广发基金刘玉：科技成长预计仍是主线 关注AI算力基础设施方向

Core Insights - The AI computing power sector is experiencing significant positive developments, with companies in the AI computing power chain being direct beneficiaries [1] - CPO (Coherent Photonic Optics) is identified as a core segment with high performance elasticity and certainty, benefiting from the capital expenditures of major overseas cloud giants like Microsoft, Google, and Meta [1] - The domestic computing power chain is accelerating its technological catch-up and market share increase under the strategy of self-control [1] Company Performance - Key players in the CPO industry include New Yisheng (300502), Zhongji Xuchuang (300308), and Tianfu Communication (300394), which have established strong competitive advantages through technology barriers and market share [1] - As of December 10, 2023, the stock price increases for these companies were 425%, 401%, and 263% respectively [1] - Funds heavily invested in these sectors, such as the Guangfa Emerging Growth Fund managed by Liu Yu, reported a year-to-date return of 73.41% as of December 10, 2023, and a cumulative return of 162.31% since October 2018 [1] Investment Strategy - The Guangfa Emerging Growth Fund's portfolio is heavily concentrated in the AI computing power sector, with over 25% of holdings in "Yizhongtian" and approximately 61% in the top ten holdings, including Industrial Fulian (601138) and Yingweike (002837) [2] - Liu Yu emphasizes a strategy focused on high performance certainty and significant future growth potential, adjusting the portfolio dynamically based on industry changes [2] - The AI industry is still in its early stages, with expected continuous demand growth for computing power infrastructure as AI models evolve and applications are implemented [2] Sector Focus - Liu Yu highlights three key sub-sectors within the computing power chain: 1. CPO (Optical Modules), with Nvidia planning to mass-produce CPO switches, leading to significant annual shipment growth [3] 2. OCS (Optical Circuit Switching), which offers high bandwidth and low latency, gaining attention due to Google's application in TPU architecture [3] 3. PCB (Printed Circuit Boards), crucial for AI server upgrades, with potential shifts in technology that require ongoing monitoring [3]