Tri Chemical Laboratories Inc. (4369.T) - Deep Dive Research Report

Prepared 2026-05-31. Fiscal year ends 31 January. "FY2026" = year ended 31 January 2026.

1. What the company does

Tri Chemical Laboratories makes the ultra-pure chemical compounds that semiconductor factories use to grow atom-thin films on silicon wafers. These compounds are called precursors. When a chip maker wants to deposit a layer of material - say, an insulating film a few atoms thick on top of a DRAM capacitor - they do not pour on a solid. They flow in a vapour of a carefully designed molecule that carries the target element (hafnium, zirconium, titanium, silicon, tungsten and so on), let it react on the wafer surface, and build the film one atomic layer at a time. The precursor is that carrier molecule. Tri Chemical designs it, synthesises it, and purifies it to a cleanliness measured in parts per billion and parts per trillion.

The company was founded in December 1978 in Sagamihara, Kanagawa, and is now headquartered in Uenohara, Yamanashi. It did not begin in semiconductors. Its first business was supplying ultra-pure chemicals for optical fibre manufacturing, where glass purity determines how far a light signal can travel before it degrades. Learning to make and handle chemicals at extreme purity for fibre optics gave the company a capability that turned out to be exactly what the semiconductor industry needed as chips shrank. As transistors got smaller and silicon dioxide stopped being good enough as an insulator, the industry needed new materials - high-dielectric ("high-k") films built from elements like hafnium and zirconium - and it needed someone who could synthesise novel molecules and purify them obsessively. Tri Chemical pivoted into that gap. Today semiconductor precursors are the core of the business, with optical fibre, compound semiconductor and coating materials around it.

The core value proposition is narrow and deep: when a chip maker or an equipment maker needs a new film for a next-generation process, they need a molecule that does not yet exist, made to a purity that almost nobody can hit, qualified into a production line that costs billions. Tri Chemical's pitch is that it will sit inside the customer's R&D from the start, invent the molecule, and then make it reliably at production purity. The hard part is not one of those things - it is doing all of them together, repeatably, for a customer who cannot tolerate a single contaminated batch.

A useful way to hold the business in your head: Tri Chemical sells grams and kilograms of liquid that look unremarkable, but each one is a custom-designed molecule purified to parts-per-trillion, qualified into a fab line where switching it out means re-validating the entire process. That combination - novel chemistry plus extreme purity plus qualification lock-in - is the whole company.

A concrete example. A DRAM maker wants to improve the capacitor that stores each bit of memory. To pack more capacitance into a smaller space, it deposits a high-k dielectric on the capacitor using atomic layer deposition (ALD). It needs a hafnium-bearing molecule that vaporises cleanly, reacts at the right temperature, leaves no carbon residue, and arrives free of metal contaminants that would short the device. Tri Chemical co-develops that molecule with the customer (and its deposition method, branded HAC), patents it, scales the synthesis, purifies it, and ships it. Once that precursor is locked into the customer's qualified process, the customer keeps buying it for the life of that process node, because changing it means re-qualifying a production line.

2. Business segments

Tri Chemical reports its financials as a single business segment - high-purity chemicals - so there is no segment revenue table to dissect. But the company organises its products into distinct application areas with genuinely different technology, customers and competitive dynamics, and management and the company website present the business through those areas. Treating them as sub-segments is the right way to understand the company even though the income statement does not split them out.

Si semiconductor (the core)

This is the engine. It covers the precursors used to build the films inside logic chips, DRAM and 3D NAND flash. Two families matter most. High-k precursors (hafnium, zirconium, titanium based) are used in DRAM capacitors and gate stacks, deposited by ALD to make ultra-thin high-dielectric films that hold charge in a shrinking footprint. Low-k / silicon-based precursors are used for the silicon-oxide and related films in 3D NAND, where the industry keeps stacking more layers vertically and needs precursors that deposit uniform films deep into high-aspect-ratio structures. The core capability here is the combination of molecular design and parts-per-trillion purification, plus the deep co-development relationships with chip makers and deposition-equipment makers. This is where the AI/datacenter demand wave lands hardest, because advanced DRAM (including the memory that feeds AI accelerators) and leading-edge logic both consume more and newer high-k materials.

