Woofun AI reports that XianGong Intelligence has officially listed on the HKEX, marking a pivotal moment for the robot brain sector with shares surging over 38% on the first trading day. The public offering was oversubscribed nearly 6,000 times, resulting in a mere 5% success rate for individual investors, while eight cornerstone investors including Hillhouse Capital and Yuanbao Family Office committed a total of HK$462 million. This overwhelming capital support for a six-year-old company with annual revenue of only RMB 442 million and ongoing losses underscores the market's intense valuation of its core asset: the robot controller. XianGong Intelligence commands a 24.8% global market share and a dominant 45.2% share in China for robot controllers, ranking first in both categories. The listing validates a broader industry trend where the focus is rapidly shifting from the physical 'body' of the robot to its intelligent 'brain.'

The timing of this listing coincides with a fundamental restructuring of the embodied intelligence sector. For the past two years, industry attention was fixated on the physical attributes of robots, prioritizing stable designs, agile movements, and human-like appearances.

However, the trajectory shifted decisively in 2026 as capital began flowing directly toward the development of intelligent systems rather than mechanical structures. Statistics reveal that in the first half of 2026, the domestic embodied intelligence sector attracted approximately RMB 43.8 billion in financing. Of this total, more than half was directed toward companies developing robot brains, while those manufacturing physical robots received less than 20%. This disparity indicates a clear migration of industrial value as the sector transitions from concept validation to large-scale implementation. As hardware supply chains stabilize and mature, the differentiating factor for robot performance is no longer the mechanical structure but the sophistication of the intelligent system controlling it.

A retrospective look at 2024 highlights the speed of this transformation. During that period, the sector's hottest entities were manufacturers of complete humanoid robots. UBTECH listed on the HKEX, while companies such as Zhiyuan, Yushu, and Fourier launched new products that captured the tech community's imagination. The prevailing narrative then was that embodied intelligence represented the next generation of computing platforms, with humanoid robots viewed as the ultimate form, and mass production capability seen as the primary competitive advantage. Within just over a year, this capital allocation logic inverted completely. The financing patterns of the first half of 2026 demonstrate that companies focused on the 'brain' captured nearly 70% of total funding, marginalizing those specializing solely in hardware bodies. The velocity of investment also accelerated dramatically; many brain-focused companies completed financing rounds every month, with the fastest securing two rounds within a mere two weeks. Such rapid capital deployment is atypical for the hard-tech sector and mirrors the dynamics previously seen only in internet and software industries.

The driver behind this capital reallocation is the rapid erosion of technical barriers in robot body manufacturing. Yushu Technology recently announced that its dual-legged humanoid robot, the Unitree R1, is now available for spot purchase at a reduced price of RMB 29,900. Previously, humanoid robots were considered premium products costing hundreds of thousands of yuan. This dramatic price reduction stems from the maturation of the supply chain and the onset of mass production, which have significantly lowered the costs of core components such as joints, motors, and reducers, showcasing the efficiency of Chinese manufacturing. When the 'body' ceases to be a scarce resource, the intrinsic value of the 'brain' becomes the primary determinant of market success. The logic is straightforward: identical robot hardware equipped with different controllers can perform vastly different tasks. Low-end controllers restrict robots to repetitive tasks along preset routes, whereas high-end intelligent controllers enable environmental perception, autonomous path planning, multi-machine coordination, and natural language understanding via large models. The former sells hardware, while the latter sells functionality.

This value distinction is starkly reflected in gross profit margins. XianGong Intelligence's prospectus discloses that its controller business achieves a gross profit margin of 79.8%, while its software business reaches an even higher 89.3%. In contrast, the gross profit margin for its robot hardware business stands at only 38.4%, and the accessory business lags further at 15.7%. A gross profit margin nearing 80% is exceptionally rare in the robot hardware industry and aligns more closely with the profitability profiles of pure software companies. This margin differential explains the capital market's eagerness to invest in brain developers. Compared to hardware bodies, which demand heavy capital expenditure, offer low margins, and entail long development cycles, controllers and intelligent systems benefit from lower marginal costs and stronger scale effects. Once an ecological barrier is established, the profit potential for these software-centric entities becomes substantial.

The definition of the 'robot brain' has also expanded significantly in the current context. Historically, robot controllers were essentially motion control boards managing motor movements and joint coordination, with technical challenges centered on real-time performance and stability. Today, robot brains have evolved into complex intelligent systems integrating perception, decision-making, and control. These systems incorporate technologies such as SLAM positioning and navigation, visual semantic recognition, reinforcement learning, multi-machine scheduling, and integration with large language models and world models. XianGong Intelligence's prospectus indicates that the global supply of independent robot controllers grew from 6,000 units in 2021 to nearly 50,000 units in 2025, with projections to exceed 300,000 units by 2030. Revenue-wise, the global robot controller market expanded from RMB 700 million in 2021 to RMB 2.4 billion in 2025, and is expected to reach RMB 8.4 billion by 2030, representing a compound annual growth rate of 28.8% between 2026 and 2030. Industry reports further suggest that in 2025, the market for robot brain controllers will reach RMB 2.236 billion, while the market for robot motion control systems will hit RMB 6.073 billion. Including new-generation AI-enhanced intelligent controllers, along with accompanying software, algorithms, and cloud services, the market potential expands even further. A report by Future Market Inc. predicts the global physical AI market will grow from approximately US$383 billion in 2026 to US$3.26 trillion by 2040. Physical AI fundamentally involves equipping physical entities with intelligent brains, with robots serving as the most prominent example.

