Fujimi: The Hidden Champion of CMP and Silicon-Wafer Polishing Materials, but 3,680 Yen Already Prepays a Good Deal of the Next Few Years

Vertical history and financial review

Company vertical history

Fujimi was born in 1950, early in Japan's postwar industrialization, originally named Fujimi Kenmazai Kogyosho and founded in Nagoya. The entry point was direct: domestic supply of abrasives was short while manufacturing upgrades were lifting demand for precision machining, and the company started out as the first domestic producer of abrasives. It incorporated in 1953, expanded the Inazawa plant in the 1970s, established sales and then production bases in the United States during the 1980s, and in 1991 formally renamed itself Fujimi Incorporated after merging its sales and related companies. Its growth path was walked step by step through process accumulation, moving close to overseas customers and extending into niche applications, without rolling up through acquisitions.

The company's real turning point was technological rather than the listing itself: the shift from traditional abrasives to high-purity, high-consistency semiconductor materials. It developed the PLANERLITE series of CMP slurry in 1995 and launched a new series for copper CMP in 2000, cutting into one of the highest-barrier consumables just as semiconductor interconnect and planarization processes were expanding. That decision shaped the company we see today: it sits where the customer's yield constraint is tightest and makes consumables that must be qualified by the process, building neither tools nor commodity chemicals.

The first-stage capital-markets narrative was Japanese precision industrial materials exporter. The company registered over the counter with the Japan Securities Dealers Association in 1995, moved to JASDAQ in 2004, listed simultaneously on the first sections of the Tokyo and Nagoya exchanges in 2007, and transferred to the Prime Market in 2022 with the Tokyo Stock Exchange reform. The listing path is orderly, with no SPAC, reverse merger or complex restructuring, which matches a management style that is conservative, manufacturing-heavy and oriented to long customer relationships.

Splitting the history into four stages makes the logic clearer. The first, from the 1950s to the early 1990s, was import substitution and the accumulation of basic manufacturing capability, centered on building out powder, processing and quality systems and beginning to internationalize. The second, from the mid-to-late 1990s to around 2010, was the formation of the semiconductor precision-materials business, as CMP and silicon-wafer-related materials gradually became the real profit pool. The third, from the 2010s to the fiscal year ended March 2023, was global expansion and advanced-node penetration, with CMP revenue growing faster than the semiconductor market and both margin and ROE rising. The fourth, from the fiscal year ended March 2024 to now, is a phase of riding through the cycle while re-levering to expand: the 2024 profit decline showed demand is not linear, but the fiscal years ended 2025 and 2026 recovered quickly, accompanied by cross-region expansion, added lines in Taiwan and the United States, a new Japanese plant and a second R&D center, moving the operating model from a low-capex high-return asset toward a growth platform with a clear reinvestment character.

Several milestones still bear directly on today. The first is the 1995 entry into CMP products, which established that what it wanted to make were consumables inside the process window rather than ordinary abrasives. The second is the production and sales footprint across the United States, Taiwan and Malaysia, which lets it sit close to fab customers worldwide today while keeping most Asian sales inside a nearby supply system. The third is the new medium-to-long-term plan launched in 2023, which positions the company explicitly as a Powder and Surface Company and sets a sales target near 100 billion JPY for the fiscal year ending 2030, meaning management is no longer content to defend its traditional abrasives advantage and wants a second growth curve in non-semiconductor and non-polishing directions. The fourth is the large-scale borrowing and capex that began in fiscal 2026, which will raise depreciation and pressure returns over the next two years, though if advanced-node demand absorbs this capacity it could also turn the supply bottleneck into a new share moat.

Financial vertical review

Start with revenue and profit. Fujimi produced 51.731 billion JPY of revenue and 12.059 billion JPY of operating income in the fiscal year ended March 2022, rose further to 58.394 billion and 13.243 billion in the fiscal year ended March 2023, then fell sharply to 51.423 billion and 8.251 billion in the fiscal year ended March 2024, with operating margin dropping from 22.7% to 16.0%. That downturn shows that although it sells consumables and is therefore less violent than equipment, it can still be dragged by fab inventory adjustments, customer utilization and downstream end demand. The recovery was fast: the fiscal year ended March 2025 brought 62.503 billion of revenue and 11.78 billion of operating income, and the fiscal year ended March 2026 brought 69.404 billion and 13.826 billion, with margin back at 19.9%. That pattern of quick repair after a downturn is steadier than a pure equipment cyclical and stronger than commodity chemicals.

