Well, you read the headline right, and we are talking about a critical element in AI chips that is not only in tremendous shortage but also originally comes from an MSG maker.
Ajinomoto, the MSG Maker, Controls the Production of ABF That Is Absolutely Crucial For Advanced Packaging
The aggressiveness of the AI demand cycle has left almost every supply chain entity involved facing shortages, whether in semiconductors, advanced packaging, OSAT services, and more. Traditionally, the computing industry has an idea of demand cycles, given how they have historically evolved, but with the AI datacenter buildout, the scale of customer requirements has reached a level where suppliers have little clue about how to address it, apart from hiking prices in the short term. One such element that isn't often discussed but is a crucial part of modern-day chips is the ABF substrate, produced by an entity traditionally known for food seasoning.
Allow me to go a bit technical in the upcoming paragraph or two, since ABF substrates deserve this. ABF, or Ajinomoto Build-up Film, is a crucial part of advanced packaging. For a general overview, it is a thin insulating film that acts as a 'bridge' between the silicon die and PCB connections. You'll know more about the ABF substrate from the graphics below, but this film is important because it enables high-end chips to achieve. high I/O density and signal integrity at multi-gigahertz frequencies, which is crucial for chips like NVIDIA's Blackwell or Rubin, which operate under intense environments.
The ABF substrate supply chain is a complex structure to look at, since it basically depends on multiple entities, such as Ajinomoto Fine-Techno for the film we are talking about above, along with Ibiden, which acts as the substrate manufacturer, and then Taiwan's Unimicron and many other suppliers that are responsible for the final stages. However, in a more abstract view, Ajinomoto, the MSG maker we are talking about, holds the dominance among all supply chain entities we discussed above, since without the film itself, final packaged AI accelerators cannot ship at all.
Now that I have established the context for ABF, let's discuss the issue. With AI accelerators, the use of the film relative to the substrate used in other components, such as GPUs, roughly increases by 15-18 times, since a traditional accelerator package requires 8 to 16+ layers of ABF, depending on the actual package size. So, the larger chips like Rubin and Rubin Ultra become, the more ABF becomes a bottleneck, which is why you can clearly see the actual constraint here. As a general reader, you might think about this: what if Ajinomoto scales up production, and then voila, solves the problem? Well, the devil is in the details.
One of the biggest issues is that the ABF supply chain depends on Ajinomoto Fine-Techno for the film material itself, and given that there's a single supplier, you cannot address this demand on your own. The Japanese firm has made efforts to increase production, but expansion is also accompanied by the fear of overcommitment, which means that substrate producers like Ibiden will always face a supply cap with ABF. At the same time, as AI packages grow in size, ABF layer requirements increase, and techniques like semi-additive patterning (SAP) introduce the risk of compromising yield rates and ruining the entire multi-layer process.
Well, the AI hype cannot wait for Ajinomoto to give us ABF, so what's the option now? Hyperscalers have already realized this constraint, which is why they are responding with prepayments, helping the Japanese company bring newer production lines and securing long-term contracts. Yet at the same time, the problem with every demand cycle is that there isn't enough to fulfill all customers' needs, which is why only a few entities will have their requirements serviced around ABF and substrate packages.
ABF demand is forecast to grow by double digits annually, and according to DigiTimes, we expect a three-year demand cycle, which means supply will remain constrained for a long time. ABF is one of the more 'quieter' elements in short supply in the AI supply chain, but we highlighted it specifically here because it is proving to be one of the bigger bottlenecks in scaling the advanced packaging supply.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
