Shenwan Hongyuan Group: 2026 is the critical year for physical AI, with a focus on core capabilities of closed-loop data and situational ontology companies.

Shenwan Hongyuan Group released a research report stating that 2026 is a crucial year for physical AI to separate from screen AI. The Siasun Robot&Automation industry is compared to a hybrid of "smartphones + autonomous driving", and investments should follow a new paradigm of "intelligent layer > collaborative layer > hardware layer" focusing on core capabilities and ecosystem building. The key focus should be on ontology companies (mechanical + automotive) with closed-loop data and scene capabilities; followed by companies in the data scene and metaverse-related fields (computer coverage); and lastly, high-quality component companies. Key points from Shenwan Hongyuan Group: There is a high degree of homogeneity between the new energy vehicle and humanoid Siasun Robot&Automation industries. The development milestone of humanoid Siasun Robot&Automation in 2026 corresponds to the period of 2012-2014 for new energy vehicles. The industrial evolution path of new energy vehicles also provides a clear benchmark framework for the Siasun Robot&Automation industry. Both industries rely on mature large-scale manufacturing and AI algorithm upgrades. The domestic new energy vehicle industry has benefited from national strategic drive, completing the evolution from policy to market, technology to ecosystem. In 2026, the technology of Siasun Robot&Automation has just crossed the threshold of usability, with unprecedented policy drive and capital heat. The similarity with the characteristics of new energy vehicles that landed after the Model S in the past is apparent, but the commercial model has not yet formed a closed loop. Although humanoid Siasun Robot&Automation and the new energy vehicle industry have stage benchmarks, their industrial natures are different. Intelligence is the core industry anchor of the former, comparable to the core industry anchor level of new energy power batteries for new energy vehicles from 2008 to 2020. China achieved a significant reduction in battery costs relying on economies of scale during this period and established the rationale of "battery supremacy" in hardware investment. Currently, Siasun Robot&Automation benchmarks the new energy vehicle industry of 2012, and while the hardware logic is temporarily correct, the core contradiction is the "intelligence deficit". The hardware body quickly commoditizes with a drastic reduction in costs along the supply chain, and the industry's focus shifts to embodied intelligence. The differentiation in service capabilities is crucial, with the embodied intelligent brain as the core moat. Hardware and intelligence are not mutually exclusive, and there is still room for significant iteration in core hardware in 2026. The two form a positive cycle of "intelligence defining hardware, hardware feeding back intelligence", with the iterative direction of hardware defined dynamically by intelligence needs. Data is a core resource in the age of embodied intelligence, akin to lithium ore. The acquisition and efficient production capabilities of data determine the model's upper limit, and the data industry chain has become a core investment direction. In the face of a severe data bottleneck in physical AI, the gap between the trillion-level physical interaction data required by VLA models and the existing million-level public datasets is vast. The scarcity of data sensed by the ontology poses the biggest challenge, with companies in the industry competing for data acquisition through data collection centers, VR remote operations, motion capture, and other methods. However, obtaining such high-value data currently has high costs and low efficiency. Data factories are the core starting point for the intelligent body of Siasun Robot&Automation, and enterprises that can build a large-scale remote operation data collection pipeline at low cost and high efficiency will establish a deep moat, forming a positive cycle of "data - capability - orders". Simulation technology and synthetic data have become important accelerators for data production, with NVIDIA's Isaac Lab and synthetic data startups taking the lead. In terms of investment, in addition to hardware companies, companies in the data industry chain such as data collection service providers, simulation platform ecosystem partners, and scene operators are worth paying attention to as potential beneficiaries. Core risks: fluctuations in raw material prices, geopolitical risks, industry recovery falling short of expectations.