More important is the change in revenue mix. The company's fiscal 2026 sales by application show silicon-wafer-related products at 21.081 billion JPY, up 3.1%; semiconductor device CMP products at 36.135 billion JPY, up 17.9%; hard-disk-related at 2.238 billion JPY, down 12.1%; and general industrial plus functional materials and thermal spray combined at 9.88 billion JPY, up 13.2%. Placing that mix back on the long-term chart makes it just as clear: CMP grew from roughly 15 billion JPY in the fiscal year ended 2018 to 36 billion JPY in the fiscal year ended 2026 and has become the absolute core engine; the silicon-wafer business grew from about 12 billion to 21 billion and is relatively steady; the hard-disk business keeps shrinking as a share. Fujimi today is a heavily semiconductor-weighted, and especially heavily CMP-weighted, materials company rather than a diversified abrasives maker.

Earnings quality is acceptable overall. On public company data, cumulative operating cash flow from the fiscal year ended 2022 to the fiscal year ended 2026 ran at about 1.11 times net income attributable to owners. The fiscal year ended 2023 was the weak point in that ratio, mainly because operating cash flow that year was 7.377 billion JPY against 10.594 billion JPY of net income attributable to owners, noticeably weaker than other years, but the fiscal years ended 2025 and 2026 returned to about 1.38 and 1.39 times. Over the long run this is not a company whose accounting profit is stacked up on paper, and cash conversion is broadly healthy. Using depreciation as a rough proxy for maintenance capex, owner earnings for the fiscal year ended 2026 come to about 10.236 billion JPY, or roughly 138 JPY per share, close enough to accounting EPS of about 122 JPY that the headline P/E is not badly overstating real cash generation.

The balance-sheet shift deserves watching. As of the end of March 2025 the company still carried almost no interest-bearing debt; by the end of March 2026, current plus long-term borrowings had reached 17.231 billion JPY, with long-term borrowings at 16.16 billion JPY and equipment-related payables at 4.092 billion JPY, while fixed assets jumped from 34.779 billion JPY to 60.116 billion JPY, within which the net book value of buildings rose from 6.868 billion to 26.713 billion. At the same time cash and deposits still stood at 31.346 billion JPY, net assets at 84.715 billion JPY and the book equity ratio at 69.1%, nowhere near dangerous, but the zero-leverage cash-cow picture is out of date. The company is trading balance sheet for future capacity, which will make capital returns fluctuate while the new lines ramp.

The direction of that investment also has clear industrial logic. The medium-term capital plan management gave at the results briefing is specific and points at three places: the new Kakamiyama plant in Japan to expand silicon-wafer and CMP product capacity; expansion at the United States subsidiary to widen the CMP production footprint; and the Taiwan site installing new facilities in stages to add CMP capacity. There is also a second R&D center focused on new non-semiconductor businesses. The schedule management gave is that the new Japanese plant broke ground in October 2024, completed at the end of 2025, ships samples in 2026 and reaches full production in 2027, while the United States and Taiwan projects add lines more flexibly. Capex has genuinely become heavier, but the direction is not scattered.

Price and valuation history

Fujimi's valuation center has migrated noticeably over the past few years. When profits peaked in the fiscal years ended 2022 and 2023, the market began repricing it from a mid-cap Japanese industrial materials stock to an advanced semiconductor consumables stock; that growth label took a hit when profit fell in the fiscal year ended 2024; and after the recovery in the fiscal years ended 2025 and 2026 the multiple rose again. By the end of July 2026 the 52-week range was roughly 2,015 JPY to 5,090 JPY, with the July 28 close at 3,680 JPY, meaning the market has been switching back and forth between a high short-term multiple and worry about earnings delivery.

The current multiple sits at the expensive end of the high-quality materials group. On net income attributable to owners of 9.059 billion JPY for the fiscal year ended 2026, earnings per share come to about 122 JPY, implying a trailing P/E of about 30x; on the company's guidance of 10.4 billion JPY of net income for the fiscal year ended 2027, expected EPS is about 140 JPY, implying a forward P/E of about 26x. That level is not absurd, but it is clearly no bargain. Two things have to keep holding for the current price to be supported: CMP demand must keep growing faster than the industry average, and the depreciation from new capacity must not depress margin for too long.

Business model, moat and industry cycle

Business model and moat

The most worthwhile thing about Fujimi is that its revenue structure and its technology structure lock into each other. On the revenue side, CMP and silicon-wafer polishing materials are the profit core. On the technology side, what the company keeps emphasizing is the coupling of three process capabilities, powder, filtration and classification and purification, and chemistry, rather than brand. That phrasing sounds like a materials company's marketing line, but in a CMP setting it carries hard industrial meaning. Advanced CMP slurry has to control particle distribution, impurities, stability, reaction selectivity and long-run batch consistency, which is far more than mixing abrasive grain and chemicals together. If Fujimi really does make more of the critical abrasive grain and purification in-house, it can customize and iterate the process for customers faster than a pure formulation supplier. The company's disclosed core-technology and product pages emphasize exactly this point repeatedly.

