Nvidia(US:NVDA) Minsheng Securities·2025-05-29 11:57Investment Rating - The report recommends a positive investment outlook for key companies in the AI infrastructure sector, highlighting specific stocks such as 瑞可达, 博创科技, 禾望电气, 江海股份, 申菱环境, and 芯原股份 as potential investment opportunities [4][46][47]. Core Insights - The ComputeX 2025 showcased NVIDIA's strategic shift from a hardware supplier to a global AI ecosystem builder, introducing groundbreaking technologies like the Grace Blackwell supercomputer and NVLink Fusion [1][10]. - The collaboration between Vertiv and NVIDIA aims to address energy bottlenecks in AI infrastructure through the development of an 800V HVDC power solution, enhancing energy efficiency and system reliability [2][17]. - NVIDIA's NVLink Fusion technology redefines the collaboration paradigm between ASIC and GPU, allowing for seamless communication and integration of third-party ASICs into NVIDIA's ecosystem [3][31]. - The expansion of NVIDIA's supply chain to 123 partners reflects its growing influence in the AI sector and the diversification of its collaborative efforts across various industries [4][37]. Summary by Sections 1. NVIDIA ComputeX Showcases Next-Gen AI Solutions - NVIDIA's GB300 AI computing platform is set to deliver 40 petaflops of performance by Q3 2025, significantly enhancing inference capabilities [10]. - The introduction of the DGX personal computing system allows users to run large AI models locally, democratizing access to AI computing [11][14]. 2. HVDC: Energy Foundation for AI Computing Era - The strategic partnership with Vertiv focuses on developing an 800V HVDC power solution, which is expected to improve energy efficiency by 5% and reduce copper usage by 45% [18][25]. - The collaboration with Navitas Semiconductor aims to leverage third-generation semiconductor technology to enhance the efficiency and reliability of AI infrastructure [26][29]. 3. NVLink Fusion: Embracing the ASIC Ecosystem - NVLink Fusion enables the integration of custom ASICs with NVIDIA's GPUs, facilitating high-bandwidth communication and enhancing the performance of AI applications [31][34]. - The technology supports a throughput of 800 Gb/s and allows for the construction of semi-custom AI servers [32][36]. 4. New Suppliers Join NVIDIA Ecosystem - The addition of new partners across various sectors, including education, healthcare, and finance, signifies NVIDIA's commitment to expanding its AI applications [38][41]. - 瑞可达's involvement in providing high-speed data transmission solutions is crucial for meeting the demands of AI data centers [43][44]. 5. Investment Recommendations - The report emphasizes the importance of focusing on companies that enhance speed and power in AI infrastructure, recommending specific stocks for investment consideration [46][47].