VIII
Substrates, resist & wafer materials
ABF substrates, advanced photoresists, and 300mm silicon are quiet few-supplier markets sitting under every chip.
Every AI accelerator is a stack of expensive things sitting on top of cheap-sounding things, and the cheap-sounding things are where the supply chain quietly knots. Underneath the GPU die and its HBM stacks is an ABF build-up film substrate; the patterns on the die were printed with a photoresist made by a handful of Japanese chemical houses; and the die itself started life as a polished 300mm silicon wafer from one of two companies. These materials are a rounding error in dollar terms next to the chip they enable, but they are genuine single- or few-supplier markets with multi-year qualification cycles, and when one of them tightens it gates accelerator output as hard as any tool or foundry. The catch for an equity investor: almost every node in this layer is owned by a giant, diversified Japanese or German conglomerate where the relevant business is a small, if highly profitable, slice — and the purest pure-plays are not US-investable at all.
Who owns it
The single most concentrated node is ABF build-up film — the organic dielectric that insulates the layers of a flip-chip package substrate. Ajinomoto, the Japanese food and amino-acid company, invented it and is estimated to hold roughly a 90% global share, with ABF commanding over 95% of high-performance chip packaging material; Nvidia, AMD and Intel packages all route through it. Ajinomoto sells the film; the substrate is then built by a separate set of converters. High-end FC-BGA substrates are a fragmented multi-supplier market rather than a monopoly: no single converter holds more than roughly 18% share, with Unimicron usually cited as the largest and Ibiden, Shinko Electric, AT&S and Nan Ya PCB rounding out the top tier. Ibiden is a leading substrate supplier into Nvidia's accelerator packages, and of the top converters only it has a usable US line (the unsponsored OTC ticker IBIDF); Shinko, Unimicron, AT&S and Nan Ya are reachable only on their home exchanges and are out of scope here.
The second node is advanced photoresist — the light-sensitive polymer that defines every transistor. It is a ~$5.5B advanced-lithography photoresist market dominated by five Japanese and Korean houses, and the EUV sub-segment is more concentrated still: JSR, Tokyo Ohka Kogyo (TOK) and Shin-Etsu together hold north of 90% of EUV resist, with JSR alone over 22% of the broader market in 2024. This is where the listing problem bites hardest: JSR, arguably the EUV-resist leader, was taken private by Japan Investment Corporation, delisted from the Tokyo Stock Exchange on June 25, 2024 and is no longer investable. The remaining listed surfaces are TOK (thin US OTC lines TOKCF/TKCYY), Shin-Etsu (SHECY), Fujifilm (FUJIY, which completed a 30% EUV-resist capacity expansion at Kumamoto in 2025 for TSMC) and Merck KGaA's EMD Electronics (MKKGY).
The third node is 300mm silicon wafers, a genuine duopoly: Shin-Etsu and SUMCO together account for more than half of global 300mm capacity, and five incumbents — Shin-Etsu, SUMCO, GlobalWafers, Siltronic and SK Siltron — control ~85% of supply. Shin-Etsu (SHECY) and SUMCO (SUOPY) are US-investable via OTC; Siltronic trades thinly as SSLLF; GlobalWafers and SK Siltron have no usable US line (GlobalWafers is Taipei-listed only). Finally, the CMP slurries, deposition materials and filtration that purify and planarize every wafer are dominated by Entegris (ENTG) — the one clean, large-cap, NASDAQ-listed name in the entire chokepoint — alongside Merck KGaA, Resonac (SHWDF, the former Showa Denko) in packaging materials, and Fujifilm.
What breaks it
Three forces make this layer a chokepoint rather than a commodity. First, qualification lock-in: a substrate, resist or wafer is qualified into a specific node and package over months to years, so customers cannot dual-source on short notice, and an incumbent's share is sticky in both directions. Second, capacity that does not flex quickly: ABF substrate and 300mm wafer lines are capital-intensive and lumpy. AI demand has already split the market — AI substrates sold out with prices up 15-40% while PC substrates sit in oversupply with flat-to-declining prices, and SUMCO management projects AI-server wafer demand reaching 800,000-1,000,000 wafers/month by 2027. Third, structural AI tailwinds in the spec itself: AI accelerators demand larger, more-layered substrates (the 8-16 layer segment is already ~48% of the substrate market by volume, with HPC/AI roughly 32% of total demand), ultra-flat wafers for 2nm/3nm, and more EUV-resist exposures per wafer as layer counts climb — all of which raise both volume and price per chip in this layer.
