Group 1: Macroeconomic Insights - The report discusses the potential for PPI to turn positive, with September CPI at -0.3%, PPI at -2.3%, and expectations for both being slightly negative [5][9][12] - Core CPI has shown a seasonal decline, with September's core CPI remaining at 0%, marking the first time since April 2025 that it fell below seasonal expectations [10][11] - The report anticipates that if PPI remains at 0% from October 2025 onward, the average PPI for 2026 could be around -0.7% [14] Group 2: Electronic Industry Insights - The report highlights the acceleration of commercialization in domestic AI hardware, focusing on the synergy of computing power, storage, and operational capacity [18][19] - The demand for Scaleup and Scaleout hardware is expected to grow significantly, with the global market for Scaleup exchange chips projected to reach nearly $18 billion by 2030, with a CAGR of approximately 28% from 2022 to 2030 [19] - The report identifies a low domestic production rate for operational hardware, indicating a significant opportunity for domestic replacements in the market [21] Group 3: Robotics Industry Insights - The report introduces Figure03, a humanoid robot designed for mass production, emphasizing safety and comfort in home environments [23][24] - Figure03 features advanced capabilities such as tactile sensors for stable operation in limited visibility environments and supports wireless charging for continuous operation [24][25] - The report notes that Figure aims to produce over 100,000 units within four years, with a projected valuation of nearly $40 billion, supported by significant investments from major tech companies [24][26]

针对Scale up、Scale out分别有不同的通信协议，大厂自研私有协议+第三方与中小厂推动公有协议将成 为未来的发展趋势。具体而言，Scale up层面英伟达NVlink、AMD Infinity Fabric(Ualink)与华为UB mesh 是私有协议的代表，而博通所推动的SUE与行业历史悠久的PCIe是公有协议的代表。Scale out层面，英 伟达Infiniband为私有，博通则致力于推动基于公有以太网的RoCE2，此外海外众多厂商还一起推动超 级以太网联盟，成为Scale out新力量。 运力硬件国产化率极低，有望成为下一个高赔率的国产替代新方向 智通财经APP获悉，开源证券发布研报称，当下传统算力架构已难以满足高效、低耗、大规模协同的AI 训练需求，超节点成为趋势，其通过提升单节点计算能力，大幅带动了Scale up相关硬件需求。另一方 面，超大规模AI集群的建设需要横向推动大量节点之间的互联，带动Scale out相关硬件需求，随着更大 规模的集群逐渐出现，单一地区的电力资源成为瓶颈，于是跨数据中心的Scale across方案也将在未来逐 步采用。针对Scale up、Scale ...