BOC International: From display to deployment of industrial scenarios and accelerated verification of physical AI
The Zhitong Finance App learned that BOC International released a research report saying that one of the main meanings of the 2026 World Artificial Intelligence Conference (WAIC) held in Shanghai is that the evaluation criteria for the intelligent industry are shifting from “can complete action” to “can continue to create production value.” Industrial and warehousing scenarios may become an important verification window. Subsequent industry competition will gradually shift to task success rate, manual takeover rate, fault recovery ability, and customer payback period, rather than simply comparing freedom, computing power, or sports performance. From an industrial chain perspective, investment opportunities will gradually expand from ontology concepts to operational capabilities and production line integration, but component performance still needs to be based on fixed mass production points.
The main views of BOC International are as follows:
incident
On July 17-20, the bank conducted field research at the 2026 World Artificial Intelligence Conference (WAIC) held in Shanghai, where over 1,100 companies concentrated on showcasing more than 3,000 products. As far as the advanced manufacturing industry is concerned, in addition to the increase in the number of exhibitors and robots, it is worth noting that intelligent enterprises have begun to focus more intensively on practical tasks such as warehousing and handling, loading and unloading, quality inspection, precise operation, and public services. Among them: Zhiyuan released new products such as Expedition A3 Ultra, Genie G2Max, Lingxi X2EDU, and Dexterous Hand. Among them, G2Max cooperated with JD Logistics and plans to enter the Zhilang warehouse to carry out handling and palletizing operations; Shenji, a subsidiary of NIO, launched a “Ruidong” embedded intelligent development platform equipped with the NX9031U to extend the car-side sensing, control and software tool chain to the field of robotics.
The bank believes that one of the main meanings of this year's WAIC is that the evaluation criteria for the intelligent industry are shifting from “can complete action” to “can continue to create production value”
In the past, the exhibition focused more on walking, jumping, and single grasping. This year, companies began actively showcasing the continuous operation capabilities of robots in warehousing, semiconductor loading and unloading, assembly, quality inspection, and public services. Zhiyuan revealed that the G2Max double arm peak load is 50 kg and can support 7×24 hours of palletizing; during the conference, multiple robots were also used to carry out guided tours and public services. The bank determined that in the future, the competitive focus of the industry will gradually shift to task success rate, manual takeover rate, fault recovery ability, and customer payback period, rather than simply comparing freedom, computing power, or sports performance.
The industrial and warehousing scenario is expected to become an important verification window for intelligent commercialization in the second half of 2026
Compared to full-sized biped humanoid robots, wheeled two-arm and specialized robots may land in customer scenarios and generate revenue faster. The boundaries of warehousing, handling, unloading and quality inspection tasks are relatively clear, the working environment is relatively manageable, and it is also easier to measure economic returns by saving labor, improving equipment utilization, and extending operating time. G2Max entered the JD Logistics Zhilang warehouse, reflecting that downstream customers are shifting from exhibition trials to real-world scenario deployment. The bank judged that in the short term, priority should be given to robot companies with existing customer scenarios and system integration capabilities, rather than assuming that a single universal humanoid platform can quickly cover all manufacturing processes.
From an industrial chain perspective, investment opportunities will gradually expand from ontology concepts to operational capabilities and production line integration, but component performance still needs to be based on fixed mass production points
Dexterous hands, vision, touch, force control, and multi-machine collaboration solutions increased markedly at this conference. Companies such as Mecamand showcased “one brain polymorphism” and collaborative eye, brain, and hand operation. Zhiyuan also launched a new generation of dexterous hands. The bank believes that robots extend from handling to insertion, assembly, and fine operation, and the importance of joint modules, speed reducers, servo motors, screws, machine vision, and force sensors has increased accordingly. Relevant targets include Sanhua Intelligent Control, Tuopu Group, Green Harmonics, Double Ring Drive, Mingzhi Electric, Buke Co., Ltd., and Huichuan Technology. Currently, most of the cooperation is still in the sample delivery, joint development, or small batch stage, and the supplier share and stand-alone value may still be adjusted according to the main product plan.
At the valuation level, WAIC forms a thematic catalyst for the robotics and automotive physics AI sector
Based on the 5nm automotive-grade NX9031U, Shenji launched a “smart” intelligent platform reflecting the multi-sensor sensing, real-time control, functional safety, and mass production tool chain accumulated by the automotive industry, which is expected to spill over to robots. Its high-spec NX9031X has been deployed on NIO and Ledao models. For robot body and component companies, whether valuation can shift from themes to fundamentals still depends on paid orders, repeated procurement, customer acceptance, stand-alone price, and project gross profit margin in the second half of 2026. The bank reiterated the views of the previous report and recommended that priority be given to companies with mature manufacturing capabilities and clear customer projects and revenue conversion paths.
Risk warning: The introduction and acceptance of intelligent customers was slower than expected; changes in supply chain share due to product plan adjustments; robot stability and economic returns did not meet customer requirements; there was a difference between the exhibition operation data disclosed by the company and the actual production environment.