The binding risk on the other side is geographic concentration. ABF film, the leading substrate converters, EUV resist and 300mm wafers are overwhelmingly made in Japan (with substrate conversion also in Taiwan), which is the same single-region fragility that defines the foundry and packaging chokepoints. That concentration is now a policy variable: the US has opened a Section 232 semiconductor investigation and begun imposing tariffs, with advanced-logic tariffs effective January 15, 2026 and a phase-2 expansion of covered categories under negotiation. Materials are not yet the headline target, but any broadening that sweeps in substrates, resist or wafers would reprice the supply chain — and conversely, onshoring incentives (GlobalWafers' $5B+ Texas plant ramping toward ~1.2M wafers/month at full capacity) point to where new capacity lands.
What forces the reprice
The cleanest catalyst clock is substrate-maker capex, because it is both a demand tell and a margin lever. Ibiden's board approved ¥500 billion ($3.3B) of IC-package-substrate investment across FY2026-2028 on February 3, 2026, aiming to roughly multiply high-performance capacity 2.5x, with the new Gama Plant (phase-1, ~¥220B, mass production from FY2027) and the Ono Plant Cell 8 ramping through FY2027. Ajinomoto is investing alongside it, including a ¥1.2B land purchase for a new ABF plant targeted for 2032 as ABF margins top 50% on the AI boom. Each capacity announcement, and each quarterly read on AI-substrate price/mix, moves the substrate names.
The second clock is EUV-resist adoption and qualification wins — High-NA EUV ramps and new node qualifications (TSMC N2, Intel 14A/18A) pull volume toward the EUV-resist leaders and validate the smaller resist names; the photoresist market is forecast to compound >11% through 2034 on EUV/High-NA. The third is wafer pricing and utilization: Shin-Etsu and SUMCO's ~$1B joint 2025 investment in ultra-flat 300mm capacity for 2nm/3nm and SUMCO's exit from 200mm at Miyazaki by end-2026 mark the pivot from a sluggish, inventory-heavy wafer market toward AI-grade tightness. For the one US large-cap, Entegris, the reprice is unit-driven materials growth at advanced nodes: it posted full-year 2025 net sales of ~$3.2B and guided Q2 2026 sales of $815-845M, with CMP consumables, liquid filtration and selective-etch products winning at the 2nm node. Finally, the Section 232 policy timeline — including a Commerce update to the President on datacenter semiconductors due by July 1, 2026 — is the macro overlay that could suddenly make geography the story.
Be honest about size and earliness
This is the chokepoint where the gap between strategic importance and investable purity is widest, and it deserves blunt caveats. Ajinomoto is a food company. Its electronic-materials business — which contains the near-monopoly ABF film — generated ¥100.7B of sales in FY2025, about 6% of group revenue, on ABF revenue of roughly $570M; the film is wildly profitable (over 50% operating margin and ~20% of group profit), but you are buying a seasonings-and-amino-acids conglomerate to get it, and you are buying it through a thinly-traded sponsored ADR (AJNMY) — the unsponsored AJINY line has gone effectively non-trading. Shin-Etsu is the world's largest PVC maker before it is a wafer or resist maker; the semiconductor-silicon and resist lines are a slice of a chemicals giant, accessed via OTC line SHECY. Merck KGaA is a pharma/life-sciences company where Electronics is 17% of group sales and semiconductor materials ~65% of that — real, but diluted, and reached via MKKGY. Resonac is a broad chemicals group where semiconductor and electronic materials is ~30% of sales. Fujifilm is a healthcare/imaging conglomerate with a genuine but small resist/CMP line.
The cleaner exposures come with their own asterisks. Ibiden is the purest large-cap substrate play — Electronics is ~65% of consolidated revenue and it is the leading Nvidia substrate supplier — but its only US surface is the unsponsored, thinly traded IBIDF. SUMCO (SUOPY) is the most wafer-levered name but is mid-cycle and only ~23% of the 300mm market, again via OTC. Tokyo Ohka Kogyo offers EUV-resist leadership but trades only on Pink-tier US lines (TOKCF/TKCYY) with minimal disclosure and liquidity, so it is speculative at best. Tokyo Electron is sometimes listed here because its coater/developer track holds ~90% share and 100% of EUV coater/developer — but that is equipment that applies and develops resist, not a material, and it belongs to the lithography/deposition chokepoint, so we name it rather than tag it. The two genuinely unambiguous, large-cap pure-plays in this entire layer — JSR in EUV resist and Zeiss SMT in EUV optics — are both private and uninvestable. The only clean, sponsored, exchange-listed US name is Entegris (NASDAQ: ENTG), and even there CMP/filtration materials are one node among many in a broad semiconductor-materials franchise. Treat this chokepoint as a real, structurally tight, Japan-and-Germany-anchored materials layer whose US access is mostly thin OTC wrappers on diversified parents — important to understand, easy to over-own, and where the purest names cannot be bought at all. Educational only; not investment advice.