But making abrasive grain in-house does not automatically mean overwhelming pricing power. The share structure Fujimi itself gives already tells the story: its share in FEOL CMP is high, yet its share across all CMP steps combined is only 13%. That means customers very likely use different suppliers at different material layers and process stages, and that Fujimi's position is very strong in certain steps while its pricing power across the whole CMP basket is partly diluted by multi-supplier strategies. So I prefer to define its moat as qualification lock-in plus deep process embedding, rather than industry-wide dominant scale pricing power. Those two sound close and differ a great deal in practice.

The second layer of the moat is customer qualification. Fujimi does not give a standard qualification period in public disclosure, but industry trade sources generally describe qualification of CMP slurry at an advanced fab as taking 12 to 24 months, while advanced-process chemicals more broadly can take 18 to 36 months from qualification to production. More persuasive still, the Chinese challenger Anji Microelectronics lists "qualified and supplying stably at multiple customers" and "sales achieved through qualification at more advanced nodes" as key results in its annual report, which shows the real bottleneck in competition is still getting into the customer's process flow. Fujimi's strength is therefore to take a seat in the process window first and then stay there for a long time on stability, not to quote a lower price.

The third layer of the moat is a globalized supply footprint. By the fiscal year ended 2026, Fujimi's sales destinations were roughly 15 billion JPY in Japan, about 46 billion in Asia and Oceania, about 6 billion in North America and about 3 billion in Europe, putting overseas sales close to 78% of the total, with Asia and Oceania alone at 66%. From a customer-service standpoint that footprint matters: the further front-end materials move toward advanced nodes, the less willing customers are to accept a distant, long-chain, unpredictable supply system. Fujimi has manufacturing or key operating nodes in the United States, Taiwan, Malaysia and Japan, and almost all of the new expansion points at the CMP and silicon-wafer businesses, unlike many small materials makers that can only export from a single country.

On management and governance, the company shows the typical features of Japanese manufacturing: managers are mostly developed internally and rotate for long stretches through the United States, CMP and silicon-wafer business lines. Keishi Seki served as president from 2008 and moved to chairman in June 2026; Katsuhiro Suzuki came up through the United States subsidiary and the silicon-wafer and CMP businesses and took over as president and representative director in June 2026. That kind of succession lowers the risk of an abrupt strategic break and suggests the next several years will continue along the existing semiconductor-materials path rather than attempting a large acquisition-driven transformation.

Industry and cycle

Fujimi does not sit in a single market. It is exposed at once to silicon-wafer processing materials, semiconductor CMP consumables, hard-disk substrate lapping, and general industrial and thermal-spray materials. What clearly decides the valuation center of the stock is CMP and silicon wafers, and both follow a consumables logic tied to wafer starts and process steps rather than the equipment logic of one-off purchases tied to capital spending. Outside public estimates of the CMP materials market use different definitions, but the direction is quite consistent: on TECHCET's basis, the global semiconductor CMP polishing materials market was about 3.8 billion USD in 2025 and about 4.2 billion USD in 2026, compounding at about 8.8% from 2025 to 2030; market research focused on CMP slurry gives a range of about 2.1 billion to 2.6 billion USD in 2025 with mid-to-high single-digit growth into the first half of the 2030s. The numbers do not agree, but they all point to one conclusion: this is a process-materials market that is not small, grows faster than traditional chemicals, and is not easily displaced.

The demand mechanism matters more. Anji Microelectronics writes it plainly in its annual report: CMP slurry accounts for more than 50% of the value of polishing materials, and its consumption rises with wafer output and with the number of CMP planarization steps. SEMI likewise shows global silicon-wafer shipment area up 5.8% year over year in 2025 and up another 13.1% year over year in the first quarter of 2026, explicitly naming AI data centers, advanced logic and memory demand as drivers. Put those two together and Fujimi's revenue elasticity becomes easy to understand: it is of course affected by the cycle, but the more direct variables are actual wafer starts at fabs and process complexity, rather than equipment makers' order intake. The further advanced nodes go and the more complex material and wiring layers become, the higher the CMP value per wafer, and that is the root of why bulls treat it as a growth consumable rather than an ordinary cyclical material.

Policy and geopolitical risk cut both ways for Fujimi. On one side, the global focus on semiconductor supply-chain security is pushing Japan, the United States and Taiwan to localize production, and Fujimi benefits as a domestic and near-shore materials supplier. On the other, Chinese domestic substitution is accelerating too, and Anji Microelectronics' annual report already lists production introduction and domestic-substitution qualification for advanced-node copper and copper-barrier, tungsten and ceria abrasive slurries, plus several advanced-packaging and new-material slurries, as core progress. Fujimi does not disclose its China revenue share in public materials, so its direct China exposure cannot be quantified precisely; but it has a sales-support entity in Shenzhen and a high Asia and Oceania weighting, which means a slowdown in Chinese fab demand or an acceleration in local substitution would still transmit back through regional orders and share pressure.

