BOCOM INTL released a research report stating that the valuation logic for humanoid robots is gradually shifting from a technology Beta to a delivery Alpha, while maintaining a positive outlook on core component segments that possess high value and cross-platform reuse capabilities. Two major events further confirmed the firm's previous judgment that 2026 is a critical stage for embodied intelligence to advance from concept validation to commercial deployment verification. As real-world scenarios are introduced and orders gradually materialize, the market's pricing focus is expected to further shift toward order fulfillment, scale delivery, and scenario replication capabilities. With the gradual maturation of the core component supply system, the firm anticipates that valuation differentiation across the industry chain will accelerate. Upstream core component manufacturers with technological and mass-production barriers are expected to continuously benefit from the ramp-up of complete machines, while embodied intelligence companies capable of stably replicating demonstrated capabilities into real-world scenarios and consistently achieving order and revenue realization are also likely to secure more sustained valuation premiums.

Key viewpoints from BOCOM INTL are as follows:

Event

The 2026 World Robot Conference and the 2nd World Humanoid Robot Games were both held in Beijing in August, ushering the robotics industry into a period of intensive validation. The 2026 World Robot Conference (WRC) took place from August 19-23 at Yizhuang, Beijing, under the theme "Human-Robot Symbiosis, Production and Demand Integration," attracting 373 domestic and international enterprises and 3,000 innovative products, with 311 new products unveiled. Compared to previous editions, this year's conference placed greater emphasis on real-world applications, with manufacturing, warehousing and logistics, catering and retail, medical and elderly care, and emergency rescue becoming key display scenarios. The 2nd World Humanoid Robot Games opened on August 22 at the National Speed Skating Oval in Beijing, with competitions running from August 22-26. A total of 666 teams and 2,056 robots from 16 countries participated in 51 events and 1,301 matches, with both the scale of competition and autonomy requirements significantly elevated from the inaugural edition. The firm believes the two events collectively reflect that the industry's evaluation system is shifting from "can it be done" to "can it complete tasks stably and economically," though the two events serve different validation functions. The WRC is closer to a commercial test oriented toward real customer needs, with the core focus on verifying whether robots can achieve autonomous operation in factories, warehouses, retail, and other scenarios, while delivering acceptable investment returns in terms of efficiency, stability, and cost. The Humanoid Robot Games, on the other hand, more closely resemble an engineering stress test for the complete machine, joints, motion control, and autonomous algorithms. From this perspective, industry competition is shifting from single-machine parameters toward a system-level capability of "hardware + model + data + scenario." The firm judges that as the core component supply system matures, the barrier to complete machine integration is expected to marginally decline, while true differentiation will increasingly manifest in generalization capabilities, data closed loops, scale delivery, and long-duration stable operation.

The most notable change at WRC is the elevated importance of real-world operation scenarios, with commercial evaluation standards shifting from demo effects to efficiency and economics

At this year's WRC, "real-scenario operation" has become a major display direction for most exhibitors, with robots further entering practical application scenarios such as transportation, logistics, healthcare, and catering. The industry is beginning to pay greater attention to generalization capabilities under varying materials and environmental changes, operational precision, and work rhythm. Data from public testing and actual deployments outside the exhibition is also starting to provide quantitative references. For example,自变量 (unlisted) completed a one-hour fully autonomous logistics sorting public test on August 12, prior to the opening of WRC, achieving 1,816 pieces per hour with an accuracy rate exceeding 98%. 星动纪元 (unlisted) previously achieved a maximum feeding volume of 1,200 pieces per hour in its actual deployment at China Post's Guangzhou Mail Processing Center, with operational efficiency reaching over 85% of manual labor. The firm believes that robot commercialization is gradually entering a quantifiable validation phase, where metrics such as success rate, UPH (units per hour), continuous operation time, human intervention rate, and unit task cost are expected to gain importance. Meanwhile, in certain logistics and industrial scenarios, simpler hardware solutions that can achieve superior efficiency and cost may achieve scale deployment faster.

As the complete machine and execution hardware continue to iterate, model generalization and high-quality data are becoming important variables for robots to move from "knowing how to do" to broader deployment

Based on the exhibits and vendor presentations at this year's WRC, VLA remains an important engineering direction, with multiple companies also exploring the use of world models for environment understanding, state prediction, and action planning. Model capabilities are further evolving toward cross-task and cross-platform generalization. 星海图 (unlisted) released the G0.5 in June, which uses a unified heterogeneous action codec to enable a single action representation to cover different robot platforms, reflecting the trend of cross-platform model reuse. On the data front, this year's WRC saw a significant increase in multi-modal collection solutions, including first-person view devices, data gloves, and tactile sensors. Orbbec (688322 CH/Buy) also showcased body-agnostic data collection hardware such as EGO, UMI, WristCam, and Hub. The firm believes that the key constraint on improving robot generalization capabilities is shifting toward "whether there is enough high-quality data and whether an effective data closed loop can be formed." The future competitive focus lies in whether companies can form a "deployment-data feedback-model iteration-redeployment" closed loop through real-world deployments and high-efficiency simulation. Companies with broader real-world scenario coverage and more operating devices are expected to continuously enhance model generalization capabilities.

The World Humanoid Robot Games further intensified the validation of software-hardware system collaboration capabilities

Autonomy requirements were significantly elevated at this year's event. According to the organizing committee's latest rules, with the exception of the 100m hurdles and 400m hurdles, all track events and other competitive events require robots to complete tasks fully autonomously. Additionally, the number of events increased from 26 in the inaugural edition to 51 this year, including 30 competitive events and 21 scenario-based events, with new high-intensity events such as long jump, weightlifting, tug-of-war, and table tennis added. The 21 scenario-based events revolve around nine real-world scenarios: home cleaning, hotel services, industrial production, emergency rescue, library management, supermarket retail, office services, garden inspection, and vehicle charging. The firm believes that high-dynamic, continuous-motion, and adversarial tasks can further test the performance and reliability of core components such as joint motors, reducers, body structures, and drive controls, while autonomy requirements further increase the challenge for environmental perception, decision planning, and motion control capabilities. The industrial value of the games lies more in providing a high-intensity, repeatable comprehensive validation environment for software and hardware, rather than simply breaking movement records.

Supply chain maturation is expected to continue driving complete machine cost reduction, but competition in core components will also shift from "availability" to performance, reliability, and mass production capability

This year's WRC featured a full industry chain exhibition covering complete machines and core components, with upstream categories including sensors, reducers, ball screws, motors, batteries, dexterous hands, and integrated joints. The firm believes that a mature core component supply system helps shorten the complete machine development and iteration cycle and promotes cost reduction, but it will also reduce the space for differentiation based solely on hardware combinations. In the future, upstream competitiveness will be more reflected in performance/cost ratio, mass production consistency, reliability, and cross-platform adaptation capabilities. For complete machine manufacturers, supply chain integration is the foundation for mass production, while the importance of model, data, and scenario delivery capabilities is expected to further increase.