Have over 40,000 units been shipped, and is the factory ready to pay for the "humanoid"?

From August 19 to 23, the 2026 World Siasun Robot & Automation Conference (WRC) was held in Beijing's Yizhuang, with over 300 companies showcasing more than 2,000 exhibits, including more than 150 new product debuts. This year's conference introduced a "Procurement Day" for the first time, and the theme wording has placed "supply and demand" before "showcasing technology." On the opening day, Yushu Technology (688836.SH) was listed on the Shanghai Stock Exchanges STAR Market, becoming the first humanoid Siasun Robot & Automation stock on the A-share market; in the afternoon, Xiaomi's next-generation bipedal humanoid Siasun Robot & Automation "Iron Big" made its debut at the conference. In the first half of this year, the shipment of humanoid Siasun Robot & Automation in China surpassed 40,000 units, increasing its global market share to 97%. According to IT Juzi's event library, the total financing amount in the domestic embodied intelligence sector reached 93.5 billion yuan in the first half of 2026, a fivefold increase compared to the same period in 2025. As the capital market and industry heat up, a significant industry issue has emerged: Is the market ready to pay for "humanoids"? Reporters from the Financial Associated Press observed that compared to previous years, this year's WRC exhibits are shifting from "eye-catching performances" to "practical applications." The performances of dancing, somersaults, and boxing, which were ubiquitous last year, have drastically decreased, replaced by practical demonstrations aimed at factory production lines, pharmacy sorting, unmanned retail, and educational services. Several companies presented solutions that are already operational at client sites, shifting industry focus from "can it be built" to "can it be effectively used." However, "effectively used" does not mean "well used." Multiple interviewed industry professionals and investors expressed to the Financial Associated Press that humanoid Siasun Robot & Automation still faces numerous challenges in terms of commercialization maturity, form selection for deployment, and cost control. Siasun Robot & Automation "on the job" A PCB board is placed on a tray, and the Siasun Robot & Automation picks it up and delivers it to a designated location. This action seems simple, but in actual production, facing different PCB models, sizes, and colors, as well as complex situations like transparent or reflective trays, the Siasun Robot & Automation must not only accurately identify the target but also perform a series of continuous actions, including tray material handling, loading and unloading, moving, and dual-arm coordination. At the WRC, Zhi Ping Fang demonstrated such an industrial production line application scenario. Similar changes are happening among the users of Siasun Robot & Automation. Zheng Xiaodan, head of JD Retail's embodied intelligence business, told the Financial Associated Press that from last year to this year, numerous practical application scenarios for Siasun Robot & Automation have emerged, including industrial, retail, health, and some semi-commercial cases. However, to accelerate the entry of Siasun Robot & Automation into various industries and households, merely having a good body and model is not enough; it also needs to validate and optimize continuously in real scenarios, Zheng Xiaodan stated. The emphasis among companies on "on-the-job" also signifies a shift in industry evaluation standards. Mengwen, co-founder of the Jump Union and Siasun Robot & Automation Home, provided insight at the 10th session of the WRC's Embodied Era - Smart Creation Forum, noting that the industry has previously been easily attracted to the aesthetics and stage actions of Siasun Robot & Automation. However, entering the second half of 2026, the market is increasingly focused on three other aspects: long-term stable operational reliability of the actual machines, repeat purchase conversion rates from industrial clients, and overall return on investment (ROI). Extending to the upstream of the industry chain, a directly observable change is that the humanoid Siasun Robot & Automation industry is moving towards "miniaturizing" components. Xu Yang, founder of upstream parts company Nuoshi Siasun Robot & Automation, believes that the mechanical parts industry has significant technological accumulation, but the development of humanoid Siasun Robot & Automation has not completely altered the existing mechanical manufacturing system. Compared to traditional applications, humanoid Siasun Robot & Automation demands miniaturization of core components and modules; however, the underlying technology and manufacturing logic have not fundamentally changed. The 1.5mm micro planetary roller screw and a complete set of micro electric cylinder modules (PWG series) showcased at this WRC are intended to adapt to the independent driving of humanoid robot thumbs, joints, and fingers, enabling fine actions like grasping small objects and turning keys. However, having "jobs" does not imply that Siasun Robot & Automation is currently fully