One more point needs correcting. The starting facts given to me said Asia-Pacific accounted for about 38.9% of the CMP slurry market in 2025, but the public market research I found generally gives a much higher figure, roughly in the 63% to 77% range. So I will not use the 38.9% number. Whichever basis is used, the main competitive battleground is in Asia, and on that there is no doubt.

Horizontal competition and current fundamentals

Horizontal competitive landscape

Placing Fujimi in a horizontal comparison, rather than hunting for whoever looks identical to it, it is better to split the field into three groups. The first is the closest global CMP consumables platform, represented by Entegris. Through the historical CMC Materials assets and subsequent integration it holds a more complete modular combination of CMP slurry, pads, filtration and post-CMP cleaning, and the company itself puts CMP slurries, pads, post-CMP chemistries and slurry filters into one synergy narrative. This group's advantage is a wider product stack and cross-module synergy; its weakness is that it is not necessarily deeper than Fujimi at every single front-end step.

The second group is the Chinese domestic-substitution challengers, represented by Anji Microelectronics. Where it most resembles Fujimi is that it also focuses on high-end front-end wet electronic chemicals and also writes heavily about nano abrasives, a proprietary technology platform and qualification progress at customers. Its annual report discloses that advanced-node copper and copper-barrier slurry has been qualified and is supplying stably at multiple customers, that a multi-product portfolio of advanced-node tungsten slurry is supplying stably, and that self-produced ceria abrasive has been loaded into several slurries now in volume sales. In other words, Anji is not simply a low-price mature-node chaser; it is already climbing toward more advanced nodes and higher value added. This is exactly the type of competitor Fujimi most needs to watch.

The third group is the broader Japanese electronic-materials majors such as Tokyo Ohka Kogyo and Shin-Etsu Chemical. They do not fight Fujimi head-on in CMP, but they represent how capital markets value critical advanced-process materials. TOK relies on high-end photoresist and high-purity chemicals and has expanded its share of premium products in generative-AI-related semiconductors in recent years; Shin-Etsu is the integrated flagship of Japanese semiconductor materials, spanning silicon wafers, lithography materials and more. Capital markets use these companies as the valuation anchor for high-quality Japanese semiconductor-materials assets, and because Fujimi is smaller, more focused and more purely exposed to CMP it should normally carry some growth premium, though that premium should not expand without limit.

Current fundamentals and the bull-bear split

First, straighten out the fiscal-year labels completely. Fujimi's company materials state clearly that the accounting year runs from April 1 to March 31 of the following year, so FY2026 on the company's basis means the fiscal year ended March 31, 2026, not calendar 2026. As of this report's base date of July 29, 2026, the company had not yet released first-quarter results for the fiscal year ending March 2027; the IR calendar shows Q1 results due in August 2026. So the latest complete first-hand disclosure is the full-year result through March 2026 plus the earlier half-year materials. Figures given to me such as polishing slurry growing 18.9% in the most recent quarter most likely belong to an earlier basis and cannot be used directly as the latest quarterly fact as of the end of July 2026.

In the latest complete fiscal year the fundamentals are strong. The fiscal year ended March 2026 brought revenue of 69.404 billion JPY, up 11.0%; operating income of 13.826 billion, up 17.4%; ordinary income of 14.169 billion, up 15.7%; and net income attributable to owners of 9.059 billion, down 3.9%, though that decline came mainly from about 1.22 billion JPY of additional tax for prior years and about 370 million JPY of impairment loss rather than from deteriorating operations. By business, CMP revenue was 36.135 billion, up 17.9%, and the clear main engine; silicon-wafer polishing slurry was 13.384 billion, up 5.4%; the hard-disk business kept declining. Management's guidance for the fiscal year ending March 2027 is 74.8 billion of revenue, 14.5 billion of operating income and 10.4 billion of net income attributable to owners, which says plainly that demand is still there, margin is slightly affected by depreciation, and earnings still reach a record.

What the market is trading now is which matters more, demand delivery or depreciation pressure. Bulls hold onto AI-related demand, the material complexity of advanced logic and memory, Fujimi's high FEOL share, and the company's expectation that CMP revenue continues toward 41.3 billion JPY in the fiscal year ending 2027. Bears hold onto several other things: first, a share across all CMP steps of only 13%, which says it is not an all-round platform supplier; second, the new plant and global line additions that will push depreciation up noticeably after the fiscal year ending 2027; third, Chinese domestic substitution pushing toward more advanced nodes; and fourth, a valuation that is no longer cheap. The share price falling from a 52-week high of 5,090 JPY to 3,680 JPY on July 28 is itself evidence that the market does not disagree about this being a good company, and disagrees a great deal about what price is reasonable.

The core of the bull-bear split condenses into one sentence: is Fujimi a long-term compounding asset holding critical FEOL seats, or a high-quality cyclical consumables company that has temporarily earned a high multiple because AI demand improved? I lean toward the former being closer to the truth, while holding that the latter imposes a real constraint on the share price. The business quality and technical barrier are real, and the thin valuation margin of safety is equally real.