Compound and oxide semiconductor

Metal-organic (MO) precursors for compound semiconductors - gallium, indium, germanium and related compounds. These feed LEDs, power devices, RF/communications chips and other non-silicon devices. The chemistry overlaps with the core (handling reactive metal-organics at high purity) but the customers and end markets are different, which is why it exists as its own area. It is a smaller, more specialised business than mainstream silicon precursors.

Optical fibre

The company's founding business. Chemicals for fibre-optic preform manufacturing, where purity controls signal loss. It is the historical root that built the purity capability the rest of the company now lives on. It is a mature, lower-growth area relative to semiconductors, but it is strategically important as the origin of the firm's process knowledge and as a steady, non-semiconductor end market.

Coating

Coating materials and solutions, again leaning on the same synthesis-and-purity skill set applied to surface films outside mainstream chip fabrication.

Special reagents

Specialised chemical reagents and catalysts - lower-volume, higher-customisation products that round out the catalogue and keep the company embedded with research-stage customers.

Services

Analytical and contract work: purity and impurity measurement, thermal-stability testing, film-formation tests, and contract synthesis. This is the part that makes Tri Chemical a development partner rather than a vendor. A customer can bring a problem, have Tri Chemical run the analysis and film tests, and co-develop a new precursor. It is low revenue but high strategic value - it is the on-ramp that turns into qualified production business years later.

Note one structural feature that the single-segment income statement hides: a meaningful slice of group profit comes through equity-method joint ventures, not direct sales. In FY2026 ordinary profit (which includes equity-method income and FX) ran well above operating profit, a gap driven substantially by the JVs - especially SK Trichem, which supplies hafnium precursor to SK Hynix. Understanding Tri Chemical means understanding that some of its most valuable franchises are held through partnerships rather than on its own books (see Sections 4 and 5).

3. Products and business detail

The catalogue is wide. Tri Chemical says it can make more than 2,000 chemicals, built around precursors carrying a long list of elements: Al, B, Br, C, Co, Er, F, Fe, Ga, Ge, Hf, In, La, Mg, Mo, N, Ni, P, Pt, Ru, Sb, Se, Si, Ta, Ti, W, Zn and Zr. Named products that recur in its materials include trimethylaluminum (TMA) and triethylaluminum (TEA), titanium tetrachloride (TiCl4) and tetrakis(dimethylamino)titanium (TDMAT), tantalum pentachloride and pentakis(dimethylamino)tantalum, tungsten and molybdenum chlorides, and a range of zirconium and hafnium compounds and alkoxides.

The products that matter most commercially:

• Hafnium high-k precursors - the crown jewel. Used to deposit high-dielectric films on DRAM capacitors by ALD. Tri Chemical's deposition method is branded HAC and is patent-protected. This is the product at the centre of both its biggest growth driver (advanced DRAM for AI) and its biggest legal fight (the Merck patent dispute, Section 8).
• Zirconium and titanium high-k precursors - companion high-k materials for capacitor and gate films.
• Silicon-based (low-k) precursors - for the SiO2 and related films in 3D NAND, where rising layer counts increase consumption per wafer.
• Tungsten, molybdenum, cobalt, ruthenium precursors - metallisation and contact/interconnect materials as the industry moves to new metals for scaled interconnects.
• MO precursors (Ga, In, Ge) - for compound semiconductors.
• Optical-fibre chemicals - the founding line.

What makes these hard to make is not the recipe; it is the purity and the consistency. The competitive bar is parts-per-billion to parts-per-trillion control of metallic and particulate contamination, batch after batch, because a single contaminated lot can scrap a customer's wafers and trigger a re-qualification. According to the independent deep-dive analysis, even if a competitor were handed the published recipe, it could not replicate the result, because the tacit process knowledge - how to actually hit and hold that purity at scale - lives in the company's hands, not in the patent.