competent in those roles. Xu Yang told the Financial Associated Press that many industrial and service scenarios are still primarily completed by humans, and for humanoid Siasun Robot & Automation to genuinely enter these areas, continuous iteration in product capability and operational stability is required. Consequently, as the industry extends upstream, the current demands from complete machine manufacturers concerning upstream parts companies are primarily focused on expanding production capacity and reducing prices, with stringent verification of precision, lifespan, and other metrics yet to be widely adopted. Youwei, chairman and general manager of Efort & Qizhi, also noted that while Siasun Robot & Automation manufacturers have gained a sense of "self-sufficiency," considering various factors such as cycle time, accuracy, and generalization capability is critical for actual productivity. "These three cost-constrained conditions form an impossible triangle; in many cases on-site, one point may stand out, but an overall balance isn't considered, making it difficult to deliver in the eyes of the client." This is also reflected in the downstream demand structure. According to Yushu Technology's response to the Shanghai Stock Exchange inquiry letter, in the first three quarters of 2025, the proportion of humanoid Siasun Robot & Automation revenue from research and education was 73.6%, with commercial consumption at 17.39% and industry applications at 9.01%. Li Fenzhenzi, a researcher at Zhongyan Puhua, told the Financial Associated Press that after excluding self-use iterations by automobile manufacturers and R&D procurement orders from universities and research institutions, the actual external commercial mass production orders for humanoid Siasun Robot & Automation remain low. Currently, most industry shipments are used for technological iteration, algorithm training, and scenario testing, belonging to internal ecological cycles and research investments. Overall, the commercialization of humanoid Siasun Robot & Automation is still in the preliminary verification stage, with genuinely commercial mass production orders being scarce. Another reason lies in the cost. Interviewees generally believe that humanoid Siasun Robot & Automation has not yet formed a clear advantage over humans. Currently, the comprehensive lifecycle cost of humanoid Siasun Robot & Automation remains significantly higher than that of traditional labor, including hardware procurement, scene adaptation, operational maintenance, and unexpected downtime losses, Li Fenzhenzi stated. At this stage, the selling price for humanoid Siasun Robot & Automation is relatively high, core components have not achieved scaled cost reductions, and due to the long adaptation and debugging cycles for industrial scene applications and high operational maintenance costs, only extremely high-load continuous operation scenarios can slightly reduce the cost gap. Two legs or several wheels? As humanoid Siasun Robot & Automation begins to emphasize getting "on-the-job" and deployment, another question has become applicable: What should it look like? Currently, the technological rivalry between bipedal humanoid Siasun Robot & Automation and wheeled chassis dual-arm Siasun Robot & Automation has commenced. In the process of embodied intelligence transitioning to industrial deployment, the choice of humanoid Siasun Robot & Automation form primarily depends on the actual operational demands of the manufacturing site. Currently, scenarios capable of achieving mass production and delivery first are heavily concentrated at the industrial manufacturing and warehousing logistics end. In these structured workshops, mainly composed of flat surfaces, wheeled devices have emerged as a more pragmatic choice. Guangdong Topstar Technology (300607.SZ) leads in light-load industrial Siasun Robot & Automation shipments in China. The company previously explained to the Financial Associated Press that its industrial clients evaluated dozens of Siasun Robot & Automation suppliers and that both companies advancing to the next round of evaluations utilized wheeled chassis. The core reason for this choice lies in safety: in precision manufacturing workshops, individual material values often reach several tens of thousands; should a bipedal Siasun Robot & Automation experience a power outage or system failure, it could fall and cause damage to materials or production lines. In contrast, wheeled devices stop in place during a fault and won't trigger secondary damages. Since last year, there has been news of numerous wheeled dual-arm Siasun Robot & Automation receiving training orders from factories. Models such as Zhiyuan Siasun Robot & Automation A2-W, Galaxy General Galbot S1, and Zhi Ping Fang AlphaBot2 have announced their "deployment" at leading domestic factories like Fulin Precision (300432.SZ), Contemporary Amperex Technology (300750.SZ), and Hkc Corporation (001399.SZ). Dai Fuquan, chairman of Shanshan Venture Capital, told the Financial Associated Press that in standard factories that are flat and well-modified, on most workstations, the cost-effectiveness, stability, and operational