The investment narrative around artificial intelligence is transitioning from infrastructure toward application layers.
A recent report from Goldman Sachs indicates that as AI computing infrastructure continues to expand, humanoid robots are poised to become the next major growth driver for AI. Labor shortages and automation demands provide structural support for the industry's long-term development.
Goldman Sachs' Asia trading team has concluded through recent research that the Asian robotics industry is still in its early stages of development. Compared to similar US assets, the Asian robotics and automation sector currently trades at a valuation discount while offering higher earnings growth expectations, with capital already beginning to rotate into the robotics supply chain.
The bank also emphasizes that the industry is currently in a transition phase from technology validation to commercialization, with large-scale deployment still requiring time.
Rapid Technology Iteration, Commercialization Still in Early Stages
Goldman Sachs equity strategist Jacqueline Du revealed in the latest report that the bank conducted visits to 14 robotics companies in Hong Kong, Shenzhen, and Beijing from May 18 to 22, covering China's embodied AI, robotics, and automation supply chains.
The research indicated that VLA/VTLA models and world models are accelerating their integration, driving improvements in robot planning capabilities and environmental adaptability, while model sizes continue to expand.
However, unlike large language models, robotics AI faces a more acute scarcity of real-world data. Robot training requires vast amounts of physical data such as force, torque, and motion, which cannot be easily acquired at scale like internet text and images, creating a critical bottleneck that constrains model training and capability enhancement.
To address this issue, the industry is increasing investment in centralized data factories and human-robot collaboration for data collection. Goldman Sachs believes that as demand for robot data grows, data collection and related services could become a significant revenue source within the supply chain.
From an application perspective, robot deployments in industrial and logistics settings are still primarily proof-of-concept, remaining far from large-scale commercialization. Most industry participants expect that after accumulating tens of millions of hours of high-quality data and developing deployment-ready models, humanoid robots may gradually enter the scaled commercialization phase between 2027 and 2029.
Goldman Sachs projects global humanoid robot shipments will reach 76,000 units by 2027, rising to 502,000 units by 2032.
From "Able to Move" to "Able to Use," Technical Hurdles Remain
Goldman Sachs maintains a long-term positive view on the industrial potential of humanoid robots, seeing them as a potential next-generation widely adopted terminal following smartphones and automobiles. As scaled production advances, average selling prices and material costs for robots are expected to decline, improving the business model.
However, the technological inflection point has not yet truly arrived.
Goldman Sachs notes that Tesla's Optimus demonstration at the "We, Robot" event, as well as Unitree's H1 appearance at the 2025 Spring Festival Gala, show significant progress in hardware flexibility and robustness for humanoid robots. However, when considering large-scale deployment requirements in industrial and consumer settings, gaps remain in precision, consistency, cost, and general-purpose autonomous AI capabilities.
Among these, general-purpose autonomous AI remains the most critical constraint. Robots not only need to "see" and "understand" their environment but also must autonomously plan and execute continuous actions in complex real-world scenarios, placing higher demands on models, data, and hardware.
Meanwhile, NVIDIA is accelerating this process through its Physical AI ecosystem. The Jetson Thor edge computing chip, GR00T model, Isaac platform, and the Isaac GR00T Blueprint and Cosmos simulation framework can help robot developers generate large amounts of synthetic data in simulated environments, alleviating the scarcity of real physical data.
Focusing on Core Components, Seizing Asia's Valuation Opportunity
Goldman Sachs believes that the value in the humanoid robot supply chain will be concentrated in core components with high technical barriers. Among these, harmonic reducers have the highest technical barrier, requiring strict standards for precision, lightweight design, and torque performance, offering significant growth potential. Actuator assemblies have relatively high certainty of technical adoption in high-specification robots.
In contrast, planetary roller screws are still in a phase of rapid change, with uncertainties around yield rates, production consistency, and capacity readiness. The technological roadmap for dexterous hands also remains unclear.
In terms of valuation, the Goldman Sachs Asia trading team believes that the Asian robotics sector still trades at a discount compared to similar US assets. Data shows the median price-to-earnings ratio for the Asia-Pacific robotics basket is 22 times, compared to 28 times for US peers, representing a discount of approximately 21%. On a PEG basis, the Asia-Pacific robotics basket stands at 1.5 times, while the US basket is at 2.0 times.
This suggests that Asian robotics and automation companies currently have relatively low valuations, while their earnings growth expectations do not lag behind those of comparable US assets.
Capital flows are also beginning to shift. Goldman Sachs has observed that mutual funds are gradually rotating into the robotics supply chain, but overall positions remain in early stages, with capital primarily concentrated in components, automotive automation, and industrial automation and precision manufacturing.
Goldman Sachs believes this indicates that the robotics industry has not yet entered a phase of broad capital congestion. If the industry enters a period of scaled commercialization, the current phase may still represent the early stages of capital rotation.
Comments