Editor's Note: The embodied intelligent robot industry is now at a critical intersection of technological iteration and large-scale deployment, accelerating its integration into various industries. The entire sector is swiftly transitioning from the "technology validation phase" to the "commercial popularization phase," with the outline of a new round of growth becoming increasingly clear. Meanwhile, more practical propositions are placed before every industry practitioner: where should technological breakthroughs be made, where should the market be deeply cultivated, and how should the ecosystem be co-built... In light of this, on the occasion of the 2026 World Robot Conference, China Electronics News invites representatives from multiple robot enterprises to focus on core topics such as technological evolution, market deployment, ecosystem synergy, and future trends. By gathering voices from the front lines of the industry, we aim to leave a vivid footnote for the next chapter of the robot industry.
Looking ahead to the next three to five years, what do you consider to be the most commercially valuable and feasible application scenarios for embodied intelligent robots? What is the basis for your judgment?
Liu Jiang, Party Secretary and Chairman of Changhong Holding Group From an industry perspective, the scenario of embodied intelligence entering households is evolving rapidly with broad prospects. However, the home is an unstructured environment. Moving from "capable of demonstration" to "capable of doing real work" still has a long way to go. It is imperative to solve the reliability issues of perception, interaction, and manipulation in complex environments. Relying solely on AI algorithm iteration is insufficient; extensive real-scenario training and data accumulation are essential.
Changhong has been deeply engaged in the home appliance industry. We have been actively exploring and innovating in bringing AI into homes. By equipping devices with multimodal sensors and large models, we have acquired proactive service capabilities. Changhong's AI agents now cover over 5 million home appliances across all categories, including TVs, refrigerators, and air conditioners. They can recognize over 202 dialects with an understanding accuracy rate exceeding 98%, and the response time has been reduced to under 1.5 seconds. The shift from "passively executing commands" to "proactively perceiving and serving" represents the industry consensus on the first step of bringing AI into homes and continuous innovation. Last year, Changhong took the lead in building the first provincial-level embodied intelligent robot training ground in western China. It replicates real home environments such as living rooms and bedrooms on a 1:1 scale, allowing robots to repeatedly practice high-frequency scenarios like serving tea, doing laundry, and organizing refrigerators. Leveraging the labor cost advantage in the southwest region, we have accumulated 400,000 training data entries, built VLA (Vision-Language-Action) and large language model testing capabilities, and explored body-free data collection solutions, enabling robots to transition from the demonstration stage to providing tangible services for households.
On the one hand, this training ground allows the algorithms and real home data accumulated to be directly used for iterating Changhong robots and whole-home smart appliances, enabling devices to proactively identify people and coordinate household chores, thereby truly promoting the transformation goal of "proactive perception" for AI home appliances. On the other hand, it actively outputs standardized training solutions to the industry, addressing the pain point of different brands of robots "training in isolation" with non-interoperable data. Through these explorations and practices, we strive to bring robots into ordinary homes faster and safer.
Xiao Xiangjiang, Chairman of SuperNeng Robot Over the next three to five years, commercialization will penetrate along the path of "industry first, then commercial use, and finally households; specialized first, then general-purpose." The most valuable scenarios lie in industrial verticals, smart elderly care, and commercial services. The basis is threefold: first, rigid demand and clear willingness to pay, such as the rigid need for health management brought by aging, with a clear commercial closed loop; second, commercial scenarios serve as an ideal "transition zone" with high interaction requirements but controllable environments; third, high compatibility with current technological levels and a massive market base. As for SuperNeng itself, elderly care is a new track we have opened up; for humanoid robots, we focus on domestic substitution and practicality, independently developing five models a year; for non-humanoid robots, we turn scenarios that "can truly land and sell" into cash cows, launching over 40 models in seven years with cumulative sales of tens of thousands of units. Although the household market is the ultimate blue ocean, its popularization will take time due to constraints in cost, reliability, and return on investment.
Gao Jiyang, Founder and CEO of Xinghaitu Xinghaitu's judgment is very clear: expanding layer by layer from the inside out along the main line of productivity.
The most core and first-to-materialize scenarios are industrial manufacturing—factories, warehouses, and express delivery sorting. These scenarios have three distinct characteristics: relatively fixed environments, definable task boundaries, and directly calculable input-output, making them the most ideal "first battlefield" for embodied intelligence to move from technology validation to batch deployment. Xinghaitu has cumulatively completed batch delivery verification for orders of thousands of units. Currently, we focus on five core vertical scenarios: transport and movement, pick-and-place, packaging and wrapping, fabric folding, and equipment linking. Each represents a high-frequency demand that truly exists on industrial production lines and has long lacked automation solutions.
Expanding outward, we enter the scenarios of precision manufacturing and high-precision sorting with millimeter-level accuracy—requiring higher precision but yielding correspondingly greater commercial value.
Further outward are unstructured scenarios such as agriculture, construction, and households. These environments are complex and highly variable, requiring a significant leap in generalization capabilities. However, we believe that once the model's generalization capability crosses the threshold, this layer will release market space far exceeding that of industrial scenarios.