Manufacturing is concentrated in Japan (Uenohara, Yamanashi), with a Taiwan subsidiary (Tri Chemical Electronic Materials Taiwan, Miaoli County), a China subsidiary (Tri Chemical Laboratories China, Shanghai), and a South Korea office in Suwon. The international footprint mirrors where the customers' fabs are: Taiwan for foundry/logic, Korea for memory, China for its expanding domestic fab build-out, Japan for equipment makers and domestic fabs. The company funds its capacity expansion from its own cash - the balance sheet ran an equity ratio above 80% with no reliance on debt - which lets it add capacity without leverage but also means capex discipline is self-imposed.

The two joint ventures are central to the product story:

• SK Trichem (with SK Materials, part of the SK Group) - licenses Tri Chemical's precursor patents and supplies hafnium precursor to SK Hynix. Reported SK Trichem volumes have run on the order of 220 billion won annually, with roughly 150 billion won going to SK Hynix.
• HBR Co. (with Air Liquide Japan) - a partnership with one of the global industrial-gas and electronic-materials majors.

These JVs let Tri Chemical's chemistry reach customers and geographies it could not serve alone, and they convert its IP into a profit stream that flows back through equity-method income.

4. Customers

The customers are the world's advanced chip makers and the equipment/materials companies that serve them, concentrated in East Asia. The company sells into Japan, Taiwan, China and South Korea, with China and Taiwan now the two largest markets by revenue (China and Taiwan each materially larger than Japan in the most recent nine-month disclosure), Korea and others behind.

The buyers fall into a few types:

• Memory makers (DRAM and NAND). SK Hynix is the clearest named relationship: it sources its hafnium precursor through the SK Trichem JV, which holds the right to use Tri Chemical's patents. This is a deep, structural tie. By contrast, Samsung sources its hafnium precursor from Adeka, a Japanese competitor - a reminder that Tri Chemical does not own every memory maker.
• Foundry / logic makers, concentrated in Taiwan and increasingly China.
• Deposition-equipment makers, who qualify precursors alongside their ALD/CVD tools and influence which materials get designed into a process.

The buying decision is made by the customer's process-integration and materials engineers, not a procurement desk, and the sales cycle is measured in years. A precursor gets designed into a process during the customer's R&D phase, tested for film quality, contamination, thermal behaviour and yield impact, then qualified into a specific production line. Customers choose Tri Chemical for three reasons that are specific rather than generic: it can invent the molecule the customer needs and does not yet have; it can hit purity levels competitors struggle to match; and it embeds itself in the customer's development from the start, so by the time the process is qualified the relationship is years deep.

Switching costs are the heart of the investment case. Once a precursor is qualified into a production line, swapping it for a competitor's product means re-validating the entire process - new film tests, new reliability data, new yield qualification - which is expensive, slow, and risks the customer's output. The independent analysis put it plainly: switching "requires revalidating the entire manufacturing line," which makes defection economically irrational as long as the incumbent performs. That lock-in is per-process, so the franchise is rebuilt at each new node - which is both the strength (sticky once won) and the vulnerability (must keep winning the next node).

Concentration is real and is a double-edged feature. A handful of megacap chip makers drive the demand, and the deepest tie (SK Hynix) runs through a JV rather than a direct contract. That concentration reflects quality - you do not get designed into SK Hynix's DRAM without being good - but it also means a single customer's capex cycle, or the loss of a single qualification, moves the business. Contract structure is a mix of qualified, recurring production supply (the predictable base, tied to the customer's running processes) and development/spot work that seeds the next generation. Revenue therefore tracks customers' production volumes and node transitions more than a fixed order book.

5. Competitive landscape

The precursor industry is an oligopoly of specialists. There is no single dominant player across all materials; instead a small set of companies compete material-by-material and customer-by-customer, and the winner of a given precursor at a given customer often holds it for the life of the process.