Valuation analysis

Historical and peer valuation reference

At the current price, Fujimi trades at roughly 30x trailing earnings for the fiscal year ended March 2026 and about 26x forward earnings. Horizontally, that is clearly above the roughly 22x of the broader Japanese materials leader Shin-Etsu and slightly above TOK's mid-to-high 20s, but clearly below the Chinese high-growth challenger Anji Microelectronics, whose public-market basis shows a forward P/E around 57x and an adjusted P/E higher still. That valuation position shows the market treats Fujimi as an in-between asset, high quality but still cycle-constrained: more growth than an integrated materials leader, steadier than the most aggressive domestic-substitution story.

Anji being more expensive does not make Fujimi cheap, though. What decides Fujimi's entry point is how much profit and cash flow it can deliver over the next 12 months to three years; how high the most expensive peer can rise has nothing to do with that entry point. The Japanese 10-year government bond yield has risen to about 2.75%, which means an equity with only low single-digit expected returns has no visible margin of safety. As a high-quality materials stock Fujimi should of course carry a premium, but a 30x trailing P/E in Japan's 2026 rate environment already demands fairly sustained delivery of growth.

Cash-flow pass-through and absolute valuation

Run owner earnings first. On five-year cumulative data from the fiscal year ended 2022 to the fiscal year ended 2026, the company's operating cash flow was about 1.11 times net income attributable to owners, so overall cash conversion is not bad. Operating cash flow in the fiscal year ended 2026 was 12.599 billion JPY against depreciation of 2.363 billion JPY; treating depreciation as an approximation of maintenance capex puts owner earnings for the fiscal year ended 2026 at about 10.236 billion JPY, or roughly 138 JPY per share. On the July 28 close of 3,680 JPY, the headline P/E is about 30.1x while the implied multiple on owner earnings is about 26.7x, a gap of only about 11%, well inside the 30% warning line, so the valuation work can stay anchored on the earnings basis with the cash basis as a cross-check.

I use a three-scenario method with only three core variables: whether CMP growth keeps outrunning the industry, how long depreciation from the expansion suppresses margin, and whether Chinese or localized competition first produces real share erosion at advanced nodes.

Prices in the table are scenario valuations within this research framework and do not constitute investment advice. One-year returns in the table are a rough calculation on the 2026-07-28 close of 3,680 JPY and an expected dividend of 77 JPY for the fiscal year ending 2027. Company guidance for the fiscal year ending 2027 is 74.8 billion of revenue, 14.5 billion of operating income and 10.4 billion of net income attributable to owners; on dividend policy, the full-year dividend for the fiscal year ended 2026 was 75 JPY, with 77 JPY planned for the fiscal year ending 2027.

The answer this set of valuations gives is not ambiguous. In the conservative case the current price has no margin of safety; in the base case it corresponds to a very limited return; only in the optimistic case, where CMP demand stays strong, share is not eroded and new capacity is absorbed smoothly, does the return become clearly attractive. In other words, what the current price buys is continued success at the company, not cheapness. For existing holders and for new buyers, those are two different things.

Expectation gap and margin-of-safety review

What the market implies now sits roughly between the base and optimistic cases. If the market only believed the company would walk steadily along guidance, a 26x forward P/E would not look clearly undervalued; it can stand at this level because investors assume Fujimi keeps outrunning ordinary materials stocks on advanced nodes and AI-related demand. The metrics that could actually create an expectation gap are three things rather than total revenue: whether CMP growth stays faster than total revenue; whether operating margin can hold near 19% through the depreciation upcycle; and whether customer qualification and new-product introduction start to be penetrated by Chinese domestic substitution.

The margin-of-safety conclusion is equally direct. First, the current price is well above the conservative-case valuation, so in the strict sense there is no discount to conservative value. Second, the most fragile assumption is that the expansion delivers supply capability rather than a prolonged margin drag; cutting confidence in profit delivery in the base case to 70% pulls base value quickly toward 3,000 JPY. Third, if profit is roughly flat over the next three years and the multiple does not expand, holding-period returns come down largely to the dividend plus scattered growth, which is not an overwhelming advantage over Japanese government bond yields in 2026. My margin-of-safety conclusion is thin. This is a good company, but for new money the price does not leave a wide enough allowance for being wrong.

Risks, catalysts and tracking

Core risks

The most dangerous business risk is gradual erosion of advanced-node share; short-term semiconductor swings matter less. The transmission path for that risk will not be orders going to zero one day: it starts with mainland China and some Asian customers, localizing niche steps such as copper and copper-barrier, tungsten and ceria-related slurries, taking mature nodes and regional customers first and then pushing into more advanced nodes. Anji Microelectronics' annual report already shows that path in motion: advanced-node copper, tungsten and self-produced ceria abrasive slurries all have multi-customer qualification and volume-production progress. If that substitution starts to extend beyond mainland China, or accelerates within advanced logic and advanced memory, Fujimi's high-share niches would first lose the increment and then lose pricing power. I put the probability at medium and the impact at high.