The outermost layer, and the ultimate form we have long pursued, is enabling productivity to replicate itself. Once a set of capabilities is validated and proven in one industry, it can be transferred to another with appropriate adjustments, forming horizontal reuse of capabilities. This is the ultimate manifestation of the "one brain, multiple forms" strategy at the commercial level: a universal brain with forms adapted to specific needs.
Regarding the pace of commercialization, I have a clear judgment: competing for first or second place at the stage of complete machine sales is of little significance, as hardware sales are merely the starting point. What Xinghaitu truly pursues is the second stage of "intelligence-driven commercialization"—from complete machine sales to solution subscriptions, and then to token sales. Each layer represents a deeper binding with customers and corresponds to progressively expanding market space. When your customers no longer just pay for "buying a machine," but for "continuously evolving intelligent capabilities," the commercial scale at that time will be what the embodied intelligence industry should truly look like.
Dr. Xu Chunshan, Founder and Chairman of Lingqiao Intelligent Over the next three to five years, the scenarios with the most commercial value and deployment feasibility are concentrated in three major fields: industrial manufacturing, energy and power, and educational training.
Industrial manufacturing: particularly flexible assembly and precision inspection scenarios in industries such as automotive precision parts and 3C electronics. These scenarios have an urgent demand for automation and flexibility, and the tasks are relatively highly structured, making it easiest to calculate the "economic account" for fault tolerance and return on investment. In collaboration with Shanghai Electric, we have achieved fully unmanned processes in a gear precision inspection project under Shanghai Electric and obtained certification from the United Nations Industrial Development Organization.
Energy and power: high-risk manual operations such as high-voltage transmission line inspection and live-line fine maintenance. This is a rigid demand, and the environment is relatively closed, making the safety and economic values of robots replacing humans very clear.
Educational training: cultivating professional talents in the fields of embodied intelligence and robotics is a national strategic demand. Universities and research institutes have a huge demand for open hardware platforms, datasets, and training solutions, which can quickly form a commercial closed loop.
The basis for judgment is that these scenarios have clear pain points, relatively controllable environments, and quantifiable value, serving as the best whetstone for technology to mature. In short, real scenarios are the best testing ground, and stable operation is the hardest standard.
Zhang Dapeng, Assistant Vice President of Leju Robot Over the next three to five years, we believe three scenarios will take the lead in achieving a scaled commercial closed loop:
First, industrial manufacturing scenarios, such as high-frequency transport, sorting, and material loading operations in industries like automotive manufacturing, warehousing and logistics, and 3C electronics. This is the key direction Leju is currently focusing on breaking through, and our solutions have been deployed and verified in leading enterprises such as FAW, JAC Motors, Haichen Logistics, and Zhaofeng Co., Ltd. The basis for judgment is that industrial scenarios feature strong willingness to pay, rigid demand, and quantifiable results.
Second, scientific research and educational scenarios, serving as the infrastructure for technology validation and talent cultivation. Although the ceiling of this market is limited, it is the "training ground" for technology iteration and the "seeder" for ecosystem construction. In this field, Leju has served QS100 universities such as Peking University, Tsinghua University, and Harbin Institute of Technology, as well as renowned research institutions, accumulating profound scenario understanding.
Third, commercial inspection and service scenarios, such as exhibition hall guiding, campus inspection, and smart retail. These scenarios have high requirements for the mobility and interaction capabilities of robots, but the environment is relatively controllable, making it the fastest field outside of industry to achieve standardized replication.
Leju's commercialization path follows a gradual strategy of "from controllable scenarios to open scenarios, from structured environments to unstructured environments," advancing steadily and driven by scenarios.
Yang Zhengye, Marketing Director of the National and Local Co-construction Humanoid Robot Innovation Center Over the next three to five years, the easiest scenarios to generate commercial value remain industrial manufacturing and service industries, such as workstations for sorting, assembly, and screw locking on automotive, electromechanical, and 3C production lines. Factory environments are relatively controllable, without the need to pursue ultimate general intelligence, making it easier to calculate the economic account for solutions. Next are flexible logistics transport and high-risk inspections in power and mining. Commercial inspection and guiding, as well as urban services, can be used to accumulate data, but the revenue is limited; household humanoid robots will be difficult to popularize in the short term due to overly complex environments and high safety thresholds, so there is no need to rush into them.
Lin Jiawei, CMO of KuaWei Intelligence On the pan-intelligent manufacturing side, the hardware ecosystem for robotic arms, collaborative robots, and composite robots is already relatively mature. KuaWei Intelligence empowers existing equipment through algorithms and models, covering over 50 segmented industries such as automotive parts, new energy vehicles (NEV), and 3C electronics. With the maturation of general-purpose humanoid robot products, more complex application scenarios such as flexible production and dual-arm assembly will be gradually broken through and verified. On the commercial service side, during the "May Day" holiday this year, robot stores empowered by KuaWei Intelligence received a cumulative total of over 500,000 consumers, with the highest revenue of a single store exceeding 100,000 RMB. In high-traffic scenarios such as cultural tourism, supermarkets, and theme parks, robots can already stably complete tasks such as coffee making, merchandise selling, and interactive guiding. We believe that scenarios with a real labor gap, clear willingness to pay, and verifiable ROI are those with commercial value and deployment feasibility.
