Ajinomoto Confronts the Major Invisible Bottleneck of AI Chips

Ajinomoto, known for decades for its connection to the food industry, has become a silent player in the global semiconductor supply chain. Its Ajinomoto Build-up Film, better known as ABF, is one of those materials that rarely feature in presentations by major chip manufacturers but are essential for the operation of advanced processors within their packages.

The pressure is now mounting on this material due to the growth of artificial intelligence. Chips for servers, networks, and data centers require increasingly larger, more complex substrates with more layers. This means more ABF consumption per package and a more demanding supply chain. The debate no longer focuses solely on GPUs, HBM, lithography, or TSMC’s capacity. It also involves specialized materials that connect silicon to the rest of the system.

Ajinomoto assures, according to forecasts cited by CEO Shigeo Nakamura, that it can meet customer demand through 2030. Beyond that date, visibility diminishes. This cautious approach well summarizes the industry’s current state: demand for AI seems solid, but the capacity to produce all the necessary components is not growing at the same pace.

What is ABF and Why Does It Matter So Much

ABF is an intermediate insulating material used in high-performance semiconductor package substrates. Its function is to enable the construction of thin, precise layers that connect the chip to the substrate and subsequently to the system board. In simple terms, it helps reliably route thousands of electrical signals from an advanced processor.

Since its commercial launch in 1999, ABF has become the de facto standard for many CPU, GPU, and other high-performance chip packages. It was initially important in personal computers and gaming consoles. Later, it gained significance in servers, networks, accelerators, and data centers. With the rise of Artificial Intelligence, its role has become even more prominent.

Interest in ABF is increasing because AI chips are not only more powerful—they are also larger, consume more energy, communicate with more memory, and require more interconnections. A modern accelerator may need a substrate of significantly larger surface area and more layers than an earlier-generation processor. Each of these layers involves more material, greater precision, and higher risk of bottlenecks.

Ajinomoto reflects this in its own forecasts: high-end server and network applications are gaining weight within ABF consumption and will continue to push aside traditional uses. The transition doesn’t mean that PCs will disappear but that growth will concentrate on higher-value and more complex systems.

AI Multiplies the Value of Discrete Materials

AI has transformed how the supply chain is viewed. For years, the focus was on wafer fabrication plants, manufacturing nodes, and GPU availability. Now, the market is also paying attention to packaging, substrates, chemicals, resins, interposers, glass, copper, HBM, and assembly capacity.

ABF now plays a strategic secondary role in this ecosystem. If insulating material for advanced substrates is lacking, having a designed chip or a produced wafer is not enough. The package doesn’t reach the server. And without a package, there’s no accelerator available to train or run models.

Though ABF doesn’t enjoy the commercial visibility of a GPU or the unit value of an advanced wafer, shortages can create a ripple effect. That’s why it’s once again attracting attention from investors, substrate manufacturers, and end customers eager to ensure supply for several years.

Ajinomoto Avoids a Steep Price Increase

Ajinomoto’s stance is unique. The company has a dominant position in this insulating film, and amid rising demand, it could be tempted to raise prices aggressively. However, it is acting cautiously.

Nakamura has argued that they do not want to raise prices just because the market allows it. The justification is business-oriented, not altruistic: opportunistic price hikes could damage long-term relationships with customers who’ve relied on Ajinomoto for years and depend on a stable supply chain.

This decision makes sense when considering semiconductors as a trust-based business. Major customers don’t buy critical materials solely based on price; they value continuity, quality, technical validation, support, and the ability to evolve. If a supplier exploits shortages for short-term margins, it risks pushing customers to seek alternatives or re-design processes.

This contrasts with other materials in the supply chain. Copper foil and fiberglass manufacturers have raised prices due to increased costs. Ajinomoto argues that ABF does not directly use those inputs, reducing some exposure. Still, uncertainties remain in resins, chemical loads, and organic solvents, as well as geopolitical risks linked to the Middle East. The company maintains that it is managing these risks through a diversified sourcing network.