The named competitors:

• Adeka (Japan) - a direct rival in hafnium high-k precursors and the supplier Samsung uses for its hafnium precursor. The clearest head-to-head competitor in Tri Chemical's most important product.
• Merck / EMD Performance Materials (Germany/US) - a materials major and the counterparty in an active patent dispute over hafnium precursor technology in Korea (and, reportedly, expanding to China). Merck competes both commercially and through litigation.
• Mecaro (South Korea) - a Korean hafnium-precursor maker that SK Hynix has at times evaluated as an alternative source.
• SK Materials / SK specialty (South Korea) - simultaneously a partner (through SK Trichem) and a player in the broader precursor space.
• Air Liquide (France) - a global electronic-materials and industrial-gas major; partner through the HBR JV but also a large independent force in deposition materials.
• Entegris, Versum, Gelest, Materion, DNF, Soulbrain (US/Korea) - the wider field of precursor and electronic-materials suppliers across different materials and geographies.

Where Tri Chemical wins: novel molecules it has invented and patented (the hafnium/HAC franchise), extreme purity, and incumbency at customers like SK Hynix where it is locked into qualified processes. Where it is exposed: it does not own every customer (Samsung buys hafnium from Adeka), some of its strongest positions sit inside JVs it does not fully control, and its most valuable IP is under active legal challenge from a larger competitor.

Barriers to entry are high but not absolute. The barriers are real - molecular design capability, parts-per-trillion purification know-how that does not transfer with a recipe, patents, and qualification lock-in that takes years to earn. A new entrant cannot simply undercut on price, because price is almost irrelevant next to the cost of a failed qualification or a contaminated batch. But the barriers are per-material and per-customer: a competitor that wins the next node's hafnium precursor at a key customer can displace the incumbent at the point of a process transition, and patents can be challenged (as Merck is doing). The structural shifts to watch are the build-out of Chinese domestic fabs (a demand tailwind and a geopolitical risk), the move to new interconnect metals (ruthenium, molybdenum - new precursor opportunities), and the litigation over who actually owns the hafnium-precursor IP.

This is a genuine moat - qualification lock-in plus tacit purity know-how is hard to replicate - but it is a moat that must be re-dug at every node and is currently being tested in court. It is not a fortress that competition cannot reach.

6. Industry

Demand for Tri Chemical's products is driven by one thing above all: the advance of semiconductor manufacturing. Three vectors of that advance each create precursor demand. Miniaturisation forces the industry off old materials onto new high-k and metal films that need new precursors. 3D stacking (taller 3D NAND, stacked DRAM, advanced packaging) multiplies the number of deposition steps and the volume of precursor consumed per wafer. New materials - hafnium, zirconium, ruthenium, molybdenum, cobalt - each require a molecule designed to deposit them. Every time the industry pushes any of these three, companies that can invent and purify new precursors gain.

The immediate cyclical driver is the AI wave. Generative-AI datacenter build-outs and the migration of AI features into consumer devices have driven a surge in advanced DRAM and leading-edge logic demand, which flows straight through to precursor orders. That is the explicit explanation management and analysts gave for the sharp acceleration in the most recent year.

Industry size: the ALD/CVD precursor market is a multi-billion-dollar global market within the broader electronic-materials industry, with Asia-Pacific (Korea, China, Japan, Taiwan) holding the largest share of consumption - roughly 40%+ - because that is where the fabs are. The hafnium materials market specifically has been put at around USD 1.2 billion in 2024 growing toward USD 2.5 billion by 2034 (roughly 8% CAGR) by market researchers, though hafnium precursors are a slice of that and estimates vary by source. The precise figures matter less than the shape: a specialised, growing market gated by qualification rather than price.

Tri Chemical sits upstream in the supply chain - it supplies the chip makers and equipment makers, not the end electronics market - so it is one step removed from consumer demand and one step closer to fab capex and utilisation. Import-substitution dynamics are relevant in two directions: Japan is a stronghold of high-purity materials supply globally, and China is building domestic fabs that pull in precursor demand while also raising the prospect of eventual local substitution. Regulation matters through semiconductor export controls and supply-chain fragmentation - the materials are dual-use-sensitive and the customer geographies (China especially) carry policy risk.

