Industry Rating - The industry investment rating is maintained at "Outperform the Market" [1] Core Viewpoints - CPO technology integrates ASIC chips and optical engines on the same high-speed motherboard, reducing signal attenuation, system power consumption, and costs while achieving high integration, applicable in data centers, high-performance computing, AI, and virtual reality [2] - The CPO market is expected to grow rapidly, with Yole predicting a market size of $2.6 billion by 2033, driven by AI demand and requiring collaboration across the industry chain [2] - Major overseas companies are leading the CPO industry development, with significant advancements from Broadcom, Marvell, TSMC, and NVIDIA [2][3] Summary by Sections 1. CPO Technology and Development Trends - CPO technology addresses bandwidth and power challenges, with a 25%-30% reduction in total power consumption for 51.2T switches compared to traditional solutions [12] - CPO technology offers low power consumption, low latency, high bandwidth, improved system performance, and cost-effectiveness [15] - Silicon photonics is becoming the main path for CPO technology due to its high integration and compatibility with CMOS processes [18] 2. CPO Market Size and Major Drivers - The CPO market is projected to grow at a CAGR of 69% from $600,000 in 2022 to $287 million in 2033, while OIO is expected to grow at a CAGR of 81% from $500,000 to $2.3 billion [31] - Major players in the CPO market include Broadcom, Intel, NVIDIA, and Marvell, with significant contributions from laser manufacturers and silicon photonics foundries [35] 3. CPO Switch Internal Structure - CPO switches consist of key components such as ASIC chips, optical engines, external light sources, and flexible optical backplanes [63] - The optical engine is crucial for optical-electrical conversion, with various packaging methods impacting performance and integration [67] 4. Recommended Companies to Watch - Companies actively involved in CPO technology include: - Zhongji Xuchuang, focusing on high-end optical communication modules and CPO-related products [96] - Xinyi Sheng, leveraging silicon photonics technology for CPO applications [101] - Tianfu Communication, developing multiple projects related to CPO components [106] - Taichuang, specializing in optical communication devices and flexible boards [111] - Yuanjie Technology, providing CW light sources for CPO applications [116] - Shijia Photon, offering components for CPO packaging [121] - Guangxun Technology, launching CPO ELS light source modules [126] - Robotech, focusing on silicon photonics and CPO packaging equipment [131] - Ruijie Network, releasing CPO switches with significant market presence [136]

Investment Rating - The industry maintains a "stronger than market" rating, indicating an expectation that the overall industry performance will outperform the market in the next six months [6][7]. Core Insights - NVIDIA's new generation of AI chips, announced at the GTC 2025 conference, significantly enhances computing power, creating new opportunities in the AI supply chain [2][6]. - The Blackwell Ultra and Vera Rubin platforms show performance improvements of 1.5 times and 3.3 times, respectively, with the latter expected to launch in the second half of 2026 [2][3]. - The demand for HBM memory is projected to grow substantially, with a near 200% growth rate expected in 2024 and a potential doubling in 2025 [3]. Summary by Sections Investment Ratings for Specific Companies - Luxshare Precision (002475.SZ) is rated as "Buy" with an expected EPS of 1.9 in 2024 and 2.38 in 2025, and a PE ratio of 21.44 for 2024 and 17.11 for 2025 [1]. - Jiangbolong (301308.SZ) is rated as "Add" with an expected EPS of 1.13 in 2024 and 2.37 in 2025, and a PE ratio of 93.69 for 2024 and 44.67 for 2025 [1]. - Lianrui New Materials (688300.SH) is also rated as "Add" with an expected EPS of 1.35 in 2024 and 1.8 in 2025, and a PE ratio of 43.73 for 2024 and 32.8 for 2025 [1]. Technological Developments - The introduction of CPO technology and optical networks is becoming a focal point in data center construction, enhancing the value of the optical communication industry chain [4]. - The global optical transceiver market is expected to grow by USD 12.56 billion from 2025 to 2029, with a compound annual growth rate (CAGR) of 17.5% [4]. Cooling Solutions - Liquid cooling technology is becoming essential for AI data centers, with NVIDIA showcasing advanced liquid cooling management systems that significantly improve thermal efficiency [5]. - The power requirements for the new Rubin Ultra architecture are projected to be approximately 4.5 times higher than the previous Blackwell architecture, necessitating advancements in power supply and cooling systems [5].

Core Insights - The event highlighted the transformative shift in data centers towards AI-driven computing, with projected capital expenditures exceeding $1 trillion by 2028 for data center construction, primarily focused on accelerated computing chips [2][12][13] - NVIDIA's Blackwell architecture is fully operational, showcasing significant performance improvements and a roadmap for future products like Rubin and Feynman, which promise substantial enhancements in computational power and efficiency [3][42][45] - The introduction of the Quantum-X CPO switch and Spectrum-X technology aims to revolutionize networking capabilities, reducing energy consumption and increasing deployment efficiency [5][46] - The advancements in AI applications, particularly in autonomous driving and robotics, are supported by NVIDIA's new systems and frameworks, enhancing the development and training processes [6][26][24] Capital Expenditure and AI Infrastructure - Data center capital expenditures are expected to reach $1 trillion by 2028, with a significant portion allocated to accelerated computing chips [2][12] - NVIDIA plans to deliver 1.3 million Hopper GPUs to major cloud service providers in 2024, with an increase to 3.6 million Blackwell GPUs in 2025 [2][3] AI Model Training and Inference - The demand for computational power for AI training and inference has surged, with estimates suggesting a 100-fold increase in required computing resources compared to the previous year [10][11] - NVIDIA outlines three levels of AI: Generative AI, Agentic AI, and Physical AI, each representing a different stage of AI development and application [8][10] Product Development and Future Roadmap - Blackwell has been fully launched, with significant customer demand and performance improvements, including a 40-fold increase in inference performance compared to previous models [3][42] - Future products like Vera Rubin and Rubin Ultra are set to enhance computational capabilities further, with expected performance increases of up to 15 times [45][42] Networking Innovations - The Quantum-X CPO switch is anticipated to launch in late 2025, offering substantial energy savings and improved network efficiency [5][46] - Spectrum-X technology will provide high bandwidth and low latency, integrating seamlessly into NVIDIA's computing architecture [5][46] AI Applications in Autonomous Driving and Robotics - NVIDIA's Halos system aims to enhance safety in autonomous vehicles, while the open-source Isaac Groot N1 model supports robotics development [6][24] - The integration of Omniverse and Cosmos platforms accelerates the development of AI for autonomous driving, enabling end-to-end training capabilities [26][24] Data Center Evolution - The transition of data centers into AI factories is underway, focusing on processing, analyzing, and generating AI-driven applications [12][13] - NVIDIA's Dynamo operating system is designed to optimize AI factory operations, enhancing efficiency and performance [35][36]