英伟达拟投资OpenAI，利好光模块、液冷板块

Group 1 - OpenAI and NVIDIA announced a strategic cooperation to deploy at least 10GW of NVIDIA systems for AI data centers, which is equivalent to 4-5 million GPUs, doubling NVIDIA's total shipments for 2024 [1] - NVIDIA plans to invest up to $100 billion in OpenAI, with the first 1GW data center expected to be deployed in the second half of 2026 [1] - The collaboration complements existing partnerships with Microsoft, Oracle, SoftBank, and the "Stargate" project [1] Group 2 - The demand for AI computing power is driving growth in the optical communication industry, with silicon photonics technology becoming the mainstream solution for 1.6T high-speed optical modules [2] - The competitive advantage of leading optical module companies is expected to strengthen, with increasing demand for CW lasers and certain passive components [2] - New technologies such as OCS, hollow-core fibers, thin-film lithium niobate, and CPO are emerging, presenting new market investment opportunities [2] Group 3 - NVIDIA's new Rubin CPX GPU is designed for large-scale context processing, offering up to 30 Pflops of NVFP4 precision computing power [3] - The Rubin CPX platform integrates multiple GPUs and CPUs, providing significant performance improvements and high memory bandwidth [3] - The introduction of Rubin CPX is expected to accelerate AI application inference, reduce inference costs, and increase demand for liquid cooling solutions [3] Group 4 - The industry outlook remains positive for leading companies in the optical module and liquid cooling sectors, with recommended stocks including Zhongji Xuchuang, Xinyi Sheng, Yingweike, Tianfu Communication, and Yuanjie Technology [4]