Cyclicality: the semiconductor industry is cyclical, and precursor demand follows fab utilisation and memory pricing. FY2024 to FY2025 illustrates this - after a strong period the business was roughly flat through a semiconductor downturn year, then re-accelerated sharply in FY2026 as the AI cycle lifted advanced memory and logic. The consumable nature of precursors (used up every wafer) makes the business somewhat less violent than equipment spending, but it is not immune to the cycle.

Tailwinds: AI-driven advanced-memory demand, rising deposition-step counts, new-material transitions, Chinese fab build-out. Headwinds: semiconductor cyclicality, export-control fragmentation, and the risk that customers localise or dual-source.

7. Growth triggers

Drawn from the four most recent quarterly disclosures (the company files Japanese-language earnings releases and briefing materials rather than English transcripts; statements below are attributed to those disclosures and accompanying management commentary).

• AI/datacenter-driven advanced-memory demand is the active growth engine. Management and the accompanying analysis attributed the sharp revenue and profit acceleration to generative-AI proliferation and large-scale datacenter investment lifting advanced DRAM and logic, which pulls high-k precursor demand. (Q3 FY2026 disclosure, 28 Nov 2025; reiterated through FY2026 results, 13 Mar 2026.)

Generative AI proliferation accompanied by large-scale datacenter investment drove the surge in semiconductor orders, lifting demand for the company's high-purity precursors. (paraphrase of management/analyst commentary, Q3 FY2026, 28 Nov 2025)

• China and Taiwan growth as the two largest markets. Both markets grew materially year-on-year in the most recent disclosures and are now each larger than Japan, signalling Tri Chemical is riding the China domestic fab build-out and Taiwan foundry strength. (Q3 FY2026 disclosure, 28 Nov 2025.)
• FY2027 top-line guidance of double-digit sales growth. Management guided FY2027 (ending Jan 2027) to roughly +13% sales growth, signalling continued volume expansion. (FY2026 results, 13 Mar 2026.)
• Margin expansion sustaining into the new year. Q1 FY2027 showed operating margin rising sharply year-on-year (into the mid-20s percent from low-20s), indicating the higher-value high-k mix and operating leverage are carrying forward. (Q1 FY2027 disclosure, 30 May 2026.)
• Self-funded capacity expansion. The company continues to add capacity from its own cash (80%+ equity ratio, no debt reliance), positioning it to absorb the demand surge without external financing. (FY2026 results, 13 Mar 2026.)
• JV-driven profit through SK Trichem. The SK Hynix-facing hafnium-precursor franchise inside SK Trichem feeds equity-method income that lifts ordinary profit above operating profit; continued SK Hynix advanced-DRAM ramp is a forward driver. (structural, evidenced across FY2026 results, 13 Mar 2026.)
• New-material precursors (interconnect metals, new high-k). The company's R&D pipeline targets the new elements the industry is adopting (ruthenium, molybdenum and others) as scaling forces material changes. (company R&D positioning, referenced through FY2026 results.)

8. Key risks

Hafnium-precursor patent litigation with Merck. This is the most specific and material risk. Merck has appealed a Korean court ruling that upheld Tri Chemical's hafnium-precursor patents, and has reportedly signalled intent to expand litigation to China, which could touch SK Hynix's Wuxi DRAM operations. Mechanism: the hafnium/HAC franchise (directly and through SK Trichem) is a core profit source; an adverse ruling could weaken the IP barrier, open the door to competitors at key customers, or disrupt supply at a customer fab. This is a moderate-probability, high-impact risk because it strikes the exact franchise that drives the company's premium economics. Management's own exposure here is structural - the dispute is public and ongoing.