Who owns the choke
Ajinomoto Co., Inc. (ADR)
Ajinomoto invented ABF build-up film and holds an estimated ~90% global share; ABF commands over 95% of high-performance chip packaging material. But this is a food company: electronic materials were ~6% of FY2025 group revenue (¥100.7B sales) on ABF revenue of roughly $570M, albeit at >50% operating margin and ~20% of group profit. US access is the sponsored ADR AJNMY; the unsponsored AJINY line has gone effectively non-trading.
- Bull
- If AI substrate demand keeps climbing, Ajinomoto's ~90% ABF-film monopoly at >50% operating margin compounds as the new 2032 plant adds capacity.
- Bear
- You buy a food conglomerate where electronics are ~6% of revenue, via a thin sponsored ADR (AJNMY) rather than the company's deeply-liquid Tokyo line.
Entegris, Inc.
Entegris is the one unambiguous US large-cap in this layer: a leading supplier of CMP slurries/pads, liquid filtration and purification, deposition materials and specialty gases. Full-year 2025 net sales were ~$3.2B; it guided Q2 2026 sales of $815-845M, with unit-driven growth in CMP consumables, liquid filtration and selective etch at the 2nm node. Materials are still one node among many in a broad semiconductor-materials franchise, but it is genuinely diversified across the supply chain and cleanly investable.
- Bull
- If 2nm node materials intensity rises, Entegris' diversified CMP, filtration and deposition franchise — the one clean US large-cap here — sees unit-driven growth.
- Bear
- At ~$3.2B sales, materials are still one node among many in a broad franchise, so AI attribution is partial and growth tracks the broader cycle.
FUJIFILM Holdings Corporation (ADR)
Fujifilm's Electronic Materials unit supplies EUV/ArF/KrF photoresists, CMP slurries and post-CMP cleaners, developers and high-purity solvents; it completed a 30% EUV-resist capacity expansion at Kumamoto in 2025 driven by TSMC. It is a top-five photoresist player, but inside a diversified healthcare/imaging/materials conglomerate where electronic materials are a small fraction of group revenue. US access is the OTC ADR line FUJIY.
- Bull
- If EUV/High-NA resist demand compounds, Fujifilm's top-five photoresist position and 30% Kumamoto capacity expansion for TSMC capture rising layer-count exposures.
- Bear
- Electronic materials are a small fraction of a diversified healthcare/imaging conglomerate, accessed only via the OTC ADR FUJIY.
Ibiden Co., Ltd.
Ibiden is one of the top FC-BGA package-substrate makers — a top-three-to-five supplier where no single maker holds more than ~18% share (Unimicron is usually cited as the largest) — and a leading substrate supplier into Nvidia's accelerator packages. Substrates sit in its Electronics segment (~65% of revenue), where AI-grade substrates are sold out with prices up 15-40% even as commodity PC substrates stay oversupplied. Its board approved ~¥500B ($3.3B) of high-performance IC-substrate investment across FY2026-2028 (Feb 3, 2026), targeting ~2.5x high-performance capacity. US access is the thin, unsponsored OTC line IBIDF.
- Bull
- If AI-substrate tightness persists, Ibiden's leading NVIDIA package-substrate supply, sold-out AI substrates at 15-40% higher prices and ¥500B capex multiply high-performance capacity.
- Bear
- No converter holds more than ~18% share and US access is the thin, unsponsored OTC line IBIDF, where substrates sit inside a broader electronics segment.
Merck KGaA (ADR)
Merck KGaA's EMD Electronics supplies photoresist, anti-reflective coatings, CMP materials, selective-etch chemistries and high-purity gases for leading-edge nodes. But Electronics was only 17% of 2025 group sales (semiconductor materials ~65% of that, ~EUR2.6B in 2024), inside a company whose larger businesses are Healthcare and Life Sciences. US access is the OTC ADR line MKKGY.
- Bull
- If leading-edge node materials demand grows, Merck KGaA's EMD Electronics photoresist, CMP and selective-etch franchise rides EUV/High-NA adoption across the supply chain.
- Bear
- Electronics is only 17% of group sales inside a Healthcare/Life-Sciences company, so the semiconductor exposure is real but diluted, via the OTC line MKKGY.
Shin-Etsu Chemical Co., Ltd.
One of only three firms (with JSR and Tokyo Ohka) holding 90%+ of the EUV photoresist market — the tightest materials chokepoint in the leading-edge supply chain — and also a major leading-edge silicon-wafer supplier. Photoresist is a minority of revenue; US exposure via unsponsored OTC ADR SHECY only.
- Bull
- If EUV-resist and leading-edge wafer demand tightens, Shin-Etsu's place among three firms holding 90%+ of EUV resist plus its silicon-wafer franchise capture the narrowest materials choke.
- Bear
- It is the world's largest PVC maker first, with resist a minority of revenue, accessed only via the speculative unsponsored OTC ADR SHECY.
