重點摘要

AI 運算需求正在重塑 PCB 產業，朝向具平台級特性的產品升級，表現為更多層數、更高密度、更快訊號傳輸與更佳可靠性。 這種由量轉質的結構性變化推動高層數與高階 HDI 板的快速採用，並迫使全製程設備升級。 材料正從傳統 FR-4 轉向 M7–M9 的低損耗層壓板，且像 mSAP 之類的精密製程正在加速，為設備供應商與耗材製造商創造成長潛力。

情緒分析

• 整體情緒：正面偏中性。報告指出 AI 伺服器部署與技術升級為未來數年高階 PCB、設備與耗材需求帶來明確的成長動能。然而，也提示與資本支出循環及技術路徑演進相關的執行與需求風險。
  70%

文章正文

The rise of AI workloads is becoming the dominant structural driver for the printed circuit board (PCB) industry, prompting an upgrade in product requirements toward platform-level designs characterized by more layers, higher density, faster signaling, lower loss and greater reliability. As AI server architectures shift from CPU-centric systems to GPU/ASIC clusters, the technical expectations for PCBs — in terms of material performance, layer count, hole structures and trace precision — are rising sharply. This evolution is accelerating demand for high-multilayer boards, advanced HDI (high-density interconnect) boards and high-speed/high-frequency laminates.

On the demand side, large-scale deployment of AI servers is pushing PCB demand beyond traditional motherboards to a broader range of function-specific boards: compute cards, switch fabrics, network boards, power distribution boards and liquid-cooling control boards. Industry forecasts indicate that AI server shipments could grow substantially faster than general server shipments, underpinning a structural increase in higher-end PCB volumes. Typical layer counts for AI-related PCBs are increasing into the 20–30 layer range or above, with tighter impedance control and stronger crosstalk suppression requirements, which drives a migration from mid/low-end general-purpose boards to advanced high-multilayer and HDI solutions.

On the product side, high-multilayer PCBs improve signal integrity, power distribution and electromagnetic compatibility by adding routing layers. HDI technology, using microvias, blind vias and buried vias, boosts routing density per unit area and is particularly important for GPUs, AI accelerator cards and high-speed communication modules. At the process level, precision techniques such as mSAP (modified semi-additive process) reduce the side-etch effects of traditional subtractive methods and enable finer-line manufacturing. Concurrently, materials are evolving: as SerDes speeds move from 112G toward 224G and higher aggregates, the market is shifting from standard FR-4 to M7–M9 low-loss and ultra-low-loss copper-clad laminates, while glass fiber, copper foil and resin systems are being upgraded to meet electrical and thermal requirements. This material and process migration is a core enabler of higher-performance AI PCBs.

From the equipment perspective, PCB high-endification compels upgrades across the full process chain, creating opportunities for both capital equipment and consumables. In drilling, conventional mechanical drills must adapt to thicker multilayer stacks, back-drilling and higher positional accuracy, raising the unit value of CCD-assisted mechanical drilling systems; laser drilling benefits from the microvia requirements of HDI processes. In imaging, LDI (laser direct imaging) offers superior resolution, registration accuracy and flexible production capabilities, and demand for high-end LDI tools is expanding. In plating, the surge in multilayer and HDI substrates and demand from substrate-like PCB applications drives higher penetration of VCP (vertical continuous plating) technologies; mSAP adoption further supports the rise of horizontal three-in-one and transfer VCP systems, accelerating local substitution and lifting equipment average selling prices.

Consumables also stand to benefit. The combined effects of an increased number of holes per board, a higher number of drill hits per hole and shorter drill bit lifetimes create a multiplier for drill-bit demand tied to AI server PCBs—both in volume and in share of high-performance drill bits. As AI PCB process nodes advance, the share of premium drill bits rises, supporting simultaneous increases in quantity and price. Other consumables such as advanced laminates, specialized copper foils and high-performance resins will similarly see growing demand aligned with the shift to low-loss materials and finer-line processes.

Risks remain. A slowdown in AI-related capital expenditure by major cloud providers or lower-than-expected AI server shipments could weaken demand for high-multilayer, HDI and high-frequency PCBs. Expansion by PCB manufacturers is also execution-sensitive: capacity additions for high-end products require customer qualification, yield ramp-up, production line build-out and significant capital, meaning slower downstream expansion could delay equipment order flows. Finally, technology routes for AI servers, PCB processes and material systems continue to evolve; if equipment vendors fail to keep pace with needed upgrades or if pricing competition intensifies, revenue and margin outcomes for related suppliers could be adversely affected.

In summary, AI-driven compute demand is steering the PCB industry into a phase of structural upgrade where quality and technical complexity replace pure volume growth. This trend supports robust opportunities for advanced production equipment and specialized consumables, though outcomes will depend on capex trajectories, execution on capacity expansions and ongoing technology development across the ecosystem.

關鍵洞見表