The narrative around the AI chip shortage is entering a new chapter.
For a long time, the main bottleneck was in cutting-edge chip manufacturing, like TSMC's CoWoS advanced packaging. Now, however, the pressure is shifting further down the supply chain to the substrates that act as the foundation for these powerful chips, and the raw materials needed to make them. This is highlighted by a recent announcement from Kinsus, a major Taiwanese substrate manufacturer.
Kinsus plans to increase its ABF substrate production capacity by 25% by 2027 to meet soaring demand. However, the company stated that its most immediate challenge isn't building new factory lines, but securing a stable supply of critical materials. This reveals the true nature of the current bottleneck.
So, what's causing this situation? The causal chain can be broken down into two main parts.
First, the demand for CPUs in AI systems is growing much faster than anticipated. While GPUs handle the core AI calculations, new architectures like NVIDIA's Blackwell and Rubin platforms use tightly integrated CPUs for crucial tasks like data pre-processing and orchestration. This design shift means the demand for high-performance CPU substrates—namely ABF substrates—is skyrocketing alongside GPU demand.
Second, the supply of key materials for these substrates is extremely tight. The most critical components include low-CTE glass cloth (a specialized fabric for stability, mainly from Japan's Nittobo), insulating films, and copper-clad laminates (CCL). Reports indicate that the shortage of high-grade glass cloth could last until 2027, and rising copper prices are further squeezing supply. Even though suppliers like Ajinomoto (for films) are expanding, it won't provide short-term relief.
In essence, the AI hardware race is no longer just about who can make the most advanced chip. It's now equally about who can secure the underlying materials and packaging capacity. The focus has decisively shifted from the front-end to the back-end of the semiconductor supply chain, and solving these material shortages is now the key to unlocking future AI growth.
- ABF Substrate: Ajinomoto Build-up Film substrate. A high-performance integrated circuit substrate used for high-density chips like CPUs and GPUs, essential for modern AI hardware.
- Low-CTE Glass Cloth: A specialized fiberglass fabric with a low coefficient of thermal expansion. It is a critical material for high-layer-count substrates, preventing warping and ensuring reliability under heat.
- CCL (Copper-Clad Laminate): A core material for printed circuit boards and substrates, consisting of an insulating layer laminated with copper foil. It forms the base for creating circuit patterns.
