科创综指收涨1.29% AI硬件引领结构性行情

Market Overview - The A-share market exhibited a mixed and volatile pattern, with major indices showing varied performance. The Shanghai Composite Index closed at 4146.63 points, down 0.01%, while the Shenzhen Component Index rose 0.19% to 14503.79 points. The ChiNext Index fell 0.29% to 3344.98 points, and the Sci-Tech Innovation Index increased by 1.29% to 1843.20 points. The total market turnover reached 25.566 billion yuan, an increase of 0.757 billion yuan compared to the previous trading day [1]. AI Hardware Sector - The AI hardware sector experienced a collective surge, driven by Nvidia's better-than-expected earnings report, which activated market sentiment. The CPO sector saw significant gains, with stocks like Jieput and Jufei Optoelectronics hitting the daily limit of 20%. Leading stock Tianfu Communication reached a historical high, closing at 362.02 yuan, up 6.01%, indicating high sector vitality [2]. - In addition to the CPO sector, related segments such as copper cable high-speed connections, liquid-cooled servers, and PCBs also performed well. Stocks like Chuanrun and Jialitu hit the daily limit, while leading companies Shenghong Technology and Dongshan Precision rose by 7.75% and 8.01%, respectively, showcasing a strong collaborative upward trend across the sector [2]. - The ongoing AI trend positively impacted computing chip stocks, with leading stock Cambricon rising by 7.96% and achieving a total trading volume of 16 billion yuan. Other computing chip-related stocks, including Muxi and Chipone, also saw varying degrees of increase [2]. Power Grid Equipment Sector - The power grid equipment sector also showed strong activity, benefiting from the AI wave. Stocks like Beijing Keri, Shima Power, and Xinlian Electronics hit the daily limit. The global demand for power grid equipment is robust, with AI data centers expected to significantly increase electricity demand. According to IEA data, global power grid investment is projected to reach 390 billion USD in 2024 and exceed 400 billion USD in 2025 [3]. - The energy infrastructure in the U.S. is largely below standard, and the increasing electricity demand from AI is likely to initiate a mandatory upgrade cycle for U.S. power grid equipment. The electricity consumption of global data centers is expected to rise from approximately 415 TWh in 2024 to 945 TWh by 2030, with an annual growth rate of about 15% [3]. - Longcheng Securities' research indicates that the significant growth in electricity demand from AI data centers is likely to directly drive the demand for transformers and high-voltage cables, benefiting domestic export enterprises. Customs data shows that the total export value of transformers in 2025 is expected to reach 9.036 billion USD, setting a historical high [3]. Opportunities in Interconnect Density - In the context of sustained high demand in the computing sector, institutions are highly focused on investment opportunities in domestic computing capabilities. CITIC Securities reports that during the Spring Festival, Alibaba and Tencent launched significant NPO achievements, including Alibaba Cloud's all-optical Scale-up network architecture UPN512, which connects xPU and switches directly, significantly reducing wiring complexity, heat load, power requirements, and operational costs, with power consumption reduced by 50% and costs down by 30% [4]. - Domestic computing capabilities, leveraging cost and ecosystem advantages, are expected to evolve from serving only domestic large enterprises to potentially serving global users through domestic AI models, thus expanding the demand ceiling for domestic computing [4]. - CITIC Securities emphasizes that the super-node architecture is essential for domestic computing development to catch up, with cloud vendors and equipment manufacturers accelerating the adaptation of open protocols. There is a recommendation to focus on the value re-evaluation opportunities brought by increased interconnect density, including optical communication, high-speed line modules, switching chips, and data centers [4].