The second risk is that returns on the expansion fall short. Fujimi has switched from almost no interest-bearing debt to borrowing for expansion, with the new Japanese plant, the United States line expansion and the Taiwan additions all in progress. Management itself flags that operating income for the fiscal year ending 2027 will grow only slightly, because the new plant and similar projects begin to book depreciation. If advanced-node demand after 2027 falls short, or customer introduction runs slower than the release of new capacity, the income statement absorbs depreciation first while revenue does not necessarily follow. That risk would not threaten the company's survival, but it would damage the valuation center. Probability medium, impact medium-high.

The third risk is high share with weaker pricing power than imagined. This is really a perception risk. Many investment narratives describe Fujimi as number one in the CMP world and then naturally infer strong pricing power. The fact the company itself discloses is that its share across all CMP steps is only 13%, and the genuinely high shares are in FEOL and silicon-wafer niches. As long as customers keep to multi-supplier, step-by-step diversified procurement, Fujimi looks more like a strong supplier at several critical positions than the price setter for the whole CMP basket. Once the market starts to notice that, the valuation premium compresses before profit does. Probability medium, impact medium.

The fourth risk is the invisibility of China exposure. The company discloses by region down to Asia and Oceania rather than mainland China, so it is hard to know direct China sales and China end exposure precisely from public statements. China risk of course exists; the question is how large and how fast-moving, and public disclosure is not granular enough. If Chinese domestic substitution accelerates markedly in future while the company still does not clearly disclose China revenue and order changes, investors may use the multiple to reflect the uncertainty before it shows up in profit. Probability medium, impact medium.

The fifth risk is rates and style. By July 2026 the Japanese 10-year JGB yield stood at about 2.75%. Fujimi is a high-quality growth materials stock rather than a high-dividend defensive; assets like that naturally face multiple compression when rates and risk-free yields rise. Even if results are fine, once the market switches from paying a growth premium to prioritizing cash returns, a trailing P/E around 30x has a hard time going higher. Probability medium, impact medium.

Catalysts and tracking indicators

The most valuable positive catalysts are a few verifiable events; a general AI tailwind does little. First, CMP revenue in the Q1 and Q2 reports continuing to run clearly ahead of total revenue, proving the company is still capturing advanced-node penetration and rising step counts. Second, sample shipments and full production at the new Japanese plant proceeding on schedule without depreciation visibly eating margin. Third, if the company updates its self-estimated shares to 2025 or 2026 in later IR and keeps FEOL high, or lifts its all-CMP share, the high multiple gains support. Fourth, if management can disclose more cross-growth signals on materials tied to advanced logic, HBM and advanced packaging, the market would raise its confidence in medium-to-long-term growth again.

Negative catalysts are more concrete. First, CMP growth in Q1 or Q2 coming in clearly below guidance. Second, operating margin falling below 19% with the explanation resting mainly on depreciation rather than one-off factors. Third, more frequent news of large-customer volume introduction by Chinese challengers at advanced nodes or in advanced packaging. Fourth, operating cash flow failing to grow alongside a period of high capex, which would start to cast doubt on the high-quality-growth label.

These indicators matter because each corresponds to a different gear in the Fujimi investment case. CMP revenue relative to total revenue decides whether the company is still becoming more semiconductor-weighted, more step-intensive and higher value added; operating margin and cash flow decide whether the expansion is turning into value creation rather than value dilution; regional mix and competitor progress decide whether the moat is loosening in Asia; and rates and valuation decide how high a discount multiple the market will grant this kind of high-quality asset. On the timing of the next results disclosure, the company's IR calendar already sets first-quarter results for early August 2026.

Cross-synthesis conclusion

What history has really proven about Fujimi is that it keeps standing in the position of having to solve the problem alongside the customer as customer processes keep evolving; landing one or two best-selling formulations comes nowhere near summing it up. It started in traditional abrasives and ended up a critical materials supplier for front-end CMP and silicon-wafer polishing, and that road was walked by turning powder, purification and chemistry into one system and embedding that system inside the quality management of fabs worldwide, rather than by acquisition engineering or by simply riding a good cycle. The most persuasive of the company's self-estimated shares are FEOL at 56% and the 59%, 84% and 92% in several silicon-wafer niches, rather than the 13% across all CMP. What those numbers mean for the business is that at some of the steps most sensitive to yield and defectivity, it does hold a long-run right to keep its seat.