Customer concentration and JV dependence. A handful of megacap chip makers drive demand, and the deepest relationship (SK Hynix) runs through the SK Trichem JV rather than a direct, fully-controlled contract. Mechanism: a single customer's capex pullback, a lost qualification at a node transition, or a change in the JV relationship could move group profit sharply, and the equity-method income that lifts ordinary profit above operating profit is concentrated in that structure. High-probability, moderate-to-high impact - this is the everyday reality of the business, not a tail event.

Not winning every customer. Samsung sources hafnium precursor from Adeka, not Tri Chemical. Mechanism: the franchise is per-customer and per-node; competitors hold real positions, and each node transition is a fresh contest. Losing the next node at an existing customer would erode the recurring base. Moderate probability, moderate impact, continuous.

Semiconductor cyclicality. Precursor demand follows fab utilisation and memory pricing. The FY2024-FY2025 plateau through a downturn year shows the business is not immune. Mechanism: an AI/memory demand air-pocket would slow volumes and compress the operating leverage that is currently inflating margins. High probability over a cycle, moderate impact.

Geopolitical / export-control exposure to China. China and Taiwan are now the two largest markets, with China growing fastest. Mechanism: tightening export controls, supply-chain fragmentation, or the eventual localisation of precursor supply by Chinese fabs could cut off or erode the fastest-growing market. The Merck-to-China litigation angle compounds this. Moderate probability, moderate-to-high impact.

Conservative-looking earnings guidance and the OP/ordinary-profit gap. FY2027 net-income guidance was set down year-on-year even as sales were guided up, because ordinary profit carries FX and equity-method income that management is guiding cautiously. Mechanism: a stronger yen or softer JV income would pull reported profit down independent of the core business doing well. This is more a reporting-quality nuance than an operating risk, but it means headline profit can move on items outside the precursor business.

Technology disruption. A breakthrough deposition method that needs fewer or different precursors could erode demand for an existing franchise. Low probability on any near horizon, high impact if it happens - this is the long-tail risk for any consumable-materials franchise tied to a specific process approach.

9. Walk the talk

The four most recent quarterly disclosures used here:

• H1 / Q2 FY2026 (six months to 31 Jul 2025) - released 29 Aug 2025
• Q3 FY2026 (nine months to 31 Oct 2025) - released 28 Nov 2025
• FY2026 full year (year to 31 Jan 2026) - released 13 Mar 2026
• Q1 FY2027 (three months to 30 Apr 2026) - released 30 May 2026

The most recent is 30 May 2026, one day inside the 90-day window from today (31 May 2026). A caveat that frames this whole section: Tri Chemical is a Japanese small/mid-cap that discloses through kessan tanshin and Japanese-language briefing decks, not verbatim English earnings-call transcripts. So this is an assessment of guidance versus reported outcomes across four disclosures, not a parse of management's spoken words. Direct quotes are therefore limited and paraphrased from the disclosures and accompanying analyst summaries.

The pattern that emerges is conservative guidance consistently beaten. Through the first half of FY2026 the business was running far ahead of the prior year - first-half sales up about 55% and operating profit up about 64% year-on-year - yet management held its full-year targets unchanged (around the low-¥23bn sales / ¥5.5bn operating-profit area it had set). By the nine-month mark, the company had reached roughly 83% of its full-year operating-profit target with a quarter still to run, and still did not raise guidance. That is the signature of management that sets targets it is confident of clearing and lets the beat come through rather than chasing the print.

Full-year guidance was held unchanged even as nine-month results reached ~83% of the full-year operating-profit target. (Q3 FY2026, 28 Nov 2025)

The outcome validated the conservatism: FY2026 landed above the held guidance, with sales up about 26% and operating profit up about 12% for the year. So on the one trackable, datable commitment that ran across these disclosures - the full-year target held through H1 and Q3 - management delivered and then some.

Then a tell worth noting. For FY2027, despite guiding sales up double digits, management guided net income down year-on-year. The reason is not weakness in the core business - Q1 FY2027 promptly delivered ordinary profit up more than 50% and a sharp jump in operating margin - but caution on the FX and equity-method income that sit below operating profit. This is consistent with the conservative posture: management appears to prefer to guide the volatile, below-the-line items down and let them surprise to the upside, rather than bake in a strong yen assumption and risk a miss. Q1 FY2027 immediately running hot against that cautious full-year frame reinforces the read.