The New Plant Will Arrive Late for Immediate Demand

Ajinomoto Fine-Techno has secured land in Kani, Gifu Prefecture, to build a new ABF plant. The investment in land is approximately 1.2 billion yen. Construction is scheduled to start in 2028, with operations expected in 2032. It will be the company’s third manufacturing site, after Kawasaki and Gunma.

The timeline illustrates the challenge: AI demand is growing now, but the new plant is planned to boost capacity beyond 2030. This means that in the coming years, the industry will rely on capacity expansions at existing facilities, process improvements, close planning with customers, and substrate manufacturers’ ability to handle more material.

The new facility also aims to ensure business continuity. Diversifying locations reduces reliance on one or two sites. For critical materials, operational incidents, natural disasters, or logistical disruptions could affect the entire semiconductor supply chain.

For AI customers, the concern isn’t just the presence of ABF but having enough of it with the necessary specifications for larger, multi-layer packages. The industry needs not just more volume but materials capable of supporting new packaging architectures.

Substrates Grow with Every Generation of Accelerators

The evolution of AI accelerators pushes substrate sizes and layers higher. Chips don’t grow by themselves—they are integrated with HBM memory, high-density interconnections, controllers, power elements, and increasingly complex packages. The substrate must support this architecture without degrading signals, power, or reliability.

Ajinomoto points out that packages for HPC and AI applications have increased in layer count and surface area, and this trend will continue. In other words, even if the number of chips sold grows moderately, the ABF consumption per chip could rise sharply due to packaging complexity.

This explains why ABF is now at the center of the debate. Demand doesn’t depend only on AI chip sales but also on how their designs evolve. If future accelerators require more surface area and layers, the material consumption could grow faster than chip sales alone.

Can Glass Replace ABF?

The industry is also exploring glass substrates as an alternative or complement for future advanced packages. Glass offers potential advantages in dimensional stability, interconnection, and scalability. Companies like Intel, Samsung, and others have shown interest in this pathway for upcoming generations.

However, Ajinomoto does not see glass as an immediate replacement for ABF. The company’s position is that both technologies can coexist—this is a cautious view. Semiconductor materials rarely change abruptly after years of validation, with factories, suppliers, and designs built around a particular solution.

Long-term threats exist, but they don’t invalidate current demand. Major customers need solutions that are available, tested, and qualified. ABF currently has that advantage. Glass may gain ground but will require years of industrialization, cost competitiveness, and acceptance within the supply chain.

The Question Isn’t Whether There Will Be Demand but Whether the Chain Will Respond

Ajinomoto’s CEO characterized AI demand as something beyond a passing trend. The company foresees expansion into robotics, industrial automation, and what’s often called “physical AI,” where models connect with machines, sensors, vehicles, and physical environments.

This perspective shifts the discussion. If AI extends from data centers to robots, factories, industrial devices, and autonomous systems, the demand for advanced semiconductors could continue to grow for years. The limiting factor will be the capacity of every link in the supply chain—from wafers and HBM to substrates, ABF, testing, assembly, and power.

Ajinomoto stands in a comfortable yet demanding position. It has a dominant product, global customers, and growing demand. But it also bears the responsibility of preventing what seems like a modest material from becoming a bottleneck for the entire industry.

The ABF case is a reminder of a fundamental hardware lesson: AI doesn’t scale only through better models. It scales through materials, factories, chemicals, packaging, logistics, and years of silent investment. The next major bottleneck might not be in the GPU but in a thin insulating film that was born in a company famous for umami.

Frequently Asked Questions

What is Ajinomoto Build-up Film?

Ajinomoto Build-up Film, or ABF, is an insulating material used in high-performance semiconductor package substrates, such as CPUs, GPUs, and accelerators for data centers.

Why is ABF important for Artificial Intelligence?

Because AI chips use larger packages with more layers and interconnections. This increases ABF consumption and can strain the supply chain.

Can Ajinomoto meet the demand?

The company states that, based on its internal forecasts, it can supply customer needs until 2030. Beyond that, its visibility becomes more limited.

When will the new ABF plant be ready?

Ajinomoto Fine-Techno plans to start building its new plant in Kani, Gifu, in 2028, with operations expected to commence in 2032.