Deng Xiong, Vice President of VEICHI Over the next three to five years, the domestic embodied intelligent robot industry will present a clear pattern of "accelerated deployment of mature scenarios and continuous accumulation of strength in frontier directions." As the core direction of long-term industry evolution, humanoid robots will continue to break through technological boundaries, laying a solid foundation for broader applications in unstructured scenarios. Considering the certainty of commercialization and deployment feasibility, industrial manufacturing and production line logistics transport will also be two major tracks with prominent commercial value currently.
Humanoid robots are recognized by the industry as the "ultimate form" and core carrier of embodied intelligence, representing the core development direction of the industry's future. VEICHI can provide a full set of motion actuator solutions, such as rotary joint modules, linear joint modules, coreless motor modules, and dexterous hands, empowering the upgrading of the embodied intelligence industry with independent and controllable capabilities.
Industrial manufacturing is the main track with high certainty for the deployment of embodied intelligence. Currently, the demands for flexible production and human-machine collaboration in manufacturing are continuously being released. Addressing the core demand for power control in industrial and collaborative robots, VEICHI has launched human-machine collaboration solutions. Products such as force-controlled harmonic joint modules and planetary joint modules for robot dogs have achieved innovative applications in collaborative scenarios, continuously expanding the boundaries of human-machine interaction.
Logistics transport is a typical pioneering scenario for the commercialization of embodied intelligence. This scenario features clear task objectives, a high degree of standardization in operation processes, and controllable operational risks, making input-output benefits easy to quantify. The AGV/AMR motion control solution launched by VEICHI can fully meet the operational accuracy and autonomous operation requirements of mobile robots in different working environments, providing reliable power support for applications in logistics, warehousing, and other fields.
Hu Zheqi, Think Tank Manager of YQBionics Over the next three to five years, embodied intelligence will not enter all industries evenly, but will prioritize deployment in scenarios with clear task boundaries, quantifiable economic value, and environments that can be moderately modified.
The first is industrial manufacturing, including flexible loading and unloading, irregular part grasping, line-side distribution, insertion, inspection, machine tool tending, and partial assembly tasks in fields such as automotive, 3C electronics, and new energy. These scenarios have clear demands and strong customer payment capabilities, allowing actual effects to be evaluated using takt time, yield rate, success rate, and frequency of manual takeover.
The second is warehousing and logistics, including depalletizing, sorting, parcel feeding, restocking, tote transport, and abnormal parcel handling. Logistics processes have a high degree of standardization, making them easy to replicate across workstations and warehouses, and they are also suitable for leasing and pay-per-task models. IFR data shows that transportation and logistics are already the largest application category in terms of sales volume among professional service robots.
The third is institutional elderly care, rehabilitation, and medical assistance. However, the focus should first be on clearly bounded auxiliary tasks such as transport, patrol, rehabilitation training, and item delivery, rather than directly replacing nursing staff. The market space for general-purpose household robots in open homes is huge, but due to complex environments, high safety responsibilities, and cost sensitivity, the time for scaling may be relatively later.
Yang Yuxin, Chief Marketing Officer of Black Sesame Technologies Over the next 3-5 years, closed, rigid-demand, and highly standardized scenarios will take the lead in achieving large-scale commercialization, such as transport, sorting, and inspection industrial robots in the industrial field, last-mile delivery robots in the logistics field, and reception, inspection, and cleaning service robots in public service scenarios. These scenarios have clear demands, relatively controllable operating environments, clear commercialization models, and a foundation for rapid replication.
The differences in chip requirements across different scenarios are very significant: In industrial robot scenarios, more emphasis is placed on the stability, real-time performance, and reliability of computing power. Chips are required to adapt to long-term stable operation under complex working conditions, with extremely high requirements for the latency of motion control; In commercial service robot scenarios, more emphasis is placed on low power consumption, multimodal perception, and human-machine interaction capabilities, while balancing cost control. The absolute performance requirements for computing power are moderate, but the requirements for energy efficiency ratio are high; General-purpose terminals such as humanoid robots have the highest demand for computing power, requiring simultaneous support for complex tasks such as multimodal perception, autonomous decision-making, and whole-body motion control, with the highest requirements for chip computing power density and architecture scalability.
The Kalos, Aura, and Liora computing power modules launched by Black Sesame Technologies are specifically designed for such differentiated needs, adapting to different categories such as low-speed wheeled robots, multi-legged robots, and advanced humanoid robots. Currently, pilot deployments have been carried out in scenarios such as industrial logistics and smart services in collaboration with industry partners.
Author | Editorial Department of China Electronics News Editor | Wu Lilin Art Editor | Ma Liya Supervisor | Lian Xiaodong