Goldman Sachs suggests the AI investment narrative is transitioning from the infrastructure layer to the application layer, with humanoid robots emerging as the next clear frontier for monetization. The firm advises investors to position early to capture the opportunities presented by a multi-year capital rotation cycle into the Asia-Pacific robotics ecosystem.

In a recent research report, Goldman Sachs equity strategist Jacqueline Du detailed a series of engagements from May 18 to 22, including participation in an Asian technology conference in Hong Kong and a dedicated AI robotics research trip to Shenzhen and Beijing. The team held in-depth discussions with 14 robotics companies spanning both private and public markets. The report stated the trip was "highly encouraging," noting that the rapid integration of Vision-Language-Action (VLA)/Vision-Touch-Language-Action (VTLA) models with world models is significantly enhancing robots' planning capabilities and robustness, with model scales also migrating toward roughly 40 to 80 billion parameters.

Peter Sheren, a trader at Goldman Sachs in Singapore, indicated that the AI narrative is pivoting from infrastructure to applications. Structural tailwinds like labor shortages and automation demand are accelerating industry implementation, driving capital rotation toward robotics-related stocks in South Korea, Japan, and China. From a valuation perspective, the median price-to-earnings ratio for the Asia-Pacific robotics stock basket (22x) trades at an approximately 21% discount to its U.S. counterpart (28x). Coupled with higher earnings growth expectations, this makes Asia-Pacific robotics stocks more attractive on a relative value basis.

Despite the optimistic outlook, Goldman Sachs clearly notes that scaled commercial deployment will require patience. Most current applications remain in the proof-of-concept stage, with industry participants broadly expecting large-scale deployment to materialize between 2027 and 2029. High-quality real-world data remains the core bottleneck constraining industry development.

**Five Days, Two Cities: Key Findings from Goldman Sachs' Mid-Year Research**

The research trip spanned five days in two segments: participation in the GS Asia Communacopia + Technology Conference in Hong Kong on May 18-19, followed by a specialized China AI robotics field study in Shenzhen and Beijing from May 20-22. The team engaged with 14 companies, representing a broad cross-section of the embodied AI, robotics, and automation ecosystem, including: Damo Robotics, Dobot, Estun Automation, X-Times, Gabot, Geek+, LimX Dynamics, Linker Robotics, Mech-Mind, Wanji Robotics, Pashini, Lingjing Intelligence, Ubtech, and Fangyu Robotics.

Jacqueline Du characterized the research as a "mid-year check on industry development," summarizing two core findings. First, the integration of VLA/VTLA models with world models is advancing rapidly, with model scale expanding significantly. Second, high-quality real-world data remains the primary bottleneck for current deployment, though industry discussion has shifted from macro debates over "data recipes" to how to build scalable data collection architectures.

**Technology Acceleration: Multi-Modal Stack Integration Speeds Up, Data Competition Heats Up**

From a technology roadmap perspective, industry discussions during the research have clearly moved beyond a single VLA framework toward execution-oriented multi-modal stacks. The specific path involves first rapidly integrating VLA with world models, then adding tactile capabilities (VTLA) in segments demanding high physical interaction quality. World models are no longer viewed as an independent model category but operate as a functional layer parallel to action models—VLA or VTLA handles strategy generation and action execution, while world models enhance execution quality through next-state prediction, action validation, and planning optimization.

Companies explicitly targeting the combination of VLA/VTLA with world models as their next step include X-Times, Gabot, Lingjing Intelligence, and Wanji Robotics. Among them, X-Times launched Fast-WAM in March 2026, achieving latency as low as 190 milliseconds. Lingjing Intelligence's open-source model Spirit v1.5 topped the RoboChallenge Table30 with a score of 66.09 and a 50.33% success rate, becoming the first Chinese open-source embodied model to surpass Pi0.5.

On the data front, human-centric, egocentric data collection methods are becoming an industry consensus. Pashini already operates five data factories nationwide. X-Times, Lingjing Intelligence, and Wanji Robotics are building distributed data closed loops through deployed systems, wearable devices, VR, and client-side collection. Several companies expect data-related revenue contributions to rise significantly by 2026, with Ubtech indicating that government data factory demand in 2026 will be flat or stronger compared to 2025.

Notably, Goldman Sachs observed that, due to model capability constraints and cost considerations, many current players favor robot forms with wheeled bases and two-to-three-finger grippers. This solution is seen as covering 70% to 90% of industrial application scenarios, while not precluding future evolution toward bipedal humanoid robots and five-fingered dexterous hands.

**Commercialization Path: Dominated by POC, Scaled Deployment Awaits 2027–2029**

Regarding commercialization progress, application scenarios are extending into industrial handling, logistics workflows, and structured commercial settings, but the overall landscape remains dominated by proof-of-concept projects, with large-scale deployment yet to arrive.