While startup financing enthusiasm signals internal industry momentum, the entry of major internet and cloud service providers indicates that competition has escalated to an ecological level. This year, Huawei, Tencent, Baidu, and Alibaba have all launched embodied intelligence platform products, uniformly choosing to focus on developing 'brains' rather than manufacturing physical robots. These giants provide intelligent systems, development tools, and cloud infrastructure to robot manufacturers. Huawei has adopted the most aggressive and comprehensive approach. At the Huawei Cloud INSPIRE Creator Conference this month, Huawei Cloud officially launched the CloudRobo embodied intelligence development platform, described as 'the world's first end-to-end development platform for embodied intelligence.' The platform covers the entire lifecycle from data synthesis to model development, simulation verification, and cloud-edge deployment. It includes millions of data assets and more than 20 Ascend-compatible models, enabling robot deployment in hours and model deployment in minutes. Huawei's strategy is clear: leverage cloud power to lower barriers to robot intelligence. Previously, manufacturers had to independently develop algorithms, train models, and build simulation environments, a costly and time-consuming process. With CloudRobo, everything from data collection to training to deployment is streamlined, allowing manufacturers to focus on high-quality hardware production and scenario implementation.

On the launch day, over 20 companies, including Youai Zhige and Huayuan Robotics, announced their adoption of the CloudRobo platform. Huawei also introduced the 'Hundred Models, Thousand Forms: Win-Win Through Cloud Collaboration' ecological initiative. Leveraging its strengths in Ascend chips, industrial software, and 5G networks, CloudRobo is effectively building an 'end-edge-cloud' collaborative infrastructure for embodied intelligence. Tencent opted for a more modest entry strategy. For the first time, the Tencent Robotics X Laboratory showcased a complete technology suite and launched the Tairos embodied intelligence open platform. It also open-sourced the Hy-Embodied series of mixed-element embodied large models and demonstrated RoboFusion technology for interconnecting robot bodies. Zhu Yajuan, head of the Tairos product ecosystem at Tencent Robotics X, stated that Tencent positions itself as an 'indispensable component' in the robot industry, focusing on software and cloud services rather than hardware manufacturing. This statement acknowledges Tencent's limitations in hardware production while emphasizing its core value at the system level. Specifically, the Hy-Embodied large models address the robots' ability to 'understand and think,' enabling environmental perception, instruction comprehension, and autonomous task planning. The Tairos platform improves development efficiency through standardized toolchains, while RoboFusion solves interconnectivity issues between different robots. Together, these three elements form Tencent's comprehensive solution for robot brains.

Baidu follows a 'data + model + infrastructure' approach. At the Create 2026 Baidu AI Developer Conference in May, Baidu Smart Cloud announced increased investment in AI infrastructure, scenario integration, and industry standard establishment. In April of the same year, Baidu collaborated with several robot companies to launch the 'Embodied Intelligence Data Supermarket,' establishing a hierarchical data labeling system. Baidu has also made significant financial commitments, participating in the RMB 1 billion Series B financing for Zhi Ping Fang and the RMB 700 million Series A financing for the Beijing Humanoid Robot Innovation Center. Baidu's strategy utilizes the capabilities of its Wenxin large models to connect with hardware companies, creating a closed AI loop spanning the virtual and physical worlds. In February of this year, Alibaba DAMO Academy released the RynnBrain embodied basic model and open-sourced seven models, including the industry's first 30B MoE architecture embodied model. Official data indicates this model enables robots to possess temporal and spatial memory and spatial reasoning abilities, setting new benchmarks in 16 embodied intelligence evaluation metrics and surpassing Google's Gemini Robotics ER 1.5. Although ByteDance has not launched a dedicated platform, it listed the world model as its top priority for 2026. At the Volcano Engine FORCE conference on June 23, ByteDance noted that its Seedance video generation model could be utilized in the data synthesis phase of embodied intelligence. Given ByteDance's expertise in multimodal large models and video generation, its entry into robot brain development is considered inevitable.

The collective shift of major companies toward 'brains' reflects practical economic considerations. Hardware bodies are tangible but require significant investment, offer low margins, and involve complex supply chain management, often leading to price wars similar to the smartphone industry. In contrast, while initial development costs for controllers and platforms are high, marginal costs decrease rapidly once an ecosystem is established, generating continuous revenue through cloud services. Cloud service providers possess natural advantages here, as training embodied intelligence requires massive hash rates, large-scale simulation environments, and multimodal large models, all of which are core competencies of cloud giants. It is neither economical nor practical for hardware manufacturers to build their own AI infrastructure from scratch. Consequently, the future structure of the embodied intelligence industry is likely to mirror the smartphone industry: chips and hash rate at the bottom, operating systems and intelligent brains in the middle, and various robot hardware forms at the top. Cloud service providers are fiercely competing for this critical middle layer.