Put the vertical history and the horizontal competition together, though, and Fujimi's past success is not mysterious either. It caught the era's dividends of advanced nodes, more CMP steps and the globalization of manufacturing, and it also relied on long-run steady execution by management and deep coordination with customers to turn those dividends into higher margin and higher share. Most of those factors are still in place today; only the marginal relationship has changed. The biggest positive in the past was the global expansion of advanced processes, while the biggest variable ahead is whether Chinese substitution will penetrate the higher-end steps that used to be steadier, and whether its own new capacity can be fully absorbed before the depreciation peak. In other words, Fujimi's moat is still there, but it has left the comfortable stage where industry growth alone carried the company and entered a new stage of both defending share and proving the new capex was worth it.

Horizontally, its advantage over Entegris is single-point depth and Japanese-style manufacturing stability; over Anji it is accumulated global customers and existing qualification lock-in; over TOK and Shin-Etsu it is purer exposure to CMP and front-end planarization with higher growth elasticity. Its weaknesses are equally clear: less platform breadth than Entegris, more direct pressure from the domestic-substitution counterattack than either TOK or Shin-Etsu, and a valuation more expensive than an integrated materials leader. The capital market is not underrating its quality at present; on the contrary, it has already prepaid a considerable part of its success over the next few years. The current price looks more like a reward for capability proven in the past plus an advance purchase of the probability of continued success.

There are two places I think the market is most likely to get wrong. First, it is easy to misread high FEOL share as unchallengeable pricing power across all of CMP; in fact the company itself has told you its share across all CMP steps is only 13%, which makes this business more a combination of critical positions. Second, it is easy to misread consumable as low volatility; in the fiscal year ended 2024 Fujimi already proved that profit is clearly affected by the cycle, with recovery merely faster than for tools and wafer equipment. Bulls who ignore the first point overestimate how far the moat extends; bears who ignore the second underestimate its ability to repair profit after passing through a cycle. The true answer sits in between.

The most critical variables over the next year are whether CMP revenue growth can stay faster than total revenue, whether operating margin can hold near 19% through rising depreciation, and whether Q1 and Q2 show early signals of advanced-process demand slowing. Over three years, the key variable is whether the new plant and the United States and Taiwan line expansions can convert global supply capability into secure share rather than drag margin down. Over five years, the key variable is whether Chinese and Asian localized competition will genuinely break through Fujimi's qualification lock-in at more advanced steps. For investors, the condition for this company to become a better investment is simple: either the price returns to a level that offers enough margin of safety, or the company uses new share, margin and cash-flow data to prove it deserves today's valuation or a higher one. Conversely, if advanced-process share starts to loosen, CMP growth falls back toward the industry average, and depreciation rises at the same time, the original judgment has to be re-examined.

Bull and bear arguments

The bull case is not hard to write, and every piece of it lands on a specific fact. First, the company self-estimates very high share in FEOL CMP and in several silicon-wafer polishing materials, which says it holds an advantaged position at exactly the steps that need long validation and batch-to-batch stability. Second, CMP revenue grew 17.9% year over year in the fiscal year ended 2026, clearly faster than total revenue at 11.0%, showing the company is still capturing the expansion of advanced nodes and high-value steps. Third, five-year cumulative operating cash flow was about 1.11 times net income attributable to owners, so profit quality passes overall. Fourth, management's expansion is tightly focused on CMP and silicon-wafer materials rather than scattered diversification. Fifth, the company's medium-to-long-term target of 100 billion JPY of sales by the fiscal year ending 2030 means management is still allocating resources to near double-digit medium-term growth.

The bear case is equally concrete. First, share across all CMP steps is only 13%, with the high shares concentrated in niche positions rather than the whole category, so pricing power may be weaker than the world number one narrative intuitively suggests. Second, the fiscal year ended 2026 already showed a sharp rise in interest-bearing debt and equipment payables, and margin has to absorb depreciation pressure first over the next two years. Third, Chinese domestic substitution is moving from mature processes toward more advanced steps, and Anji Microelectronics' qualification and volume-production progress is no longer merely a concept. Fourth, the current price corresponds to roughly 30x trailing and about 26x forward earnings, which is not a thick margin of safety for new money. Fifth, in the fiscal year ended 2024 Fujimi already proved it is not a cycle-free materials company, and profit still contracts noticeably when demand falls back.

Pre-mortem

If this investment halves three years from now, one of the most likely scripts runs like this. Between 2027 and 2028, Chinese domestic CMP and related wet-electronic-chemical suppliers take more copper and copper-barrier, tungsten and ceria-related steps at mainland advanced memory and some logic customers, and Fujimi's incremental advanced-node orders in the Asian region slow. At the same time, depreciation from the new Japanese plant and the overseas line expansions lands fully in the accounts, operating margin falls back from 19% to 20% down to 15% to 16%, the market re-rates the stock from high-quality growth consumable to a good but ordinary cyclical material, the P/E compresses from 26x to 30x back to 18x to 20x, and the share price could fall into the 1,800 to 2,200 JPY range. That path is not exaggerated at all, because profit and valuation would fall together.

