ABF Substrate Price Increase and the Material Bottleneck Behind AI
On September 8, 2026, market reporting drew attention back to ABF material upstream of IC substrates. Ajinomoto produces ABF, the key interlayer insulating material used in high performance semiconductor package substrates, and company information indicates its global market share exceeds 95 percent. In May, Ajinomoto proposed a price increase of approximately 30 percent to substrate manufacturers, with the new pricing expected to take effect in the third quarter of 2026.
The more useful way to read this is not as a raw material price event. It is a signal that increasing advanced packaging complexity has changed the demand profile for a specific upstream material.
Why Packaging Complexity Drives Material Demand
GPU, CPU and AI accelerator die areas, I/O counts and package complexity have all continued to increase. These devices connect to the board through FC-BGA style packaging, and ABF is the key insulating material in the buildup structure of such substrates. It has to support finer lines while simultaneously meeting insulation, thermal stability and high frequency transmission requirements.
Ajinomoto has disclosed that ABF layer thickness is approximately 10 micrometers, and that it has long been used in FC-BGA substrates for high performance computing and data center servers.
When advanced package dimensions grow and buildup layer counts increase, demand growth is not simply a matter of using more film. It is a rise in consumption of specific ABF grades. A larger package with more buildup layers requires more of the material, and it requires material that meets tighter specifications for thickness uniformity and dielectric behavior.
Material qualification cycles are considerably longer than for ordinary industrial materials, which means new supply cannot quickly enter the core customer base of a leading packaging house. Qualification involves reliability testing over extended periods, and a substrate manufacturer cannot switch materials on a production line serving a qualified AI accelerator package without revalidating the entire assembly.
A Concentrated Supply Structure Amplifies Timing Gaps
Ajinomoto holding more than 95 percent of the market does not by itself mean the market will be short. But such a concentrated supply structure amplifies capacity elasticity problems when demand rises suddenly.
AI server demand for high performance packaging is growing faster than the traditional PC cycle that previously set the pace. Even when substrate manufacturers are willing to expand, they must wait for material, equipment and customer qualification to proceed in step. Each of those has its own timeline, and the slowest one determines the outcome.
This is why the price increase deserves attention from the PCB industry. The bottleneck in electronics manufacturing does not always sit at final PCB capacity. It can propagate further upstream to copper foil, resin, glass cloth, copper clad laminate, and then to materials specific to packaging substrates. A material that accounts for a modest fraction of total product volume can nevertheless determine how much high end capacity can actually be released.
Once material supply becomes concentrated, supply chain competition stops being purely about purchase price. It includes whether a manufacturer can lock in specifications early, whether it can qualify substitute materials, and whether it is positioned to develop the next generation of products jointly with the upstream supplier. Those are engineering and relationship capabilities rather than procurement transactions.
IC Substrates Are Not Simply Advanced PCBs
It is important to distinguish ABF substrates from ordinary PCBs, even though both belong to the electronic interconnect manufacturing family.
An IC substrate connects a bare die directly to the mainboard. Its line precision, microvia structure, material system and manufacturing environment are closer to the semiconductor packaging stage than to board fabrication. Ordinary HDI cannot substitute for an ABF substrate.
Because of this, an ABF price increase does not mean ordinary PCBs will rise by a comparable proportion, and it does not imply that conventional HDI can move into that role. What is worth observing is the shared upstream pressure. AI servers are simultaneously raising demand for advanced packaging, high speed high layer count PCBs and low loss materials. When several critical materials tighten at the same time, cost and lead time propagate downward through different segments of the chain, each at its own rate.
For PCB manufacturers, this raises the importance of material management. Substituting a high speed material grade, adjusting a stackup, accommodating an impedance change and revalidating process parameters are all linked activities. A straightforward material swap tends to introduce new yield risk rather than remove it, particularly on high layer count boards where the stackup has been tuned around the original material’s properties. For programs sourcing through component procurement channels, understanding that interaction is part of managing the schedule.