Despite the compelling narrative, business realities present significant challenges. XianGong Intelligence's prospectus offers critical insights. While the high gross profit margins of its controller and software businesses demonstrate commercial value, the controller business generated only RMB 85 million in revenue in 2025, accounting for 19.3% of total revenue. Conversely, the hardware robot business, with lower margins, generated RMB 300 million, accounting for 67.9% of revenue. This reveals a paradox where the most profitable segment is the smallest in scale, while the revenue-driving segment yields lower profits. This is a common challenge for third-party controller manufacturers. Currently, robot manufacturers, particularly in the industrial sector, prefer purchasing complete robot solutions over individual controllers. Complete robots offer immediate delivery and usage, eliminating the need for customer integration.

Furthermore, as core components, controllers require significant secondary development capabilities, creating a high barrier for traditional manufacturing customers. Consequently, controller manufacturers often must sell products as part of complete solutions, integrating controllers into their own products before reaching end-users. The 'brain' is essential but must be combined with the 'body' to realize value. Xu Zhaoyun, a partner at Lihang Investment, noted that domestic humanoid robot manufacturers are currently scaling mass production with a primary goal of controlling hardware costs. This forces third-party controller manufacturers to compete on price, discounting software value due to hardware constraints. This creates an awkward dynamic where high-barrier, high-margin controllers are difficult to sell separately, while lower-barrier, lower-margin complete robots become the primary revenue source.

XianGong Intelligence has not escaped this pattern. Between 2023 and 2025, the company's revenue grew from RMB 249 million to RMB 442 million, a compound annual growth rate of 33.2%.

However, net losses during these three years were RMB 47.7 million, RMB 42.3 million, and RMB 47.07 million respectively, totaling a cumulative loss of approximately RMB 137 million. Although the adjusted net loss decreased from RMB 20.91 million in 2023 to RMB 2.87 million in 2025, the company remains far from profitability. This is not unique to XianGong Intelligence but a common phenomenon in the industry's early stages. Customers currently prefer paying for tangible hardware over software, algorithms, or intelligent systems. Similar to the early smartphone era, consumers initially paid for hardware, and the value of software and services only became apparent after the mobile internet ecosystem matured.

Another challenge arises from the self-development efforts of major companies. As Huawei, Tencent, and Baidu launch their own platforms, the survival of independent third-party controller manufacturers faces uncertainty. Currently, the roles appear distinct: cloud platforms focus on general large models, training tools, and cloud scheduling, while companies like XianGong Intelligence focus on low-level real-time motion control and on-site execution. These functions are complementary rather than competitive.

However, long-term boundaries may blur. Internationally, the four major traditional industrial robot manufacturers—Fanuc, ABB, Yaskawa, and Kuka—all integrate controllers and hardware bodies, keeping core technologies in-house to maintain high profit margins. Domestic manufacturers aiming for the high-end market will likely follow suit by developing their own controllers. Therefore, the real opportunity for independent third-party controller manufacturers may lie in serving a large number of small and medium-sized enterprises and addressing niche applications. XianGong Intelligence's 'Nebula' open system includes more than 1,000 robot models, covering forklifts, material handling robots, wheeled humanoid robots, and robotic dogs. Customers can customize these models modularly. This 'Lego-style' approach appeals to small and medium-sized enterprises needing quick deployment without starting from scratch. By the end of 2025, XianGong Intelligence had served more than 2,000 customers across 35 countries and regions. Establishing a broad customer base and industry coverage is essential for independent manufacturers to build ecological barriers.

XianGong Intelligence's listing is a landmark event, signaling that the capital market recognizes the independent value of robot brains and that industrial division of labor is becoming more refined. Industry discussions have shifted from whether robots can perform basic tasks like walking or grasping to how intelligent they can be. The second half of this development journey will no longer be defined by sophisticated mechanical structures but by the intelligence of brains, the openness of systems, and the vibrancy of ecosystems. Capital is flowing to robot brains, major companies are investing in platforms, and startups are deepening application scenarios. In the short term, hardware bodies will remain the primary revenue source, with robot brain value reflected in technical barriers and competitive advantages.

However, in the long run, as robot hardware becomes homogeneous, intelligent systems will become the true centers of profit and the key to establishing competitive ecosystems. It is still too early to declare a final outcome, as embodied intelligence remains in its early stages. only 3,000 general-purpose embodied intelligence robots were shipped globally in 2024, with expectations to reach 2.6 million by 2035. There is a long road ahead, but the direction is clear: in the second half of embodied intelligence development, it will be the 'brains' that determine success.