通信行业点评：海外AI逐步闭环，继续看好AI基建带来高增长

Investment Rating - The report maintains a "Recommended" rating for several key companies in the AI infrastructure sector, including 德科立, 仕佳光子, 中际旭创, 新易盛, 天孚通信, and 润泽科技 [5]. Core Insights - The report highlights a significant investment by NVIDIA in OpenAI, amounting to up to $100 billion (approximately 711.47 billion RMB), aimed at deploying at least 10GW of NVIDIA systems, which includes millions of GPUs [1][2]. - The transition from "selling shovels" to "delivering shovels" is noted, with the estimated cost of building 1GW of computing power being around $50-60 billion, where NVIDIA's chips and systems account for approximately $35 billion [2]. - The AI sector is entering a phase focused on infrastructure expansion rather than just model competition, with OpenAI's mission to drive "superintelligence" being heavily reliant on computational power [3]. - The report suggests that the current focus in AI infrastructure has shifted from cloud vendor spending to AIDC cloud computing centers, recommending companies with a first-mover advantage in GPU cards [4]. Summary by Sections Investment Opportunities - Key companies to watch include Oracle, Coreweave, Nebius, AppliedGigital, and Iris Energy for their GPU card advantages [4]. - In the domestic market, companies like 润泽科技, 奥飞数据, and 润建股份 are highlighted for their potential in AI infrastructure [4]. Financial Projections - The report provides earnings per share (EPS) and price-to-earnings (PE) ratios for several companies, indicating strong growth potential for 2024-2026 [5]. - For example, 中际旭创 is projected to have an EPS of 12.94 RMB by 2026 with a PE ratio of 34, while 新易盛 is expected to reach an EPS of 13.10 RMB with a PE of 27 [5]. Market Trends - The report emphasizes the increasing demand for AI infrastructure, with OpenAI's rapid growth and significant contracts with cloud service providers indicating a robust market environment [3]. - The establishment of a 1GW data center by OpenAI in 2026 is anticipated to mark a milestone in the infrastructure arms race, paving the way for large-scale application deployment [3].