计算机行业全球科技动态追踪：DeepSeek火爆全球，AI即将迎来新一轮繁荣期

Investment Rating - The report maintains a "Recommended" rating for the computer industry [5]. Core Insights - DeepSeek is rapidly gaining popularity globally, significantly impacting the AI industry landscape, with AI applications like AI Agents and edge AI expected to experience explosive growth. As of February 9, DeepSeek has begun operations in Saudi Arabia and has surpassed 30 million daily active users [2][25]. - Major US tech companies are increasing their capital expenditures on infrastructure, reflecting optimism about the future of AI. Microsoft, Amazon, Google, and Meta are projected to reach record capital expenditures in 2024, with Amazon planning to invest up to $100 billion primarily for data center construction [2][17]. - DeepSeek has innovatively reduced the computational power requirements for training and inference, lowering costs to an industry-low level, while still indicating that computational demand will continue to grow in the long term. This low-cost model is expected to drive widespread application adoption, with inference power becoming the primary driver of computational consumption [3][25]. Summary by Sections Global Market Performance - From February 5 to 11, the US tech market experienced a slight pullback, while the A-share computer sector saw a cumulative increase of 15.65%. The S&P 500 index rose by 0.51%, and the NASDAQ composite index fell by 0.05% [12][13]. Industry News Computing Power and End Devices - CoreWeave has launched new instances based on NVIDIA's GB200 NVL72, enhancing AI inference model development. These instances support up to 110,000 GPUs and are designed for next-generation AI applications [19][20]. - NVIDIA's new RTX 50 series graphics cards are optimized for generative AI workloads, providing significant performance improvements and enabling local processing of AI tasks [21][22]. Large Models and Cloud Applications - DeepSeek has emerged as a leading AI application, achieving over 20 million daily active users within 20 days of launch, significantly outpacing competitors like ChatGPT. Its training costs are substantially lower than those of OpenAI, making AI development more accessible [23][24].