GB300 赋能 AI 工厂，机器人引领未来

Investment Rating - The report assigns an "Accumulate" rating for the industry, indicating a positive outlook for investment opportunities [4]. Core Insights - Nvidia is repositioning itself as an AI infrastructure company, emphasizing the distinction between traditional data centers and AI factories, which generate high-value "tokens" from power and data inputs [4]. - The next-generation GB300 AI system is set to launch in Q3 2025, boasting a 50% increase in computing power and memory compared to the GB200, with each B300 GPU featuring 288GB of HBM3e memory [4]. - Nvidia's DGX Spark, aimed at individual developers and small teams, will provide significant AI computing capabilities, supporting the training and inference of large AI models [4]. - The robotics sector is projected to become a trillion-dollar industry, with Nvidia showcasing its leadership in AI-driven robotics development at COMPUTEX 2025 [4]. Summary by Sections AI Infrastructure - Nvidia's focus on AI data centers as "AI factories" marks a fundamental shift in its business model, highlighting its central role in the AI hardware and software ecosystem [4]. Product Development - The Blackwell architecture and GB300 AI system are set for significant upgrades, enhancing AI computing capabilities and supporting more complex AI models [4]. Robotics - Nvidia's advancements in robotics, including the AI training system and Omniverse platform, position it well for future growth in industrial, medical, and logistics applications [4]. Financial Projections - The report includes a financial forecast for SMIC, projecting an EPS of 0.46 RMB for 2024, increasing to 0.91 RMB by 2026, with a PE ratio decreasing from 178.26 in 2024 to 90.11 in 2026 [5].