Resonac Holdings Corporation (foreign ordinary)
Resonac (the former Showa Denko, renamed January 2023) supplies semiconductor front- and back-end materials including epoxy encapsulants, die-bonding materials, high-purity gases and CMP slurries. The semiconductor and electronic-materials segment is roughly 30% of sales inside a diversified chemicals group spanning mobility, chemicals and innovation-enabling materials. US access is the foreign-ordinary OTC line SHWDF, which is thin.
- Bull
- If back-end materials demand grows, Resonac's encapsulants, die-bonding materials, high-purity gases and CMP slurries ride AI packaging intensity across front- and back-end nodes.
- Bear
- Semiconductor materials are ~30% of a diversified chemicals group, and US access is the thin foreign-ordinary OTC line SHWDF.
SUMCO Corporation (foreign ordinary)
SUMCO is the second half of the 300mm wafer duopoly at roughly 23% of the market, producing ~1.8M 300mm wafers/month. Management projects AI-server wafer demand of 800k-1M wafers/month by 2027 and is exiting 200mm at Miyazaki by end-2026 to pivot to AI-grade 300mm. It is the purest wafer exposure but cyclical, inventory-sensitive, and accessed only through the foreign-ordinary OTC line SUOPY.
- Bull
- If AI-server wafer demand reaches 800k-1M/month by 2027, SUMCO's ~23% of the 300mm duopoly and pivot from 200mm to AI-grade wafers make it the purest wafer exposure.
- Bear
- SUMCO is mid-cycle and inventory-sensitive at only ~23% share, accessed solely through the foreign-ordinary OTC line SUOPY.
Tokyo Ohka Kogyo Co., Ltd. (foreign ordinary)
Tokyo Ohka Kogyo is a top-three EUV photoresist maker (TOK + JSR + Shin-Etsu hold north of 90% of EUV resist) and an Intel sub-2nm resist development partner, making it the purest listed resist play now that JSR is private. But its US surfaces are Pink-tier lines TOKCF and the unsponsored ADR TKCYY, which carry limited disclosure and minimal trading liquidity, so it is speculative for US investors. Resist is also one of two segments (Materials + Equipment) inside the company.
- Bull
- If EUV-resist demand grows, Tokyo Ohka's top-three position (>90% EUV resist with JSR and Shin-Etsu) and Intel sub-2nm partnership make it the purest listed resist play now JSR is private.
- Bear
- US surfaces are Pink-tier TOKCF/TKCYY with limited disclosure and minimal liquidity, and resist is one of two segments, so it is speculative at best.
Catalyst calendar
- 2026-07-01Commerce Section 232 datacenter-semiconductor market update due to the PresidenthighThe Secretary of Commerce must update the President on the datacenter-semiconductor market by July 1, 2026, framing whether phase-2 Section 232 tariffs broaden to cover more categories. Any expansion that sweeps in substrates, photoresist or wafers would directly reprice this Japan/Germany-anchored materials supply chain on geographic-concentration grounds.
- 2026-07-30Entegris Q2 2026 results (guided sales $815-845M)mediumEntegris is the only clean US large-cap in the chokepoint; its guided Q2 2026 sales of $815-845M and the read on CMP consumables, liquid filtration and selective-etch wins at the 2nm node are the cleanest listed proxy for advanced-node materials demand. Date estimated from the late-July cadence of prior quarters.
- 2026-08-04Ajinomoto Q1 FY2026 results (ABF film read)mediumAjinomoto's quarterly print is the only window into the ABF film near-monopoly: pricing, mix and the >50% operating margin that is now ~20% of group profit despite ABF being ~6% of revenue. The prior-year Q1 was reported August 4, 2025, so an early-August 2026 cadence is expected; date estimated from that pattern.
- 2027-03-31Ibiden Gama Plant / Ono Cell 8 AI-substrate ramp through FY2027mediumIbiden's ~¥500B ($3.3B) FY2026-2028 high-performance IC-substrate program (board-approved Feb 3, 2026) centers on the new Gama Plant (phase-1 ~¥220B, mass production from FY2027) and added capacity at the Ono Plant Cell 8, targeting roughly 2.5x high-performance capacity. The ramp is the cleanest read on whether AI-substrate demand is pulling through to real, qualified output; this is a program window, not a single confirmed event date.
- 2026-10-13SEMICON West 2026 (Moscone Center, San Francisco, Oct 13-15)lowSEMICON West is the materials/equipment industry's main venue; substrate, photoresist (incl. EUV/High-NA), wafer and CMP/filtration suppliers (Fujifilm, Entegris, Merck/EMD) use it to signal capacity, qualification wins and AI-driven demand. It is a recurring data point for the entire substrates/materials layer.