The typical path for industrial implementation involves four steps: POC phase (typically lasting 3 to 6 months, averaging 2 to 3 iterations) → small-batch testing (typically factory orders below 50 units) → an approximately 12-month validation period → pilot deployment (order sizes gradually moving toward 50 to 100 units per customer). Recent representative application opportunities include sorting, material handling, pick-and-place, inspection/testing, and other standardized or semi-structured workflows.

The industry broadly anticipates that large-scale commercial deployment will be achieved between 2027 and 2029, following the accumulation of tens of millions of hours of high-quality data and the development of deployment-ready models. Goldman Sachs believes that despite current challenges, "the long-term investment prospects for this industry are highly promising," but "this journey still requires patience, as companies must navigate the complex transition from proof-of-concept to large-scale commercialization, with quality stability and sustained cost reduction being key milestones."

**Supply Chain Focus: Targeting High-Barrier Core Components**

Within its supply chain investment framework, Goldman Sachs favors products with high content value, focusing on two main directions: harmonic reducers and actuator assemblies.

For harmonic reducers, Goldman Sachs views them as having the highest technical barriers, with stringent requirements for precision, lightweight design, and torque. Among high-specification products, Harmonic Drive Systems and Leader Harmonious Drive Systems are established players. Goldman Sachs has raised its market share forecast for Leader Harmonious Drive Systems in high-spec robots for 2025-2030 to 30%, maintaining Harmonic Drive Systems at 70%. The firm anticipates the two may trend toward a 50%/50% long-term equilibrium post-2030, primarily because the growth of Chinese humanoid robot players benefits Leader Harmonious Drive Systems more.

For actuator assemblies, Goldman Sachs sees higher certainty in technology adoption. Against the backdrop of automakers developing humanoid robots, actuator assembly companies possess unique value due to their ability to manage complex supply chain relationships. Goldman Sachs expects the market share allocation between Sanhua Holding and Tuopu Group to shift from an initial 50%/50% to 70%/30%, based on Sanhua Holding's established global leadership in Tesla EV thermal management module assemblies.

Regarding planetary roller screws, Goldman Sachs believes the market landscape is still evolving rapidly, with significant uncertainty around yield rates, production consistency, and capacity readiness. Schaeffler is currently estimated to hold about 50% to 60% market share, but the long-term winner remains unclear. For dexterous hands, the technology roadmap also shows considerable divergence, with mainstream solutions still competing.

**Valuation and Trading Strategy: Asia-Pacific Discount Coupled with High Growth Highlights Early-Cycle Opportunity**

From a valuation standpoint, the mean price-to-earnings ratios for Goldman Sachs' Asia-Pacific and U.S. robotics baskets are similar (approximately 29-30x). However, the Asia-Pacific median P/E (22.0x) is significantly lower than the U.S. median (28.0x), implying a typical Asian industrial/robotics stock trades at an approximate 21% discount to its U.S. peer. The median price/earnings-to-growth ratio for the Asia-Pacific basket is 1.5x, lower than the U.S.'s 2.0x, reflecting that Asian robotics/automation stocks offer higher implied earnings growth per unit of valuation. This primarily benefits from the relatively higher expected growth rates of Chinese and South Korean automation companies like Ningbo Tuopu, Sanhua Holding, and Hengli Hydraulic.

Regarding fund flows, mutual funds have begun rotating into robotics-related supply chains, but overall positioning remains in early stages, concentrated on components rather than pure-play robot manufacturers. Capital is primarily flowing into South Korean and Chinese auto parts suppliers, Chinese industrial automation/precision manufacturing firms, and select robotics component suppliers. Passive fund rebalancing is driving noticeable rotation in the South Korean market (with Mobis receiving inflows and Hyundai Motor experiencing outflows). Liquidity-adjusted flow data shows mid-cap stocks like BizLink and Hengli Hydraulic have particularly benefited.

On shipment forecasts, Goldman Sachs takes a more cautious stance than the market: the firm predicts global humanoid robot shipments of 76,000 units in 2027 and 502,000 units in 2032, below prevailing market expectations. Goldman Sachs notes the market may have already priced in 500,000 units for 2027, corresponding to a 40x exit P/E ratio. Upside valuation potential stems from the sum of core business value plus the optionality value of humanoid robots; downside risk arises if the robot optionality value goes to zero, leaving valuation supported solely by core business value.

Goldman Sachs concludes that with structural demand accelerating and Asia still trading at a significant valuation discount to the West while exhibiting stronger growth momentum, "this represents an early-cycle opportunity to position ahead of a multi-year capital rotation fully entering the robotics ecosystem."