maintenance costs of wheeled dual-arms significantly surpass those of bipedal humanoid systems, with faster mobility, lower energy consumption, simpler balance control, and shorter payback periods. Cai Bingzhen, director of the High-Tech Robot Research Institute, also expressed to the Financial Associated Press that wheeled devices are generally recognized in the industry as the first to achieve deployment. Simultaneously, the costs for wheeled humanoid Siasun Robot & Automation are also lower. Haitong's research report notes that due to their structural design already possessing stability, wheeled Siasun Robot & Automation can reduce the use of components and lower model dimensions, consequently decreasing overall manufacturing costs. Under equivalent height, the BOM cost for bipedal designs is ten times that of wheeled chassis. In terms of operational maintenance, bipedal Siasun Robot & Automation requires human assistance more significantly, currently not achieving work efficiency comparable to that of human workers, and many still require tethered support for walking. These necessary safety measures increase operational maintenance costs; meanwhile, wheeled Siasun Robot & Automation, with higher stability, can reduce maintenance and repair costs. The temporary preference for wheeled forms in factories does not imply that bipedal humanoids have lost their rationale for existence. Cai Bingzhen pointed out that the movement control requirements for bipedal Siasun Robot & Automation are higher; however, in unstructured environments, narrow spaces, stairs, and outdoor scenarios, bipedal systems can navigate by adjusting their movement posture, providing unique advantages that are incomparable. Dai Fuquan stated: "The core value of bipedal Siasun Robot & Automation lies in adapting to the physical environments shaped by long-term human modifications, especially in older factories with stairs, thresholds, and cable trenches, as well as specific inspection scenarios in power and chemical industries, without requiring costly modifications to the sites surface infrastructure. Bipedal designs can reuse existing human pathways and operational heights." Interviewees generally believe that although wheeled designs offer quicker deployment now, the two robot forms will present a complementary relationship in the future. "Wheeled humanoid Siasun Robot & Automation suits flat and hardened surfaces, while bipedal humanoid Siasun Robot & Automation can accommodate uneven terrains, filling the gaps in complex industrial tasks; the two forms complement each other in their respective scenarios," Li Fenzhenzi noted. The replaceability and versatility of bipedal configurations at workstations also motivate some large manufacturers to continue investing in research and development. Recently, Xiaomi mentioned in an interview with the Financial Associated Press that if only addressing two or three workstation issues in a vertical, singular scenario, humanoids are not significantly important; however, in the long run, companies hope to explore a universally applicable operational method for multiple workstations, rather than customizing automated devices for each individual workstation. Currently, training for their full-sized bipedal Siasun Robot & Automation production line is ongoing, but it's still facing efficiency limitations in lengthy task chains. Xiaomi shared data regarding their new generation bipedal humanoid Siasun Robot & Automation "Iron Big" showcased at this year's WRC, revealing it stands 1.70 meters tall, weighs 66 kilograms, and possesses 66 degrees of freedom, achieving a success rate from 90% in March to 98% in July at the self-tapping nut workstation on the automotive production line. However, regarding operational efficiency, it operates at 70% to 80% of human efficiency for short tasks of 3 to 5 seconds; once entering longer task chains exceeding 1 minute involving grasping, moving, searching, and judging, the efficiency declines to around 30% of human output, necessitating ongoing optimization of model inference efficiency and long-tail conditions. Underlying technology attracts 80% of investments 2025 has long been viewed as the year of an explosion in humanoid Siasun Robot & Automation. This year, humanoid Siasun Robot & Automation started to enter the public eye on a large scale while also drawing considerable investment. According to Quantum Bit Think Tank data, in 2025, the total number of investment events in the embodied intelligence sector surged from 173 in 2024 to 447, with total capital inflow jumping from 13.7 billion to 55.4 billion yuan, corresponding to increases of over 250% and 400%, respectively. This year, compared to the previous year's fervor, more skepticism has been heard within the embodied intelligence realm, with ongoing debates regarding deployment scenarios and forms raising questions on whether investment enthusiasm is cooling. This is not the case. IT Juzi data indicates that in the first half of 2026, the domestic embodied intelligence sector saw 322 financing rounds, totaling 93.5 billion yuan, a fivefold surge compared to the same period in 2025, exceeding 57 billion