Hundred-billion giant hits the daily limit at the close, doubling its stock price this year. Institutions are optimistic about the computing infrastructure sector (with stock picks attached).

On March 24, A-share fiber-optic concept stocks surged. Fiber-optic leader Yangtze Optical Fibre and Cable (YOFC) hit the daily limit in a straight-line rally near the close, closing at 233.31 yuan per share. Its total market value was 1932 billion yuan, with trading volume of 43.21 billion yuan. The stock price has already doubled within the year. In addition, Tongguang Cable rose more than 10% to lead the gains, Tengjing Technology rose more than 9% to set a new all-time high. Several concept stocks, including Xinengtai Shan, Yunnan Germanium, and Fuguang Technology, all hit the daily limit. Multiple other stocks, such as VUC, Hengtong Optic-Electric, and Glux Technology, rose more than 7%.

On the news front, at the 2026 Optical Fiber Communications Conference and Exhibition held last week, all optical communications manufacturers emphasized their plans to expand production of core optical-module components and materials. Among them, Lumentum stated that by the end of 2026, the capacity of EML will increase by more than 50% compared with 2025. The company is one of the key suppliers of optical components for global AI data center infrastructure buildouts.

Guotai Junan Securities believes that at the OFC conference, Lumentum disclosed that capacity would be basically sold out by the end of 2027, EML capacity growth would exceed 50% by the end of fiscal 2026, and that by 2030 the AI data center’s demand for indium phosphide would have a compound annual growth rate of 85%. This directly confirms the rigid demand for high-speed optical modules driven by inter-rack connectivity.

Bank of China International Securities Co., Ltd. noted that CPO, liquid cooling, and LPU rebuild AI computing infrastructure foundations. As agentic AI and physical AI become the development trend in the next stage, CSP providers will shift their focus to the performance and profitability of AI at the inference end. To this end, NVIDIA’s Rubin and Feynman focus on three major changes: LPU, CPO, and liquid cooling. LPU significantly enhances inference performance through high-capacity SRAM. CPO reduces signal-transmission latency through high-speed interconnects. And liquid cooling helps break through the compute density bottleneck via energy-efficiency management.

（Statement: The article content is for reference only and does not constitute investment advice. Investors act at their own risk.）