2025-12-23 09:00Investment Rating - The report rates the industry as "Outperform" [1] Core Insights - The AI infrastructure upgrade wave is driving a revolution in materials, with electronic cloth, copper foil, and resin forming the core dielectric performance barriers for AI PCBs [1][3] - The demand for low dielectric materials is critical for AI PCB design, as GPU and ASIC manufacturers are actively enhancing chip efficiency and interconnect bandwidth [3][13] - The market for AI-related materials is expected to experience rapid growth, with projected global market sizes for HDI boards and high-layer boards reaching approximately $3.098 billion in 2025 and $3.891 billion in 2029 [1][3] Summary by Sections Investment Recommendations - Quartz fiber cloth and low-dielectric electronic cloth are recommended for investment in companies such as Feilihua, Zhongcai Technology, and Honghe Technology. HVLP copper foil investments should focus on Defu Technology, Longyang Electronics, and Tongguan Copper Foil. High-frequency and high-speed resin investments are recommended for Dongcai Technology and Shengquan Group [3] Industry Trends - The AI industry is shifting focus from training to inference, leading to increased demand for AI infrastructure. Major cloud vendors are ramping up capital expenditures to meet this demand, with Alibaba and Tencent expected to spend a total of approximately 380 billion RMB over the next three years [13][14] - The performance requirements for PCBs are evolving, with AI servers requiring more layers and tighter line widths compared to traditional servers. The layer count for AI servers typically ranges from 20 to 30 layers, while traditional servers range from 8 to 22 layers [42][44] Material Innovations - The core materials for M8.5 and M9 PCBs/CCLs are expected to reach a critical point of development, with Nvidia's Rubin server anticipated to adopt advanced materials combinations for its PCB solutions [1][3] - Low dielectric constant (Low-Dk) and low dielectric loss (Low-Df) materials are essential for reducing signal loss and maintaining signal integrity in AI PCBs [1][3]