Author: Zihao
Recently, TOTO announced the construction of a new mass-production plant for core components of semiconductor manufacturing equipment in Buzen City, Fukuoka Prefecture. The investment amounts to approximately CNY equivalent of JPY 60 billion. Combined with an equipment investment of about JPY 11 billion across two plants in Nakatsu City, Oita Prefecture, the total reaches roughly JPY 71 billion.
Is it surprising that the toilet giant TOTO has crossed over into semiconductors? If you glance at the top gainers on the Tokyo Stock Exchange in 2026, you might get a strong illusion: the biggest winners of the AI computing wave are not NVIDIA's foundry supply chain or TSMC's ancillary factories, but rather a bunch of "brands you expect to see in supermarkets and building material stores." According to CNBC statistics from July 2026, by late July2026, TOTO's stock price had risen by 78%. Its most profitable segment is not sanitary ware, but the electrostatic chuck (ESC) it has been producing for over thirty years. Ajinomoto, known for MSG, saw its stock rise by 61%, having turned amino acid fermentation by-products into ABF build-up film, effectively becoming the "monopolist of AI packaging materials." Nitto Boseki, a textile fiberglass manufacturer, rose 63% because CoWoS substrates suffer from thermal expansion warping, and they integrated T-Glass low-expansion fiberglass cloth into AI server packaging substrates. Kao Corporation, a shampoo seller, was renamed in research reports as the "leader in AI cleaning materials" after applying surfactant technology to create precision cleaners for 2nm wafers. Even Fujifilm is no longer just a "film factory," but has converted its silver halide colloid coating experience into photoresists and CMP slurries.
These companies do not design chips, build wafer fabs, or manufacture GPUs. However, they respectively control how wafers are "held," how they are "insulated," how they avoid "warping," how they are "cleaned," and how photoresist is evenly coated. Moreover, this is not a sudden pivot or chasing trends into semiconductors; rather, they have repositioned decades-old processes within chip factories.
01. Powders and Furnaces: Another Undervalued Path
TOTO is undoubtedly a century-old enterprise. Founded in 1917, its core competencies include ceramic glazing, water efficiency control, and seat heating. In 1984, it applied the molding and firing technologies of sanitary ceramics to industrial ceramics, establishing a ceramic business division. For the subsequent thirty years, multiple product lines lost out to domestic and international competitors, resulting in long-term losses. Japanese media called this a "healthy loss-making business," meaning the company continued investing despite knowing it wasn't profitable. The turning point came in 2016 when it cut loss-making product lines and concentrated resources on electrostatic chucks.
During etching, deposition, and ion implantation, wafers must remain stationary. Mechanical clamps scratch wafers; vacuum chucks largely fail in plasma vacuum environments. Therefore, advanced nodes use "electrostatic chucks": relying on electrostatic fields within ceramic dielectric layers to "stick"300mm silicon wafers to the surface while controlling temperature, conducting heat, and resisting plasma bombardment. How demanding are the requirements? Surface flatness must be at the micrometer level; temperature uniformity fluctuations must be less than 10°C; materials must resist plasma, shed no particles, and not contaminate the backside of the wafer; during low-temperature etching of 3D NAND, the surface must operate stably at -100°C or lower; holes with aspect ratios exceeding 50:1 must be etched hundreds of millions of times without any jitter. A single jitter ruins an entire wafer.
Why can TOTO do this? Because making high-end sanitary ceramics and making semiconductor precision ceramics share the same underlying fundamentals: alumina powder formulation + molding + high-temperature sintering + micro-defect control + dimensional precision. TOTO uses alumina with purity above 99.4%, doping titanium dioxide to adjust resistivity, stabilizing volume resistivity between 10⁸-10¹¹Ω·cm. Kiln sintering temperatures approach 1700°C, controlling grain uniformity, densification, warping, and micro-cracks. This expertise, originally used to fire toilets that are "leak-proof, crack-free, and have extremely flat glazes," translates directly to ESCs that are high-purity, low-particle-contamination, highly plasma-resistant, and temperature-controlled.
