Goldman Sachs' latest research shows that China's humanoid robot industry is experiencing a transition from; Universal Imagination "; Towards'; Dedicated landing "; Strategic transformation. This pragmatic approach, combined with significant progress in motion control capabilities and rapid iteration cycles, is driving major manufacturers to set their 2026-2027 shipment targets for several times the growth of 2025.
According to the Wind Chasing Trading Platform, Jacqueline Du, an analyst at Goldman Sachs, pointed out in her latest report that in a survey conducted from January 15-20 on eight humanoid robot and industry chain companies, including Yushu Technology, Ubiquitous, Fourier, and Yunshen, the team observed that the industry is shifting its focus from pursuing universal capabilities to vertical scenarios that can utilize existing task planning, mobility, and interaction capabilities, such as security patrols, public place guidance services, and factory logistics sorting.
Based on feedback from major manufacturers and supply chain companies, Goldman Sachs expects global shipments of humanoid robots to reach approximately 15000 to 20000 units by 2025, with Chinese companies contributing the majority of shipments. Looking ahead to 2026-2027, top manufacturers are expected to achieve several fold growth, increasing their scale from hundreds to thousands of units in 2025 to thousands to tens of thousands.
Goldman Sachs points out that 2026 may become a critical year; Volume verification+expected reset; In this year, investors will focus on; Million Robots "; Whether milestone expectations have been revised, as well as the evolution of market share and single machine value for individual supply chain companies.
Radical shipment targets pose challenges to production capacity and testing
According to Goldman Sachs' research, the global shipment of humanoid robots in 2025 is expected to be between 15000 and 20000 units, which is close to Goldman Sachs' previous expectation of 20000 units and consistent with third-party data's range of 13000 to 16000 units. Chinese companies currently contribute the vast majority of shipments, with demand mainly coming from scientific research, robot AI training, education, entertainment performances, and data factories.
In this still early-stage market, top manufacturers have set ambitious growth targets for 2026-2027. Based on the shipment volume ranging from hundreds to thousands of units in 2025, various enterprises have set their targets for 2026-2027 at a scale of thousands to tens of thousands of units, which means several fold growth.
The supporting factors for this growth expectation include increasingly mature supply chains, optimized cost curves, and expanded application scenarios. However, Goldman Sachs pointed out that achieving these goals will face challenges such as ensuring production consistency and the inherent multi-stage testing process of this emerging industry.
Significant improvement in motion control, with iteration cycle reduced to 6-8 months
During the on-site product demonstration, Goldman Sachs analysts observed that humanoid robots have made substantial progress in motion control, demonstrating stronger robustness and flexibility in both wheeled upper body platforms and fully bipedal systems, with significant improvements compared to the previous year.
A manufacturer claims to have achieved '; Cerebellar level; The robot has full body control capability and provides two practical evaluation criteria: the ability to navigate on terrain without pre mapping, and the ability to achieve full body remote control instead of segmented upper and lower body control.
The integration capability of the supply chain is accelerating product iteration. Multiple companies have revealed that the product iteration cycle of the humanoid robot platform has been shortened to about 6-8 months per generation. This rapid iteration is largely attributed to 80% -90% of the independent design capability of components, which is crucial for ensuring seamless integration of software and hardware, and optimizing their respective performance limits within a compressed R&D and testing cycle.
Application Focus; Dedicated landing "; Avoiding the difficulty of agile operation
" Simulate to Reality "; The gap is still a bottleneck in the industry. The current pre training of robots heavily relies on simulation and synthetic data, and the accuracy of 80% -90% in simulated environments often drops below 50% in real scenarios. Due to the time required for large-scale, high-quality real data collection and world modeling methods, China's leading humanoid robot developers are prioritizing the development of; Dedicated "; Commercial deployment.
These application scenarios include security patrols and guidance services in public places such as hotels, banks, museums, exhibition centers, car dealerships, and supermarkets, which can effectively utilize existing task planning, mobility, and interaction capabilities while avoiding the complexity of highly agile operations.
In industrial applications, humanoid robots that require dexterous hands or fixtures are currently limited to logistics tasks such as container handling and simple item sorting. This is mainly due to the limitations of AI in handling unpredictable corner cases in factory environments. According to Uber, in sorting and logistics applications, customers are willing to invest when robots reach about 50% of human worker capacity, which can result in a payback period of about two years (assuming about 10 hours of operation per day). Even in environments where labor is particularly scarce, a three-year payback period is considered acceptable.
Data strategy becomes the core competitiveness, and world models receive attention
In recent times, humanoid robot manufacturers have increasingly adopted standardized methods to integrate with mature Large Language Model (LLM) and Visual Language Model (VLM) technology stacks such as Alibaba (Tongyi Qianwen), Doubao, and Tencent. This strategy makes proprietary data engines a key differentiating factor in developing deployable robot intelligence.
High quality real-world data is considered a major constraint in bridging the gap between mature hardware technology and scalable practical applications. Therefore, companies are carrying out; Data Formula "; The arms race is driven by differentiation in its target terminal applications.
Although all robot manufacturers are pursuing data collection strategies, they have adopted different combinations of three main data inputs: remotely operated human or expert demonstration data, which is highly controllable but often expensive; Simulated data, with low cost but imperfect authenticity for each additional sample; And real-world video datasets, which have the highest data availability but may have lower accuracy when converted to the real world.
Goldman Sachs found in this survey that the mention of world model methods is increasing, which may give robots some common sense about their environment, leading them to shift from reactive behavior to active agents capable of complex planning and adaptation.
Business model differentiation: 2C emphasizes experience, 2B focuses on investment return
Different target markets have given rise to differentiated profit models, mainly divided into 2C (consumer oriented) and 2B (enterprise oriented) applications.
Companies targeting 2C applications primarily focus on providing differentiated features and enhancing user experience, typically emphasizing the importance of; Emotional Value "; And capture specialized vertical segmented markets, obtaining premiums through unique features or interactions. The goal is to create products that stand out through their abilities and user engagement.
In contrast, companies targeting 2B applications anchor their pricing strategies on customers' return on investment (ROI), typically by demonstrating how robots can improve production capacity, enhance efficiency, or reduce labor costs. Ubiquitous stated that in sorting and logistics applications, customers are willing to invest when robots reach about 50% of human worker capacity, which can bring about a return on investment of about two years. Even in environments where labor is particularly scarce, a three-year payback period is considered acceptable, highlighting the value proposition of automation in addressing critical operational challenges.