yuan for the entirety of 2025 within just six months. Dai Fuquan remarked: "I believe the overall investment and financing enthusiasm has not cooled down, but the investment logic has undergone a fundamental shiftfrom two years agos 'seeing prototypes and betting on overall stories' to a more rational focus on 'heavy orders, realizing mass production, and core supply chains. "Two years ago, obtaining financing was possible as long as a walking prototype could be demonstrated; by the second half of 2026, capital is unwilling to pay solely for demonstration outcomes. The critical criteria for pre-investment now involve obtaining small-batch commercial orders, whether yield rates can sustain improvement, and if prices can be continually lowered. Leading firms are still able to secure large capital from industry and local state-owned assets, whereas financing for many projects without real-world deployment capabilities has become significantly more challenging, resulting in notable differentiation within the industry," Dai Fuquan stated. Regarding the funding direction changes in the Siasun Robot & Automation industry, Xu Yang noted that with primary enterprises having secured substantial capital investments and leading company valuations rising, some investment firms are shifting their focus from complete machines to upstream core components. Currently, the production levels of the three core components of humanoid Siasun Robot & Automation display a differentiated pattern, with varying degrees of domestication and yield rates. Li Fenzhenzi informed the press that domestic harmonic reducers and multi-degree-of-freedom dexterous hands are relatively mature, but internal micro hollow cup motors and high-precision sensing components still rely heavily on imports, with high-end dexterous hands' core parts also being mostly imported. Furthermore, Li Fenzhenzi emphasized that micro screws represent a bottleneck in the supply chain for humanoid Siasun Robot & Automation, with a domestic rate of less than 20% for high-precision planetary roller screws, low rates of mass production yields, and high-end markets largely dominated by overseas manufacturers. Domestic products can only accommodate low- to mid-end scenarios, resulting in overall high dependency on imported supply chains, which is a critical bottleneck for reducing costs and achieving mass production of complete machines. In this context, Dai Fuquan analyzed several key investment directions currently in focus: first, the domestic substitution trajectory of core components remains a primary focus, particularly on dexterous hands, flexible tactile sensors, integrated joints, and six-dimensional force sensorsthese are essential and receiving ongoing attention; second, embodied intelligence algorithms with world models and Siasun Robot & Automation data infrastructures, which act as the "brain" and training base of Siasun Robot & Automation, determine whether Siasun Robot & Automation can progress from "being active" to "working autonomously"; third, vertical solution providers are not discussing general Siasun Robot & Automation but are dedicated to specific scenarios such as factory settings and hazardous inspections, with a focus on calculable ROI. The "2026 Global Investment Compass for Embodied Intelligence Industry" white paper released by the Jump Union indicates that in the first half of 2026, 84% of the collected funding in the entire sector went to core foundational tracks, including embodied intelligence models (world models/VLA integrated algorithms), standardized physical interaction data, and dexterous tactile execution hardware, while financing for humanoid complete machines dropped from 31% in 2025 to 12.8%. Following significant funding, the next step is large-scale deployment. Cai Bingzhen informed that the scaled deployment of humanoid Siasun Robot & Automation is expected to progress in three phases: the relatively short-term first stage will be primarily applied to weak-process and high-tolerance scenarios such as inspection, loading and unloading, and material handling, expected to achieve scaled deployment within three years; the mid-term phase will apply to sorting of logistics parcels, which, while still somewhat tolerant of errors, requires higher generalization for the handling of non-standard materials; the third phase, the long-term stage, will apply to industrial manufacturing scenarios that demand strict precision, pacing, and ROI. "The current core gaps hindering deployment are manifested in task success rates, continuous operational stability, and consistency in body hardware. The deep-seated technology bottleneck lies in the fact that the large models of embodied intelligence have yet to fully converge, coupled with extreme scarcity of training data. The industry broadly believes that a million hours of real scene data is necessary for models to reach a stable state, with data collection bottlenecks resulting in a surge of data collection devices this year," Cai Bingzhen stated. This article is reproduced from the "Financial Associated Press" WeChat official account, by Wang Biwei and Zhang Xiaoyu; GMTEight editor: Song Zhiying.