申万宏源：2026年是物理AI关键元年 核心关注具数据闭环和场景能力本体公司

Core Insights - 2026 is identified as a pivotal year for physical AI, marking a transition from screen-based AI, with the robotics industry being compared to a hybrid of "smartphones + autonomous driving" [1] - Investment strategy should focus on a new paradigm of "intelligent layer > collaborative layer > hardware layer," emphasizing core capabilities and ecosystem building [1] Group 1: Industry Comparisons - The development of humanoid robots is closely aligned with the evolution of the electric vehicle (EV) industry, with 2026 serving as a key milestone for humanoid robots, similar to the 2012-2014 period for EVs [2] - Both industries rely on mature large-scale manufacturing and advancements in AI algorithms, with the EV sector benefiting from national strategic support, leading to a transition from policy to market and technology to ecosystem [2] Group 2: Core Industry Challenges - The essence of humanoid robots differs from that of EVs, with intelligence being the core anchor point for the former, akin to the significance of battery technology in EVs [2] - The current challenge for humanoid robots is the "intelligence deficit," as hardware commoditization occurs rapidly, while the core value lies in differentiated service capabilities [2] Group 3: Data as a Key Resource - Data is considered a critical resource in the era of embodied intelligence, comparable to lithium in the EV sector, with data collection and efficient production capabilities determining the upper limits of models [3] - The industry faces a significant data bottleneck, with a vast gap between the required trillion-level physical interaction data and the existing million-level public datasets [3] - Companies that can establish low-cost, high-efficiency data collection pipelines will create a strong competitive advantage, forming a "data-capability-order" positive cycle [3] Group 4: Investment Opportunities - Potential investment targets include data collection service providers, simulation platform ecosystem partners, and scene operators within the data industry chain, which are likened to "mining companies" in the data sector [3]