When TOTO first started making electrostatic chucks, it did not immediately turn a profit. It suffered long-term losses for decades until the AI boom drove capacity expansion for 3D NAND and advanced logic, finally triggering explosive growth. More importantly, ESCs are consumables. When fab utilization rates rise, chuck wear necessitates replacement, typically within a lifespan of less than two years. Thus, this is not "selling equipment once," but continuously collecting revenue from replacement parts.
Today, TOTO's profit structure has changed. In fiscal year 2025 (ending March 2026), TOTO's total revenue was approximately JPY 737.4 billion. The "Advanced Ceramics" business generated JPY 67.4 billion in revenue, accounting for only about 9% of the group, yet contributed operating profits of JPY 28.9 billion, with an operating margin of about 43%, providing more than half of the group's total operating profit. Meanwhile, the residential equipment business, which accounts for over 60% of revenue, has a margin of just over 4%. This 9% of the business is essentially "electrostatic chucks."
By 2024, in the global market for wafer electrostatic chucks, Shinko Electric Industries led with about 44%, while TOTO ranked second with about 17%. Furthermore, TOTO holds a unique card that others lack. Aerosol Deposition (AD) technology allows ceramic particles to impact the substrate at high speed at room temperature, solidifying directly into a film, overturning the conventional wisdom that "ceramics must be sintered at high temperatures." This phenomenon was discovered by the National Institute of Advanced Industrial Science and Technology (AIST) in Japan in 1994. TOTO joined joint research in 1999 and achieved commercialization in 2011.
It sprays plasma-resistant yttrium oxide ceramic films onto the inner walls of etching chambers, reducing dust generation to one-tenth or one-twentieth of previous levels and extending component life by approximately tenfold. In 2015, it won the Prime Minister's Award at the "Manufacturing Japan Grand Prize." In June of this year, it disclosed plans to invest JPY 80 billion in semiconductor materials over the next five years, with JPY 39 billion already committed, targeting the 1-nanometer node.
The logic here is clear. Sanitary ceramics and semiconductor plasma ceramics share the entire powder engineering system involving powder blending, molding, sintering, densification control, and surface modification. While the tolerance margins allowed for sanitary products are far larger than those required inside etching chambers, the craftsmanship remains identical.
02. From MSG to Insulating Film: Separated by One Formulation Line
Consider another incredible legend—Ajinomoto. Its most famous story is how an MSG factory produced the key insulating film for chip packaging—ABF. ABF stands for Ajinomoto Build-up Film. High-end CPUs, GPUs, and AI accelerator dies, once diced, must be mounted onto packaging substrates. These substrates contain dozens of circuit layers requiring insulation between them, low dielectric loss, extremely flat surfaces, direct laser drilling and copper plating capability, thermal stability, and low expansion. Traditional epoxy inks cannot handle high-density fine line widths. ABF is a dry-film type insulating build-up material that can be directly pressed and patterned. ABF surface roughness must be controlled to one-thousandth the width of a human hair. Slight unevenness causes distortion in 112G/224G SerDes signals, significantly dropping AI chip yields.
How did it originate? In the 1970s, Ajinomoto obtained a class of high-purity insulating resin precursors from amino acid fermentation by-products. Engineers did not treat them as waste but spent twenty years developing them towards "film-forming, modifiable, and highly insulating" properties. Around 1996, Intel approached them: high-density FC-BGA packaging lacked suitable insulating films. Ajinomoto developed amino-acid-derived polymers into thin films, commercializing them in 1999 under the name ABF. The result: over 90% global market share in high-end ABF; a single H100 consumes 8-10 times more ABF than a standard PC processor; AI servers, data centers, GPUs, and CPUs all depend on it; during capacity crunches, lead times stretch beyond 30 weeks. Ajinomoto plans to invest JPY 25 billion before 2030 to expand production. Why can't others copy it? Not because "the MSG recipe is secret," but because the precursors come from amino acid fermentation chemistry, where purity control relies on decades of fermentation separation experience; film formation, modification, and process coordination with packaging factories have undergone thirty years of iterative development; qualification cycles for substrate factories at TSMC, Intel, AMD, and NVIDIA take years, and switching suppliers requires rerunning yield validation. Ajinomoto didn't "switch careers to make chips"; it always operated on an amino acid chemistry platform. Previously, amino acids went into soup; now they go into packaging interlayers.
