Semiconductor Geopolitical Risk Sits Past the Fab, Not Inside It

TL;DR: Semiconductor geopolitical risk mapping usually stops at the wafer fab, tracking TSMC's capacity and the Netherlands' export licences for lithography tools. Three chokepoints sit past that point and carry comparable concentration: the design software every chip is built with, the substrate film that packages it, and the assembly and test capacity that finishes it. A single 2026 supply cut in one of those three, announced by a company most procurement teams have never screened, shows how fast the risk moves once it is outside the fab conversation.

Key takeaways:

• Synopsys, Cadence and Siemens EDA hold more than 85 percent of the global electronic design automation market combined, and Synopsys tools alone cover 84 to 90-plus percent of specific design steps like synthesis and signoff.
• Ajinomoto controls the large majority of the global ABF substrate film market, cutting shipments to China by 30 percent starting in the third quarter of 2026 after a roughly 30 percent price increase announced in May.
• Taiwan-based outsourced assembly and test firms held about 54 percent of global OSAT revenue in 2024, with ASE alone accounting for 44.6 percent of the ten-company total.
• Goldman Sachs projects the ABF substrate supply-demand gap widening from roughly 10 percent in the second half of 2026 to 21 percent in 2027 and 42 percent in 2028, driven by AI chip demand outrunning new capacity.

Ajinomoto is not a name most semiconductor procurement teams screen. It is a Japanese food and chemical company, and it also supplies an estimated 90 to 95-plus percent of the world's ABF film, the insulating material that lets a chip package route signals in multiple layers instead of one.[^1] In August 2026 it cut ABF shipments to China by 30 percent, on top of a roughly 30 percent price increase announced three months earlier.[^2] That is a single supplier decision reaching a share of global chip output that no fab-capacity map would have flagged as at risk.

The Four Single Points of Failure Nobody Fab-Maps

Most corporate semiconductor risk assessments concentrate on the wafer fab, tracking TSMC's Taiwan capacity, Samsung's foundry share and export licensing for extreme ultraviolet lithography tools. That coverage is real but partial. At least three other stages of the chain carry concentration levels as high as anything in the fab, and a fourth has already shown what disruption there looks like:

• Design software. Synopsys, Cadence and Siemens EDA hold more than 85 percent of the global electronic design automation market between them, and at the level of individual tools the concentration is sharper still: Synopsys's PrimeTime covers more than 90 percent of signoff timing analysis and Siemens's Calibre covers more than 85 percent of physical verification.[^3] No advanced chip is designed without software from one of these three companies.
• Substrate film. Ajinomoto's ABF film share sits above 90 percent by most industry estimates, and ABF accounts for roughly 30 percent of a finished substrate's bill of materials, so a price or supply change there moves the cost of every advanced package built on it.[^1][^2]
• Assembly and test. Taiwan-based outsourced assembly and test companies held about 54 percent of global OSAT revenue in 2024, with ASE alone at 44.6 percent of the top ten firms' combined sales, concentrating the final packaging and test step in one geography.[^4]
• Specialty gases. Ukraine supplied a large share of the world's semiconductor-grade neon before 2022, and Russia's invasion that year briefly disrupted that supply, a dated but instructive example of how a input most chipmakers never priced as strategic became one within weeks.[^5]

Each of these sits downstream or alongside the fab, not inside it, which is exactly why a risk map built around fab capacity and export licences misses all four.

The Substrate Squeeze Shows How Fast a Chokepoint Moves

The ABF case is worth tracing in full because it moved from a niche materials story to a global cost pressure inside a single year. What the record shows: Ajinomoto raised ABF prices by roughly 30 percent in May 2026, citing demand growth from AI GPU packaging that requires one to two large substrates per chip, combined with new substrate fab lines that take 18 to 24 months to bring online.[^2] In August, the company cut shipments to mainland China by 30 percent, describing the move as prioritizing Japanese and other overseas accounts.[^1] The timing sits close to China's own rare earth export curbs, which cut exports to Japan by roughly 51 percent year over year in the first half of 2026.[^1]

What we infer: reporting on the August cut treats the proximity of those two events as more than coincidence, though Ajinomoto's own stated rationale does not claim a retaliatory link, and this piece does not assert one either. What is verifiable independent of motive is the structural outcome: China's domestic ABF self-sufficiency sits below 5 percent, so a 30 percent cut from the dominant supplier has nowhere close to a domestic substitute to absorb it.[^1] Goldman Sachs, cited in the same reporting, projects the resulting global supply-demand gap widening from about 10 percent in the second half of 2026 to 21 percent in 2027 and 42 percent in 2028 as AI-driven demand keeps outpacing new capacity.[^1]

What we recommend: a supplier concentration review that stops at wafer fabs and assembly plants will not catch this kind of exposure, because Ajinomoto does not appear on most semiconductor supplier lists at all. It appears on chemical and specialty materials lists, filed under a category few chip-focused procurement teams monitor for geopolitical triggers.

Mapping the Stack Nobody Put on One Page

The practical gap is not analytical difficulty. Concentration data for EDA, substrates and assembly and test is publicly available, the way fab capacity data is. The gap is that these three sit in different functional silos inside most companies: design tooling reports to engineering, substrate sourcing reports to procurement, and assembly and test capacity reports to manufacturing operations, and none of the three functions is typically asked to think in geopolitical terms about its own supplier base. A single stack map, listing every stage from design software through final test with its top-two-supplier concentration and country exposure noted at each stage, would surface the Ajinomoto-scale risk before a price increase does it instead.

This sits alongside two related exposures already mapped elsewhere. A separate Fortius Intel field note on the bifurcation checklist for product teams covers the architecture decisions that follow once a chokepoint like this turns into an actual split market, and another on China's rare earth export controls traces the upstream materials dispute that the ABF cut sits downstream of. Both assume the stack map in this piece already exists somewhere in the reader's organization. For most, it does not yet, and the running Fortius Intel brief is as reasonable a place as any to start building it.

Fortius Intel note: What happens to a company's AI hardware roadmap when the supplier it never screened controls a step it never mapped? Most procurement teams cannot answer that today, because the stack map that would let them answer it does not exist yet.

Methodology: this analysis draws on SemiAnalysis's May 2026 EDA market primer, Tom's Hardware's August 2026 reporting on Ajinomoto's ABF supply cut to China, AtlasPCB's May 2026 reporting on Ajinomoto's price increase, Mark LaPedus's OSAT market share analysis of 2024 revenue data, and established reporting on the 2022 Ukraine neon supply disruption, cross-checking the Ajinomoto market-share and price figures against two independent trade-press sources.

---

Footnotes