全球科技(计算机)行业周报:英伟达VeraRubin平台量产,驱动AI应用规模化普及-20260112

Investment Rating - Industry rating: Overweight [1] Core Insights - On January 6, 2026, NVIDIA CEO Jensen Huang officially launched the latest NVIDIA Rubin platform at CES 2026, stating that it has entered full-scale production. The Rubin platform consists of six new chips: Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-6 Ethernet switch, designed to form an AI supercomputer that significantly reduces AI training time and lowers inference token generation costs. The training performance of the Rubin platform is 3.5 times that of the previous Blackwell generation, with software performance improved by 5 times, and the cost per token for inference reduced by 10 times compared to the Blackwell platform. Additionally, the number of GPUs required for training MoE models is reduced to one-fourth of the previous requirement [3][13][15]. - The Rubin platform integrates five key technologies to achieve breakthroughs in performance enhancement and cost reduction: 1) New NVLink interconnect technology ensures low latency and high bandwidth during multi-chip collaboration; 2) The third-generation Transformer engine optimizes for AI tasks, greatly improving model training and inference efficiency; 3) Confidential computing technology provides end-to-end security for sensitive AI data, meeting compliance needs in finance and healthcare; 4) RAS engine ensures stable performance under 24/7 high-load operation; 5) Vera CPU is specifically designed for agent inference [4][13]. - Major cloud providers, including Amazon AWS, Google Cloud, Microsoft Azure, and Oracle Cloud, have confirmed plans to deploy instances based on Vera Rubin in 2026, allowing global users to access top-tier AI computing power through cloud services. This will enable AI startups and SMEs, previously limited by high computing costs, to easily utilize the Rubin platform's powerful capabilities without significant hardware investment. Efficient computing power will accelerate breakthroughs in basic scientific research and technology, promoting deep applications of AI in healthcare, education, and environmental protection [5][14]. - The Rubin platform's dual breakthroughs in performance and cost reduction will significantly lower hardware procurement and operational costs for AI companies, facilitating the widespread adoption of AI applications such as intelligent customer service, autonomous driving, drug development, industrial quality inspection, and scientific research. It is recommended to pay attention to domestic and international AI computing infrastructure and related companies, including Cambrian, Zhongke Shuguang, Yonyou Network, Dingjie Zhizhi, Kingsoft Office, Tonghuashun, Jiao Dian Technology, Saiyi Information, and Fubo Group [6][15]. Market Overview - As of the week, the Shanghai Composite Index rose by 3.82%, the ChiNext Index increased by 3.89%, and the CSI 300 Index went up by 2.79%. The computer industry index surged by 8.49%, outperforming the Shanghai Composite Index by 4.67 percentage points, the ChiNext Index by 4.60 percentage points, and the CSI 300 Index by 5.70 percentage points. Year-to-date, the computer industry index has also increased by 8.49% [17][20]. - The computer industry index ranked 5th among 31 industry indices this week and 2nd among the four major TMT industries (electronics, communications, computers, media) [17]. Company Dynamics - The report highlights significant movements in the software and information technology sectors, with companies like Huijin Technology, Ersan Siwu, and Zhuoyi Information showing notable performance. Future investment opportunities are suggested in the financial IT, industrial software, and trusted innovation sectors, which are expected to see upward trends [22].