The plain assessment: this is management that under-promises and over-delivers. Across four disclosures they held guidance through a period when they were obviously beating it, delivered above the held numbers, and set the next year's profit guide conservatively. There is no instance in this window of a target quietly dropped or a promise missed. The one thing to watch is that "conservative guidance + beat" is easy to sustain in an up-cycle driven by AI demand; the real test of credibility will come when the semiconductor cycle turns and the question becomes whether they guide the downturn honestly. On the evidence available, they do what they say.

10. Shareholder friendliness index

Dividends. Post the 1-for-4 stock split in February 2021, dividends per share were ¥20 (FY2022), ¥30 (FY2023), ¥30 (FY2024), ¥35 (FY2025), ¥35 (FY2026), with FY2027 guided again at ¥35. Over the last three fiscal years the trajectory is: held flat at ¥30, raised to ¥35, then held flat at ¥35 and guided flat again. So the dividend has been essentially static at ¥35 for three consecutive years even as earnings grew strongly. The payout ratio is low - against FY2026 EPS of roughly ¥170, a ¥35 dividend is about a 20% payout, and that ratio is falling as earnings rise faster than the dividend.

Buybacks and dilution. There is no active buyback programme evident in the disclosures reviewed; the company funds capacity expansion from retained cash and carries an equity ratio above 80%. Shares outstanding are roughly flat at about 32.4-32.5 million (32,453,300 as of 31 Jan 2026), with no material buyback-driven shrinkage and no large option-driven dilution. The count is essentially stable, neither retiring stock nor expanding it meaningfully.

Verdict: Hoards Capital. Tri Chemical retains the great majority of its earnings to self-fund growth and build balance-sheet strength, holding the dividend flat and a falling payout ratio while running an 80%+ equity ratio - capital is being kept inside the business, not returned.

11. Insider activities

For a Tokyo-listed company the available "insider" disclosures are EDINET Large Shareholder Reports (大量保有報告書) and their change reports (変更報告書), which capture 5%+ holders. Director-level open-market dealing is not disclosed in the granular, per-transaction form available in the US or UK, so this section relies on the 5%-rule filings. The standout signal is the founder/chairman steadily selling.

Recent material filings (most recent first):

Buys - the signal. The buying in this window is institutional, not insider. Nomura's rising position is best read as a trading/market-making book rather than strategic conviction, and Wellington Management's new ~5.66% stake plus SMTAM's and Mitsubishi UFJ's positions reflect index/active-fund flows into a strong-momentum AI-materials name. None of these are management or board members buying with their own money, so none carries the strong "insider conviction" signal that an open-market purchase by a director would.

Sells - the why. The recurring seller is Takenaka Junpei, the chairman and a founder-level individual shareholder, who has filed multiple decreases in a tight window: from 12.81% down through 11.70% to 10.66% between mid-March and mid-May 2026 - roughly a 2.6 percentage-point reduction in about two months. The filings do not disclose a reason. The most common benign explanations for a founder/chairman trimming a large holding are personal diversification, estate/tax planning, or monetising into a strong share price; selling into a period of record results and a rising stock is consistent with diversification rather than a signal about the business. But the reason is not disclosed in the filings reviewed, so it should be labelled as such rather than assumed.

Net assessment. Insiders (specifically the chairman) are net sellers; institutions are net buyers. The activity that matters is concentrated in one person - the chairman steadily reducing a founder-scale stake - against a backdrop of institutional accumulation. There is no insider buying to provide a bullish counter-signal. Read: mild concern. A founder trimming into strength is common and often benign, but a steady, multi-filing reduction with no disclosed rationale, and no offsetting insider purchases, is not a positive signal and is worth monitoring. It is not a red flag (the pace is gradual and he remains a top-three holder), but it is the opposite of the cluster-buying pattern that would be bullish.

