The next major pressure point in the AI chip cycle is no longer just in GPUs, advanced packaging, or data center power. It’s beginning to shift to a less visible but critical layer for the entire industry: materials and substrates for PCB and IC substrates. In Taiwan, several analysts and market panels are indicating that the expansion of AI servers, high-speed switches, and new data centers is starting to tighten the supply chain from the bottom up, especially in ABF substrates, CCL, and related advanced materials. This aligns with a broader context of significant investments in infrastructure: BloombergNEF estimates capital expenditures (capex) for major data center operators to approach $750 billion by 2026, while Dell’Oro projects global sector capex reaching $1.7 trillion in 2030 driven by AI growth.
The core idea is that pressure is no longer solely on chips. As accelerators grow in complexity, size, and interconnection speeds, the value and technical difficulty of supporting substrates, boards, and materials also increase. This impacts the entire stack: copper, fiberglass, high-grade CCL, precision drill bits, ABF substrates, and PCB manufacturers with higher layer counts and densities. Several Taiwanese companies are already describing this fundamental shift in their communications. For instance, Nan Ya Plastics indicated in 2025 that investments and upgrades in AI servers, data centers, and high-speed networking equipment were boosting demand for ABF substrates, CCL, fiberglass cloth, and copper foil.
This point is significant because it confirms—directly from an industry source—that demand is no longer confined to silicon. It also explains why the market is increasingly paying attention to names like Nan Ya PCB, Unimicron, Kinsus, Zhen Ding, Taiwan Glass, Topoint, EMC/TUC, Gold Circuit, and other Taiwanese ecosystem players. Many of these are not new companies or exotic bets but established firms specializing in substrates, materials, or advanced manufacturing that are now benefiting from more demanding specifications. For example, Nan Ya PCB has been listed in Nan Ya Plastics’ annual reports as a subsidiary focused for years on high-end IC substrates, serving applications in CPU, GPU, Netcom, AI, and HPC.
The bottleneck is starting to shift toward materials
Among Asian analysts, a growing consensus is that supply constraints are beginning upstream, preempting broader shortages across the entire chain. In Taiwanese financial press and market reports throughout Asia, there’s a recurring idea: the next phase of the AI cycle may be more dependent on the availability of high-performance CCL, advanced fiberglass, low-roughness copper foil, and precision drill bits than on traditional manufacturing capacity. While many of these predictions remain market estimates rather than official company guidance, the overall trend is supported by other sources. Han’s CNC, for example, explained in a recent presentation that AI server-related high-multilayer HDI boards are increasing PCB complexity due to more layers, the need for more sophisticated vias, and adoption of high-speed materials like M8 and M9, heightening manufacturing challenges and tooling requirements.
This helps explain why some niche companies are gaining visibility. Topoint, which specializes in drill bits and perforation solutions for PCBs, publicly describes itself as a global provider of drills, drilling services, and cutting tools for PCB and CNC manufacturers. Even more significantly, Zhen Ding announced a strategic alliance with Topoint at the end of 2025, focusing on advanced PCB drilling technologies, AI server applications, and next-generation substrate technologies. In other words, the market isn’t just expecting more volume: it anticipates a technical leap that favors those better positioned with finer processes and more complex materials.
High-end PCB, ABF substrates, and CCL: the top three areas closely watched by the market
Segment-wise, the first is ABF substrates, essential for advanced encapsulation and high-performance platforms. The second is high-end PCBs, especially in AI servers and high-speed networking systems. The third is high-speed CCL, where transitions from M8 to M9 and beyond have become part of industrial conversations. In this context, companies like Zhen Ding and Gold Circuit Electronics describe a clearly favorable environment. Zhen Ding explained in March 2025 that AI-related products—including AI servers and IC substrates—were maintaining strong momentum, with the company expecting AI applications to account for over 70% of its consolidated revenue in 2025. Gold Circuit’s 2024 annual report also noted that AI-driven growth continued to boost the PCB market and that increased shipments of AI servers were fueling demand for higher-value boards.
On the high-frequency/high-speed CCL side, the story is also shifting. Although not all companies disclose detailed info, the market is paying close attention to rising specifications for high-frequency / high-speed materials used in servers, switches, and ASICs. Suppliers like Han’s CNC are publicly noting that transitioning to M8 and M9 materials increases processing complexity, benefiting those with more advanced technical capabilities. While not evidence of an outright severe shortage, it indicates a market where the supply of premium materials may be more constrained relative to current demand.
Taiwan remains at the center of the cycle again
If this thesis proves correct through the second half of 2026 and into 2027, Taiwan will once again be at the core of a crucial phase of the AI boom. Not just through TSMC and advanced packaging, but via a broader industrial network encompassing substrates, PCBs, laminates, fiber, copper, and manufacturing tools. The Taiwanese market has experienced similar shifts during past technological cycles, but the current catalyst is different: it’s not a single product category but the simultaneous build-out of data centers, high-speed networks, GPU systems, and large-scale ASIC platforms. Nan Ya Plastics has already reflected this reality by speaking about the combined push of AI servers, data centers, and networking equipment across multiple electronic materials at once.
This doesn’t mean all the most aggressive forecasts will come true exactly as some brokers predict. Many are still estimates, and the ultimate impact will depend on yields, capacity expansions, customer validations, and the actual pace of physical AI deployment. Nonetheless, the overall trend is clear: the ripple effect from AI investment has moved from chips to boards and, from there, to materials. And in this next stage of the cycle, Taiwan’s supply chain appears well-positioned to once again be a major beneficiary.
FAQs
What are ABF substrates and why are they important for AI?
They are advanced substrates used in high-performance encapsulations, especially in complex chips like CPUs, GPUs, and accelerators. Their importance grows with larger chip sizes, higher interconnection density, and increased signal and power demands. Nan Ya Plastics identifies its subsidiary Nan Ya PCB as an active player in the high-end IC substrate market for CPU, GPU, Netcom, AI, and HPC applications.
What does CCL mean and why are M8 and M9 gaining so much attention?
CCL stands for copper clad laminate, a fundamental base material for manufacturing PCBs. In high-speed/high-frequency applications like AI and networking, newer generations such as M8 and M9 support faster speeds and signal integrity but also make manufacturing more complex.
Why does Taiwan appear to be a major beneficiary at this stage?
Because it hosts a significant portion of the value chain in substrates, PCBs, CCL, fiber, and precision tools. Several local companies are already reporting rising exposure to AI servers, data centers, networking, and IC substrates.
Is a severe structural shortage already confirmed?
What exists today is a combination of clear industry signals and increasingly aggressive market forecasts. Companies acknowledge rising demand for AI servers and networking; however, estimated shortages in specific materials remain the speculations of analysts for now, rather than official guidance from all involved companies.
via: ctee