This demand was not a track prepared in advance by material manufacturers, but a blank space collided upon by process iteration itself.
Ajinomoto did not develop a new discipline from scratch. It redirected existing polymer synthesis and ultra-thin coating capabilities toward new metrics. The true barrier for ABF lies not in the formula itself, but in resin monomer synthesis, film-forming processes, batch consistency, interface compatibility with copper traces, and the comprehensive ability to accompany substrate factories through lengthy reliability validations. The result is over 95% global market share. Once they expand capacity, scheduling across the entire industry shifts accordingly.
03. Cleaning Wafers and Removing Grease Use the Same Interfacial Logic
Kao Corporation started in 1887 with a bar of facial soap. To avoid being constrained by raw materials, it later developed its own oleochemistry, beginning to sell surfactants in 1951. Since then, it has mastered the entire craft of interfacial control: wetting, penetration, dispersion, and stripping. Concrete admixtures, waste paper de-inking agents, and copier toner binders seem unrelated, but their core is the same thing.
Entry into semiconductors was driven by regulations. The Montreal Protocol was signed in 1987, putting freon used for cleaning circuit boards on a phase-out countdown. Alternatives had to simultaneously clean effectively, not corrode substrates, rinse easily, and allow for waste liquid treatment. The industry was temporarily stuck between "compliant options were ineffective" and "effective options were non-compliant." Finally, in 1991, Kao provided the answer using aqueous surfactant composite systems.
More than thirty years later, this line became the brightest growth driver in financial reports. The semi-annual report for fiscal year 2026 shows it holds about 60% share in precision cleaners for packaging substrates and ranks first in share for specific front-end NAND process chemicals. Its customer list includes TSMC, Samsung, and Kioxia.
Cleaning accounts for about 30% of the front-end process steps. Although the monetary value is far less than equipment, it directly determines yield. Line width and spacing on high-end ABF substrates have entered below 10 micrometers. Micro-vias are deep and narrow; if drilling residues, plating remnants, and flux are not thoroughly cleaned, short circuits occur. The difficulty also lies in selectivity. Copper, resin, nickel, and gold coexist on the same substrate. The chemical solution must penetrate fine structures to remove dirt without damaging any substrate material. This happens to be the exact problem type daily chemical formulation engineers are best at solving.
Kao has also systematized this business. It established Precision Cleaning Centers in Wakayama and Hsinchu, opening cleaning and stripping processes for client testing. The Hsinchu center, launched in December 2025, is its first overseas facility, serving advanced packaging lines like CoWoS, CoWoP, and FOPLP. Once materials are embedded in mass production processes, switching suppliers requires clients to rerun full reliability validations, making market share exceptionally stable.
04. A Cloth and a Film Determine How High Substrates Can Stack
Nitto Boseki is a company founded on textiles. In 1984, it introduced T-glass, a special fiberglass cloth with high elasticity and low coefficient of thermal expansion (CTE). Initially used for composites, it was later found to be ideal for reinforcing packaging substrates. As chips grow larger and substrate layer counts increase, warping becomes a fatal issue.
The company holds about 90% global share in this category. New-generation products reduce the CTE from 2.8ppm to 2.0ppm. Facing severe supply shortages triggered by AI servers, it made a counter-intuitive business decision: instead of raising prices, it maintained stable pricing to secure capacity expansion and consolidate market share, while simultaneously increasing its medium-term capital expenditure budget by 50%.
A piece of cloth seems far removed from chips, but the maximum size of a substrate, the number of stackable layers, and achievable yield depend on the formulation and weaving method of this cloth. It waited nearly forty years for AI to turn its legacy products into scarce commodities.
