AI Chip Demand Is Reshaping the Advanced Substrate Industry and Its Pricing Power

Table of Contents

• You might want to know


• Main Topic


• Key Insights Table


• Afterwards...

You might want to know

• Why are advanced package substrates becoming a bottleneck as demand for AI accelerators expands?


• How could larger substrate designs and emerging materials change manufacturing economics and supplier influence?

Main Topic

The rapid expansion of demand for artificial intelligence accelerators is changing the requirements placed on semiconductor packaging. Next-generation chips combine more computing resources with high-bandwidth memory (HBM), which places new demands on package size, interconnection density, and substrate layer count. As a result, demand for high-end ABF substrates is increasing. At the same time, longer equipment lead times and more difficult manufacturing processes constrain the speed at which suppliers can add usable capacity.

The source article anticipates that the supply gap for high-end products could widen through 2028. It also describes a major shift in purchasing behavior: global technology companies are pursuing long-term agreements and advance payments to reserve production capacity. Such arrangements can help buyers secure access to materials that are difficult to obtain, while giving suppliers greater visibility into future demand. Together, capacity constraints and advance purchasing are shifting bargaining power away from buyers and toward qualified suppliers.

Substrate dimensions and layer counts are increasing

Advanced packaging allows a chip package to integrate more processing components and memory than earlier designs. The source describes package-mask dimensions increasing from 1 times their earlier size to 3 times, or even more than 9 times. These changes are accompanied by larger and more complex substrates. Specifications that were previously around 55×55mm are expanding to 110×110mm and more than 26 layers.

These changes do not translate into a simple, proportional increase in manufacturing effort. A larger substrate occupies more production area, while additional layers require more complex processing and tighter control. Consequently, the equivalent processing area can rise nonlinearly. Even if nominal unit counts appear manageable, the combination of larger formats and complex stacks can consume a substantially greater share of available production capacity.

Manufacturing difficulty also increases with size and complexity. Differences in the coefficient of thermal expansion (CTE) among materials can contribute to warpage as a package is built and processed. Larger structures can make it harder to maintain flatness and align circuit patterns across layers. These challenges affect yield as well as throughput: a manufacturer may have installed equipment but still face limits on the volume of products that can be made to demanding specifications with consistent quality.

For this reason, having production equipment alone is not sufficient to establish a strong position in the high-end market. Suppliers must also demonstrate the ability to manufacture advanced products at volume and control yield. Companies with established process expertise and reliable quality management may therefore command a technical premium, particularly when customers need predictable supply for complex AI packages.

New materials are prompting technical change

Traditional organic-core substrates face physical limitations as package dimensions increase, including concerns about warpage and heat dissipation. Glass-core substrates and ceramic interposers are being explored as potential ways to address some of these constraints in next-generation high-end packaging. These approaches are not yet free of manufacturing challenges: processing precision and the maturity of production yields remain important considerations during early adoption.

The source reports that major companies in Taiwan and Japan have begun positioning themselves around these technologies. Their work includes the introduction of ultra-low-CTE materials, automated routing, and design-for-manufacturability (DFM) rules. These efforts matter because new materials require changes not only to the substrate itself but also to process design, equipment settings, inspection, and coordination across the supply chain.

In the near term, the transition from organic cores to inorganic materials is likely to involve testing and gradual development rather than an immediate, uniform replacement. The commercial value of glass and ceramic approaches will depend on whether manufacturers can achieve the required precision, yield, reliability, and scale. Over the medium to long term, successful adoption could create another phase of materials upgrading and value growth in the substrate industry.

Taiwanese suppliers and their reported positions

The article identifies Taiwanese substrate makers as important participants in the advanced-packaging upgrade, citing their market presence and technical experience. It names four companies and outlines areas of activity. These descriptions are industry claims from the source material; they should not be treated as a guarantee of future results or as investment recommendations.

• Unimicron (欣興, 3037-TW): The source describes the company as having high-end ABF mass-production and CoWoS substrate capabilities, with strong visibility from customer long-term agreements. It also cites work on EMIB-T and innovative panel technologies involving glass and ceramic cores.


• Nan Ya PCB (南電, 8046-TW): The source says the company is expanding the penetration of high-layer-count, large-format products and working with ultra-low-CTE core materials and glass-substrate testing.


• Kinsus Interconnect Technology (景碩, 3189-TW): The article links the company to increasing AI accelerator volumes and says it is participating with partners including Taiwan Semiconductor Manufacturing Company (台積電, 2330-TW) in research and development for ceramic and glass substrate processes.


• Zhen Ding Technology Holding (臻鼎-KY, 4958-TW): The source describes a push into high-end substrates for AI servers and advanced glass-based HDI applications.

The broader industry implication is that demand growth alone does not determine which suppliers benefit. The ability to make large, multilayer products with stable yields, and to qualify new materials, will shape how much capacity can be delivered to customers. Long-term contracts may offer demand visibility, but they do not remove manufacturing risk, investment requirements, or the possibility that technology adoption develops more slowly than expected.

Capacity reservations and pricing power

Long-term agreements and advance payments are common ways for buyers and suppliers to manage uncertainty in capital-intensive supply chains. For technology companies, reserving capacity can reduce the risk that critical substrate supply will limit chip and system production. For suppliers, committed demand can improve planning and may support investment in equipment, process development, and materials qualification.

However, a reservation does not automatically mean that every planned unit can be produced. Equipment delivery schedules, process complexity, and yield development all influence effective output. In a market where demand is strong and only a limited number of manufacturers can meet advanced specifications, suppliers may gain negotiating leverage. The article’s central argument is that this supply-demand imbalance is reversing the industry’s traditional pricing relationship, but the extent and duration of any shift will depend on actual capacity additions and customer demand.

Investors and industry observers should distinguish between reported technological progress, production capability, and commercial profitability. A company can participate in research without having high-volume production, and a new product can require substantial time to reach reliable yields. Likewise, forecasts of a supply gap through 2028 describe an expectation, not a certainty. Changes in chip architecture, customer plans, or manufacturing expansion could affect the outcome.

Key Insights Table

Afterwards...

The next stage of development will depend on whether substrate manufacturers can turn promising materials and process concepts into consistent, high-volume production. Further exploration is needed in glass processing, ceramic integration, ultra-low-CTE materials, automated routing, and DFM methods. These areas may help address warpage, alignment, and thermal-management concerns, but their value will ultimately be measured by reliability, manufacturability, and cost at scale.

Industry participants will also need better ways to evaluate capacity beyond headline equipment figures. Useful measures include production yield, qualification progress, process stability, and the amount of output that meets customer requirements. Transparent progress reporting can help customers and observers distinguish research activity from commercially available supply.

Finally, continued advances in AI accelerators and packaging architectures will influence substrate specifications. Collaboration among chip designers, packaging providers, material suppliers, and substrate manufacturers may be essential to align designs with realistic production capabilities. The key question for the years ahead is not only how much capacity the industry can build, but how reliably it can produce increasingly large and complex substrates at scale.

Source: Chen Chih-Lin, analyst / Kai Xu Investment Advisory.

Risk notice: The source states that it has no improper financial relationship with the individual securities it recommends or analyzes. Past performance does not guarantee future returns. Investors should make independent judgments, evaluate carefully, and bear their own investment risks.