12. Scenarios

Bull case. The AI build-out keeps pulling advanced DRAM and leading-edge logic, and every new node consumes more high-k and more new-material precursors per wafer. Tri Chemical's hafnium franchise survives the Merck litigation intact, the patents hold, and the SK Trichem JV keeps scaling as SK Hynix ramps advanced and stacked DRAM. China's domestic fabs keep buying, Taiwan foundry demand stays strong, and the company wins design-ins for the next interconnect metals (ruthenium, molybdenum) as the industry adopts them. Self-funded capacity comes on line just as demand peaks, operating margins stay in the mid-20s as the high-value mix deepens, and the conservative guidance keeps getting beaten. The invisible, indispensable materials supplier compounds quietly as the semiconductor materials intensity of each wafer keeps rising.

Base case. Management delivers roughly what it guides: double-digit sales growth, operating profit broadly stable to modestly higher, and reported net income that comes in cautious on FX and JV income but with the core business clearly healthy. The hafnium litigation grinds on without a decisive adverse ruling. SK Hynix stays the anchor relationship; Samsung stays with Adeka. China and Taiwan remain the growth markets, with the usual semiconductor-cycle wobble somewhere in the next couple of years that flattens a year before demand resumes. The dividend stays around ¥35 and the company keeps retaining cash to fund expansion. A high-quality, sticky niche business growing with its end market, with profit that occasionally zig-zags on below-the-line items.

Bear case. Merck wins or escalates the hafnium-precursor dispute - in Korea, or by expanding to China where it could entangle SK Hynix's Wuxi fab - and the IP barrier around the crown-jewel franchise weakens, letting competitors like Adeka or Mecaro contest Tri Chemical's positions at key customers. Simultaneously the AI memory cycle cools into an air-pocket, fab utilisation drops, and the operating leverage that inflated recent margins reverses. Export controls fragment the China market or accelerate domestic precursor substitution, cutting off the fastest-growing region. The chairman's steady selling turns out to have been early. Concentration bites: a lost qualification at a node transition or a JV relationship change at SK Hynix removes a chunk of profit that is hard to replace because the franchise is per-customer and per-node. None of this requires a fraud or a blow-up - just the convergence of cyclical softness, a legal loss, and geopolitical friction on a business whose strengths are also its concentrations.

Note on Section 13 (Further Reading): a search of SemiAnalysis, Stratechery, and MBI Deep Dives returned no qualifying in-depth coverage of Tri Chemical Laboratories, so that section is omitted.

Sources:

• Tri Chemical FY2026 results & FY2027 guidance (Nikkei disclosure)
• Tri Chemical Q3 FY2026 disclosure (Nikkei)
• Tri Chemical H1 FY2026 interim disclosure (Nikkei)
• Tri Chemical Q1 FY2027 results (Kabutan)
• Tri Chemical Q3 FY2026 deep analysis (note.com / jstock_lab)
• Tri Chemical moat / barriers-to-entry deep dive (note.com / tatsuya_sabato)
• Tri Chemical corporate profile (English)
• Tri Chemical product list
• Tri Chemical dividend history (IRBank)
• Large shareholding reports database (M&A Online)
• Major shareholders (Kabutan)
• Merck vs Tri Chemical hafnium-precursor patent dispute (THE ELEC)
• SK Hynix / Mecaro / Hf precursor sourcing (THE ELEC)
• Hafnium market size & suppliers (Mordor Intelligence)
• CVD & ALD precursor market (Business Research Insights)
• Tri Chemical second-quarter FY2026 EPS (Webull/SimplyWallSt feed)

A note on what I could and couldn't verify: financial figures come from Japanese exchange disclosures and reputable aggregators; multi-year revenue values for FY2024-FY2025 in the charts are approximate (derived from reported growth rates) and labelled as such. Insider data is limited to EDINET 5%-rule filings because Japan does not publish granular per-director open-market dealing - the chairman's reductions are the key signal there.