This sector is currently in a “golden cycle” of explosive growth. Its current state can be summarized in three key points:
Explosive Demand Growth as the Core Engine: The hardware upgrades in AI servers are the primary driver. Unlike standard servers, which typically use 12-layer boards, a single AI server requires a highly complex, multi-layer “layer-cake” design of 18 layers or even 32+ layers, utilizing ultra-low-loss, high-speed transmission materials. This results in a per-unit PCB value that is 5 to 7 times higher than that of a traditional server. From 2023 to 2028, the PCB market for AI/high-performance computing servers is projected to grow at a Compound Annual Growth Rate exceeding 30%, leading growth across the entire industry.
Extremely High Technical Barriers and Concentrated Production Capacity: The requirements for materials, processes, and consistency for these products are exceptionally stringent. For instance, the dielectric loss factor must be less than 0.002, necessitating globally top-tier ultra-low-loss materials (e.g., MEGTRON 7/8 grades). Process-wise, lamination precision errors must be controlled at the micron level, and laser-drilled via diameters need to be as small as 50 micrometers. This means only a handful of manufacturers worldwide possess the capability for stable mass production. These technical barriers make it difficult for high-end production capacity to scale up quickly to meet the surging demand in the short term.
Persistent Supply-Demand Imbalance with Expansion Led by Market Leaders: Due to the factors above, the market exhibits a clear state of supply shortages. Industry leaders such as Wus Printed Circuit Co., Ltd. (a core supplier of server motherboards for NVIDIA) and Shengyi Technology Co., Ltd. (whose S8/S9 materials have entered NVIDIA’s supply chain) are operating with full order books and commanding solid gross margins. To seize this opportunity, leading manufacturers are strategically directing their new production capacity towards high-end categories with 18 layers or more and are accelerating their overseas expansion. According to industry analysis, the tight supply-demand situation in the high-end AI PCB market is expected to persist at least until 2026.
The Current State of Traditional High-Frequency PCBs: Demand Pressure and Homogeneous Competition
Relatively Moderate Growth and Structural Constraints: This segment serves traditional communications, some consumer electronics, and industrial control markets. End-user demand in these areas is recovering weakly under macroeconomic influences, with growth far less robust than in the AI computing sector. As industry capital, materials, and high-end production capacity tilt toward the AI segment, the resources allocated to traditional areas naturally become constrained.
Fragmented Competitive Landscape and Limited Bargaining Power: The market has numerous players, and product homogenization is high, leading to intense competition. This results in limited room for companies to negotiate with both upstream and downstream partners. On one hand, they must bear pressure from price fluctuations of upstream raw materials (such as copper and glass fiber cloth). On the other hand, when demand is weak, they are prone to price wars, which squeeze profit margins.
Conclusion and Essence: The Industrial Upgrade from “Following” to “Leading”
The fundamental difference between these two segments lies in whether they are part of a new value chain defined by technological innovation.
AI/computing high-frequency, high-speed PCBs represent the “leading” future. They are defined by cutting-edge technological demands (such as PCIe 5.0/6.0, 112G/224Gbps transmission) and are characterized by high added value, significant technical barriers, and strong customer stickiness. This represents the direction of upward industry value.
Traditional high-frequency PCBs, meanwhile, are situated more in a mature “following” market. Their technical standards are relatively stable, and competition revolves more around cost, scale, and delivery.