05. Printing Factories Guard the First Gate of Lithography
Dai Nippon Printing (DNP) began making photomasks in 1961. Today, it is the world's largest independent third-party photomask supplier, holding about 30% of the third-party market share. In 2022, it spun off this business into Tekscend Photomask, retaining 50.1% ownership.
Toppan Holdings occupies about 20% of the photomask market and is advancing into the 2-nanometer node and EUV processes.
Photomasks are masters used to transfer circuit patterns onto wafers. The core competency of printing factories is pattern transfer. The logic hasn't changed; only the precision requirement differs by six orders of magnitude.
HOYA is even more surprising. It is the world's second-largest eyeglass lens manufacturer and simultaneously the absolute dominant player in mask blanks, holding over 60% share in mask substrates and over 75% in EUV mask substrates. Substrates are the blanks for masks, requiring ultra-low defect density, extreme flatness, and controlled thermal expansion—essentially an extension of optical glass processing capabilities. AGC is another major player in this market.
Fujifilm's transformation is the most complete. After the decline of its film business, it shifted its particle dispersion, precision coating, and optical design capabilities from the silver halide era entirely to semiconductor materials.
From fiscal year 2021 to 2025, semiconductor material sales grew at an average annual rate of about 20%. Over five years, it invested more than JPY 100 billion. Its new plant in Oita will be completed in August 2026. It holds the global number one position in CMP slurry for copper wiring. Hybrid bonding slurry for advanced packaging has been adopted by major chipmakers, and next-generation HBM processes are under validation. In the first half of fiscal year 2025, electronic material sales reached JPY 107.3 billion, with the electronic materials business growing by 11.9%.
A company famous for film now makes money from slurries. This is not an inspirational story; it is a capability list sold to a different buyer.
06. What Is Truly Scarce in Chips Is Not Theory, But Hands That Push Metrics to the Limit
The public tends to view semiconductors as an independent discipline, assuming entry tickets require a background in semiconductor engineering. However, breaking it down reveals that many challenges in chip manufacturing do not belong to semiconductor physics, but to general foundational engineering. Material purity, interface control, powder densification, film uniformity, morphological precision, particle management, and motion control in cleanroom environments exist in other industries too, albeit with much looser tolerance margins.
What the chip industry truly does is compress the tolerances of these metrics to the limits of human craftsmanship. When an enterprise polishes the same underlying skills in its main business for decades, what it holds is not a specific product formula, but a complete set of process systems, equipment modification experiences, quality control standards, and talent pipelines. This asset can be repurposed for a different downstream application.
Therefore, true cross-industry integration is not "changing careers," but "changing customers."
07. Why This Cross-Industry Phenomenon Appears Only in Japan
Connecting these cases reveals a common path. They did not first select semiconductors as a target and then supplement technology; rather, they possessed mature craftsmanship first and then sought a downstream willing to pay for the ultimate version. The former mindset is "I want to make semiconductors," while the latter is "My skillset is exactly what semiconductors need pushed to the extreme."
They crossed the "technical foundation," not the "industry label." TOTO didn't cross from "sanitary ware to semiconductors," but from "ceramic sintering to high-purity ceramic sintering." Ajinomoto didn't cross from "MSG to chips," but from "amino acid polymers to insulating films." Kao didn't cross from "shampoo to wafers," but from "surface activity to precision cleaning."
Semiconductor material substitution costs are extremely high: customer validation takes 1-3 years; changing materials requires rerunning yield, reliability, and failure analysis; a single contamination incident eliminates the supplier. Therefore, early movers who endure loss periods build a moat based on time. TOTO's ESCs lost money for decades; Ajinomoto's ABF R&D took twenty years; JSR took forty years to transition from rubber.
Those old enterprises seemingly unrelated to chips are securing their positions in this most critical high-tech track in a quiet, slow manner heavily reliant on process accumulation. They submitted their answers early, but slowly. So slow that others refused to wait, and waiting became a barrier.