Goldman Sachs: AI Second Half Focuses on Humanoid Robots

Goldman Sachs' latest research report points out that the global AI investment narrative is shifting from infrastructure to the application layer, with humanoid robots emerging as the next clear core track.

As large models and embodied intelligence accelerate integration, the Asia-Pacific robotics ecosystem is welcoming early-cycle investment opportunities brought about by capital rotation.

Despite high-quality real-world data still being the main bottleneck, the industry expects large-scale commercial deployment to be implemented between 2027 and 2029.

Technical Integration Breakthrough and Real Data Bottlenecks

Goldman Sachs strategist Jacqueline Du found through research that industry discussions have transcended the single VLA framework and turned towards execution-oriented multimodal stack integration. The integration of VLA/VTLA with world models significantly enhances the planning capabilities and robustness of robots, with model sizes migrating towards 40 billion to 80 billion parameters.

The core focus of the industry has now shifted to how to build a scalable data collection architecture. Companies such as Pasini and Xingtu Dynamics are building closed loops through various channels. Due to model capability constraints and cost considerations, many players currently prefer the transitional form of wheeled chassis with robotic arms, thereby covering the vast majority of industrial applications.

Clear Path to Commercialization but Patience is Required

Despite rapid technological iteration, Goldman Sachs emphasizes that most applications are still in the conceptual validation phase, and transitioning to large-scale commercialization still faces complex transformation. Typical industrial deployment requires going through conceptual validation, small-batch testing, and pilot deployment, with tasks like sorting and material handling as the main application opportunities in the near term.

The industry generally expects that after accumulating tens of millions of hours of high-quality data and creating ready models, scaled deployment will be achieved between 2027 and 2029. Goldman Sachs' shipment forecast is relatively cautious, expecting global humanoid robot shipments to be 76,000 units in 2027, significantly lower than the current market consensus.

Valuation Lowlands and Capital Flows Drive Early-Cycle Opportunities

Goldman Sachs trader Peter Sheren points out that labor shortages and automation demands are driving global funds to rotate towards Asia-Pacific robotics-related targets.

The median price-to-earnings ratio of the Asia-Pacific robotics stock basket is 22 times, a discount of about 21% compared to the similar 28 times in the United States. With high growth expectations added, the cost-performance ratio is even more prominent. The median PEG ratio of the Asia-Pacific basket is 1.5 times, lower than the 2.0 times in the United States, reflecting that Asian automation companies have higher growth space per unit of profit. Currently, mutual funds have begun to rotate toward supply chain components, and liquidity adjustments show that mid-cap stocks such as Hengli Hydraulic and some automotive parts manufacturers have seen significant benefits.

Supply Chain Focuses on High Barriers and Certain Targets

In the supply chain investment framework, Goldman Sachs prefers core components with high content value, focusing on two major directions: harmonic drives and actuator assemblies.

For the high-barrier harmonic drives, Goldman Sachs has raised its market share forecast for high-specification robots for Green's Harmonic to 30% from 2025 to 2030. The technical adoption certainty in the actuator assembly area is higher, with Goldman Sachs predicting that the market share distribution of Sanhua Holdings and Top Group will be adjusted to 70% versus 30%. In comparison, the yield and capacity of planetary roller screws are more uncertain, and there is still disagreement on the technology roadmap for dexterous hands, with the long-term winners still unclear.