Why Engineering Capability Matters More as Materials Tighten
As material specifications become more complex, the value a manufacturer provides is no longer limited to whether it can obtain the material.
The relevant engineering questions are which material grade is appropriate for a given product’s performance requirements, how stackup, routing and process design can control material consumption, and whether a substitute material can be validated against signal integrity and reliability requirements without restarting the entire qualification. Each of these affects whether a project moves into production on schedule.
Material consumption control is a practical example. Choosing a higher grade laminate than the design requires increases cost and may extend lead time for no functional benefit. Choosing one grade lower produces a design that fails insertion loss verification. Establishing the correct grade requires measuring the actual channel loss requirement rather than defaulting to the highest available specification.
Substitute validation is the other half. When a primary material is tight, a design with a validated alternative has options. A design without one waits. The validation work is best done during prototype, when the cost of building additional test vehicles is low and the schedule has room for it.
The Right Approach for Development Stage Programs
For AI hardware projects still in research and small batch validation, engineering coordination is more realistic than pursuing large scale capacity.
Capability covering 1 to 40 layer fabrication, 1 to 5 stage HDI, high speed differential impedance control and connectivity through assembly allows stackup, material, routing and process to be evaluated together during design verification. The value lies in reducing the time lost to repeated design revisions, not in substituting for ABF substrate manufacturing.
Where a program includes AI hardware PCBA assembly, keeping fabrication and assembly in one process chain lets a material or stackup change be evaluated through to finished assembly rather than stopping at the board. That matters because some effects of a material change only appear after reflow, when the thermal profile has interacted with the laminate and the joints.
Reviewing the interaction between material, stackup and impedance during PCB design and layout is where most of the avoidable cost sits. A design that specifies a material without confirming the process window for the intended stackup will discover the constraint during fabrication, when changing the design is expensive. A design that confirms it during review changes nothing at all.
Second Sourcing Without Restarting Qualification
The practical answer to concentrated material supply is not to avoid the material. It is to structure qualification so that a second source can be added without restarting the program.
That requires defining performance in terms of measured outcomes rather than a specific part number. If a specification states insertion loss, impedance tolerance and reliability performance targets, several materials may satisfy it. If the specification names one laminate, only that laminate will qualify, and any supply disruption becomes a schedule event.
Building the evidence base early is what makes the difference. Fabricating test coupons from more than one candidate material during prototype, and measuring them against the same criteria, establishes whether the alternatives are genuinely equivalent for the intended stackup. The measurement should include dielectric thickness variation across the panel, not only nominal values, because thickness control is where materials most often diverge in practice.
It is also worth confirming that the alternative can be processed on the same equipment with adjusted parameters rather than requiring new capabilities. A material that requires different drilling or desmear equipment is not a substitute in any useful sense, because the capacity constraint simply moves rather than resolves.
Done during development, this work costs a modest amount of engineering time. Done during a supply disruption, the same work competes with production schedules for the same fabrication capacity, which is when it is least affordable.
Reading the Supply Chain Correctly
The ABF price increase points to something broader than one material. AI computing demand is tightening capacity at several levels of the electronics supply chain simultaneously, and the constraint frequently sits at the least flexible link rather than the largest one.
A material with a 95 percent market share and a multi-year qualification cycle is structurally less flexible than a PCB fabrication line, even though it represents a small fraction of the finished product. Recognizing where that inflexibility sits is what allows a program to plan around it.
For manufacturers, the response is to build PCB fabrication capability and material knowledge that is broad enough to accommodate changes without losing process control, and to document that capability through quality management records that customers can review. For buyers, the response is to treat material qualification as part of the development schedule rather than as a purchasing task, and to insist on measured performance data rather than nominal specifications.
As AI hardware advances, the competitive question is increasingly not who can build the most layers, but who can hold a design stable while the materials and components around it move. That is a manufacturing discipline, and it is the one that determines whether a supply chain delivers.
