Cao Zheng introduced that the technical route of China's humanoid robot industry hardware has not yet begun to converge, and the existing products have different forms, resulting in high costs. Meanwhile, the significant differences in hardware form and sensing schemes also limit the data collection and training of humanoid robot brains. The humanoid robot industry chain is highly complex, and there are still many core technologies that urgently need to be improved. Taking sensors, especially tactile sensors, as an example, "current humanoid robots mainly rely on visual and auditory solutions to achieve interaction with the real world, but touch, as an important sensory of physical contact, is also the key to achieving large-scale applications of humanoid robots. The existing solutions are difficult to reproduce the perception ability of human hand fine operations, and the high cost of six axis force sensors also hinders large-scale applications. The flexibility and durability of dexterous hands still need to be further improved," said Mu Yao, an assistant professor at the Artificial Intelligence Research Institute of Shanghai Jiao Tong University. Jiang Zhou, Vice President of Sales for Sweet Potato Robots, admitted that the mobility of humanoid robots still needs to be improved, and behind it is a game of control algorithms, material processes, and production costs. The cost advantage of China's industrial chain is irreplaceable, but it requires collaborative optimization across the entire industry rather than relying on a single enterprise. Secondly, there is a lack of high-quality training data, and the collection cost remains high. At present, humanoid robots mainly demonstrate motion control capabilities, which can only solve their mobility problems and cannot solve the problem of helping people with work. ”You Wei explained that the key to the practical application of humanoid robots lies in their autonomous intelligence, which requires massive amounts of high-quality data for training and learning. Unlike autonomous driving, which naturally accumulates data through daily driving, humanoid robots rely on manual operation or simulated environments to collect data. In home scenarios, users often find it difficult to accept the "learn while using" model, while in industrial scenarios, data collection relies on high-precision devices, resulting in high costs. Cao Zheng said that using virtual simulation scenes to collect data is relatively low-cost, but it can only be used for pre training scenarios of humanoid robots. To further implement the application, post training must be conducted in more real-world scenarios, and skilled operators are needed to obtain high-quality data, which greatly increases the cost of data collection. In addition, the interviewee pointed out that China has the most complete manufacturing system in the world and has an advantage in accumulating data in industrial scenarios, but the underlying ecology is weak and the development of algorithm ecology needs to be strengthened. Several industry experts have pointed out that China needs to accelerate the filling of the gaps in underlying technologies, leverage the advantages of scenarios and industrial chains, and promote the construction of independent ecosystems in stages. One is to jointly promote the research and development of key common technologies. Faced with the problem that the domestic humanoid robot industry has not yet formed a complete collaborative innovation system, it is necessary to establish cross departmental and cross industry cooperation mechanisms, promote the research and development of key common technologies, and open source the results to the industry, so as to connect the key technical links of the industry, "said Wei Jiaxing, the brand public relations manager of Beijing Humanoid Robot Innovation Center. You Wei also suggests integrating industry forces to strengthen the foundation of intelligent robot technology and common technology platforms, and forming an independent ecosystem as soon as possible. The universality of a general-purpose robot can be achieved by stacking specialized skill packages one by one. In Jiang Zhou's view, how to compress deep learning algorithms to the robot end side powered solely by batteries and achieve energy efficiency ratios close to the human brain is also the key to technological breakthroughs. The second is to explore the advantages of scenarios and accelerate the accumulation of data. Zhang Xiaohui, Chief Engineer of Tuosida, believes that technological maturity and price decline are the key factors for industrial explosion. Commercialization needs to be scenario driven, accelerating technological iteration through data feedback on the application side. To address the issue of insufficient data, it is recommended to start from two aspects: on the one hand, improve it by building real-life training scenarios through multi-party cooperation. Mu Yao believes that if a single enterprise or university is unable to afford the cost of building large-scale training scenarios, the government can take the lead in organizing multiple universities and enterprises to jointly build a centralized training platform with multiple scenarios, thereby diluting costs and improving efficiency. On the other hand, we can quickly explore the advantages of Chinese application scenarios and accumulate data on the application side. You Wei proposed a "closed domain" development strategy, which reduces data collection complexity by segmenting scenarios such as stir frying and welding. Cao Zheng stated that the focus of the development of humanoid robots in the future is to enhance the intelligence level of "big and small brains", relying on the combination of advanced model algorithms and massive data in actual manufacturing scenarios. On the one hand, SDIC will pay close attention to the technological progress of the industry in model algorithm, data collection and generation, and on the other hand, give full play to its industrial ecology in the manufacturing industry, promote robot enterprises to fully cooperate with downstream manufacturing customers, promote technological innovation with the actual scene demand as the traction, and accelerate the industrialization of humanoid robots in China. The third is to increase support based on the "first set" policy. Cao Zheng, You Wei and others suggest that the "first set" policy can be referenced to provide subsidies for early application scenarios, thereby reducing the cost of trial and error for enterprises, accelerating technological iteration and scenario verification. At the same time, while increasing funding support, the government is taking the lead in formulating safety standards and other regulations, deeply guiding humanoid robots into application scenarios such as public elderly care service institutions. The outbreak of embodied intelligence cannot be separated from long-term investment in policy support, industry chain collaboration, and talent cultivation. ”Mu Yao believes that the humanoid robot industry should not be overly hyped up, and technological applications need to be breakthrough in stages. Areas with moderate task complexity and high human participation risk, such as high-risk environment operations in nuclear power plants and flexible production line sorting, may become breakthroughs for the large-scale application of humanoid robots within three years. However, technological breakthroughs are only the starting point, and the co construction of industry ecology is more crucial