AI算力行业周报：Meta 27年底前推出四代自研AI芯片，OFC 2026大会于洛杉矶启幕-20260317

Investment Rating - The report maintains a "Buy" rating for the companies mentioned, including 汇绿生态, 沪电股份, 天孚通信, 太辰光, while 长电科技 remains unrated [6]. Core Insights - Meta plans to launch four generations of self-developed AI chips by the end of 2027 to support its growing AI computing needs and reduce reliance on external suppliers. The chips include MTIA 300, MTIA 400, MTIA 450, and MTIA 500, with MTIA 300 already in mass production [3]. - The OFC 2026 conference in Los Angeles is expected to attract 16,000 participants and over 700 exhibitors, focusing on AI, optical innovation, and space optical networks [4]. - The report suggests focusing on companies such as 沪电股份, 长电科技, 天孚通信, 汇绿生态, and 太辰光 for potential investment opportunities [5]. Weekly Market Analysis - The electronic industry saw a decline of 1.23% from March 9 to March 13, ranking 20th among the 31 sectors, while the communication industry decreased by 0.12%, ranking 11th [12][15]. - The AI computing sector showed mixed performance, with the printed circuit board (PCB) sector increasing by 2.95%, while other power supply equipment sectors decreased by 4.61% [19]. Company Focus and Earnings Forecast - The earnings per share (EPS) forecasts for 2024, 2025E, and 2026E for the companies are as follows: - 汇绿生态: 0.08, 0.11, 0.22 with PE ratios of 540, 384, and 198.34 respectively [6] - 沪电股份: 1.35, 1.94, 2.61 with PE ratios of 60.13, 41.85, and 31.1 respectively [6] - 天孚通信: 2.43, 2.61, 4.18 with PE ratios of 133.74, 124.52, and 77.75 respectively [6] - 太辰光: 1.15, 1.83, 3.01 with PE ratios of 111.22, 70.06, and 42.49 respectively [6] - 长电科技: 0.90, 0.88, 1.19 with no rating [6]. Industry Dynamics - Tesla is advancing a large AI chip manufacturing project to support its AI and autonomous driving needs [45]. - Applied Materials and Micron are collaborating to build a semiconductor R&D center with an investment of $5 billion to develop next-generation AI storage solutions [46]. - The U.S. Department of Commerce has withdrawn a proposed rule on AI chip exports, reflecting internal disagreements on balancing national security and global competitiveness in the AI sector [47]. - Meta's announcement of new AI chips and a significant capital expenditure plan of $115 billion to $135 billion for expanding AI data center infrastructure indicates a growing trend among major tech companies to develop in-house AI capabilities [48].