China Galaxy Securities has released a report stating that AI computing chip upgrades and AI edge devices are driving the advancement of copper-clad laminates (CCL) and printed circuit boards (PCBs). Demand from downstream sectors such as AI servers, data storage, and network communications is expected to grow steadily under AI's influence, propelling high-layer PCBs and HDI boards to maintain strong growth and leading to a simultaneous rise in volume and price for AI PCB products. The push for high-frequency, high-speed, and low-loss performance is driving material upgrades in PCBs, benefiting related electronic fabrics, copper foils, and resins. As PCBs evolve toward higher layers and greater integration, their manufacturing processes and equipment must also be upgraded. The global market share of domestically produced specialty CCL stands at only 8.3%, indicating significant room for import substitution.
AI Development Fuels Rapid Growth in Demand for HDI and High-Layer Boards
According to Prismark, the global PCB market is projected to reach $109.258 billion by 2029, with a compound annual growth rate (CAGR) of 8.2% from 2024 to 2029. In 2025, China's mainland PCB output is expected to be $48.459 billion, representing approximately 57% of the global total. By 2029, China's PCB market size is forecast to reach $62.463 billion, with a CAGR of 8.7% from 2024 to 2029. In 2025, global HDI output is estimated at $15.717 billion, up 25.6% year-on-year, while global multilayer board output is projected at $33.091 billion, up 18.2% year-on-year. Prismark data indicates that the AI server-related PCB market will achieve a CAGR of 18.7% from 2024 to 2029. Within this, AI server-related HDI is expected to grow at a 29.6% CAGR, and AI-related multilayer boards with 18 or more layers at a 33.8% CAGR, far exceeding the PCB industry average of 8.2%. Demand from downstream sectors like servers/data storage, network communications, and automotive electronics will continue to grow under AI's impetus, driving sustained high growth for high-layer and HDI boards.
AI Computing Chip Upgrades Drive CCL and PCB Advancements, with Significant Import Substitution Potential for Specialty CCL
Nvidia's Rubin chip uses a core PCB with over 20 layers, and the CCL has been upgraded to M9 grade. According to TrendForce, the Rubin platform employs comprehensive material upgrades to achieve low loss and low latency. This includes the Switch Tray using M8U grade (Low-DK2+HVLP4) and a 24-layer HDI board design, while the Midplane and CX9/CPX adopt M9 (Q-glass+HVLP4). The Midplane's ultra-high multilayer PCB has 44 layers. The PCB value per server is more than double that of the previous generation. ASIC AI servers like Google's TPU V7 and AWS's Trainium3 also incorporate high-layer HDI, low-Dk materials, and ultra-low roughness copper foils. As AI computing chips are upgraded, CCL materials are correspondingly improved, and the number of core PCB layers increases. Edge AI is characterized by high-performance SoCs/NPUs, high-density BGAs, transient high currents, and multi-sensor RF mixed signals, leading to comprehensive upgrades in PCB materials, stacking, HDI processes, and power delivery networks (PDN). Demand for high-layer, high-frequency, high-speed, and high-density interconnect PCBs in AI edge devices is surging. In 2024, China's CCL production capacity accounted for 73.3% of the global total, but the domestic CCL global market share is only about 20%, and the global market share for domestically produced specialty CCL is just around 8.3%, highlighting significant potential for import substitution.
High-Frequency, High-Speed, Low-Loss Performance Drives Upgrades in Copper Foil and Electronic Fabric for PCBs
PCBs transmitting high-speed signals at frequencies above 1 GHz require high-frequency, high-speed materials. The copper foil industry is currently evolving from HVLP2/3 to HVLP4/5 and profile-free copper foils. According to Guanyan Tianxia, AI server demand for HVLP copper foil is surging, with usage per unit being eight times that of traditional servers. Nvidia's new Rubin platform uses HVLP5 copper foil with PTFE substrates, driving up value. Key A-share copper foil companies include Tongguan Copper Foil, Defu Technology, Zhongyi Technology, Haixing Shares, and Nuode Shares. To reduce high-frequency signal transmission loss and lower the coefficient of thermal expansion, low-dielectric electronic fabrics have been widely adopted in recent years in high-frequency, high-speed communication fields such as AI servers, data center switches, and optical modules. Low thermal expansion coefficient electronic fabrics are now being promoted for use in AI chip packaging substrates and high-end mobile phone chip packaging substrates. The prospects for second-generation Low Dk/Df electronic fabrics, Low CTE electronic fabrics, and Q-fabrics are promising. Major domestic companies listed in the electronic fabric sector include Honghe Technology, Sinoma Science & Technology, China Jushi, and International Composites. Quartz fiber fabrics offer excellent dielectric properties and thermal stability, primarily used in specialized fields like satellite communications and aerospace engineering.
As PCBs Upgrade to Higher Layers and Greater Integration, Manufacturing Processes and Equipment Follow Suit
PCB manufacturing processes have now reached the micron level. The Modified Semi-Additive Process (MSAP) can achieve high-precision, high-density circuits (e.g., 0.018mm line width) while maintaining mass production feasibility and has been used to manufacture substrate-like PCBs. HDI boards require higher impedance stability, and their fine lines and thin dielectric layers make impedance control more challenging. Laser drilling technology is the core process for creating blind and buried vias (key HDI board structures), capable of drilling holes as fine as 0.05mm. Companies like Shengyi Technology, Redboard Technology, and Zhongjing Electronics can perform laser drilling below 0.07mm. Processing multilayer PCBs significantly reduces the efficiency of mechanical drilling, while simultaneously increasing the demand for laser drilling.
Risk Disclosures
Risks include the potential for PCB manufacturers' new orders and execution falling short of expectations; risks related to AI chip demand and mass production not meeting forecasts; risks of accounts receivable recovery being less than anticipated; and risks of AI application commercialization not progressing as expected.
Comments