Another script, milder but just as dangerous, is that nothing seriously breaks in the business and growth simply is not fast enough. Suppose that around 2028 the CMP business is still growing but has come back from the mid-to-high teens to the high single digits; the silicon-wafer business steadies; general industrial and thermal spray fail to contribute a meaningful second curve; and earnings per share stall near 130 to 140 JPY. The market gradually finds this is more a good company maturing faster than imagined, and the valuation center falls back to around 20x. Even if profit does not decline, the share price can still deliver long-run low returns simply because it was bought too expensively. That script is especially common for high-quality companies.

Final research conclusion

Fujimi is a good company, and good in the genuine sense. It reached today's position by spending decades turning powder, purification and chemistry into a reliability asset customers cannot easily replace, and none of that has anything to do with buzzwords. Its strong position in front-end CMP and silicon-wafer polishing materials explains why the company can show higher margin, stronger medium-term growth and faster cycle repair than many ordinary materials businesses. The problem is the price: how much investors have to pay today for that quality. The company itself is fine. At the price at the end of July 2026, the market already fully acknowledges it as an advanced-process consumable asset rather than a traditional Japanese industrial stock. A new reason to buy therefore has to be stronger than the company is excellent.

At the current level I would rather treat Fujimi as a name worth tracking and worth waiting for than one to rush into. My concerns are mainly three. First, the real boundary of the high CMP share has been disclosed by the company itself, and pricing power may not be as strong as the market imagines. Second, returns on the expansion still need time to verify, and near-term depreciation pressure certainly exists. Third, the pace at which Chinese localized competition advances in advanced processes will be the most valuation-sensitive variable over the next two years. What would actually change my judgment is whether the company can use subsequent results to prove the following, rather than adding one more AI tailwind story: that margin stays steady after the expansion, that CMP still outruns the total business, and that advanced-node share has not loosened.

【Company-profile scores】

• Fundamental quality: high
• Growth: medium
• Moat: strong
• Financial soundness: medium
• Management credibility: high
• Valuation attractiveness: low
• Risk level: medium
• Suitable investor type: long-term growth

【Investment rating】

• Rating: Watch
• One-line thesis: Company quality and the FEOL and CMP niche barriers are both strong, but the current valuation demands delivery on growth and leaves a thin margin of safety.
• Three price signals:
  • 【Ideal buy price】2300–2550 JPY Basis: a conservative value of about 2,860 JPY from roughly 130 JPY of EPS for the fiscal year ending 2027 at a 22x P/E, then a margin of safety of about 10% to 20% on top.
  • Acceptable hold price: 3100–4200 JPY
  • Clearly overvalued price: 4950–5400 JPY
• Current-price classification: acceptable hold
• Whether to wait for a better price: yes. For new money the better entry is 2,300 to 2,550 JPY, or results that prove earnings per share can hold above 150 JPY beyond the fiscal year ending 2027 with advanced-process share intact. The opportunity cost of waiting is that if AI-related demand keeps beating expectations and the multiple does not compress, you could miss upside of roughly the low teens to a bit over 20%.
• Target holding horizon: 3–5 years
• Expected annualized return: conservative about -20%; base about +1%; optimistic about +24%
• Max-loss risk: if advanced-process share is eroded by Asian localized substitution in 2027 and 2028 while rising depreciation pushes operating margin back to the mid teens, the share price risks falling toward the 1,800 to 2,200 JPY range, about 40% to 50% below the current price.
• Reassessment-trigger signals: if CMP revenue growth falls below total revenue growth for two consecutive disclosure periods; if operating margin falls below 18% for two consecutive disclosure periods; if operating cash flow over net income falls below 0.8x on a rolling two-year basis; if Asia and Oceania revenue drops markedly and management cannot give a clear explanation; if Chinese advanced-process or advanced-packaging localization announcements increase markedly and start to map onto Fujimi's main battlefield.

【Valuation Range】

• current: 3680 (close as of 2026-07-28)
• bear (conservative · ideal buy zone): [2300, 2550]
• base (fair · acceptable hold zone): [3100, 4200]
• bull (optimistic · above the clearly-overvalued line): [4950, 5400]

Research uncertainties and main sources

Research uncertainties

First, on market share, what I can safely confirm from public primary materials is the company's self-estimated basis as of March 2024: FEOL 56%, all CMP steps 13%, and 59%, 84% and 92% in silicon-wafer-related subcategories. I did not find an independent third-party share table of equal credibility, equal granularity and public citability, so on the judgment of roughly 60% FEOL CMP I can only say the direction is close to secondhand accounts, while a strict anchor still has to rest on the company's own estimate.

Second, the company does not adequately disclose China revenue share or single-customer share in public materials, so quantifying China exposure and customer concentration can only be handled conservatively. That means some risk judgments can rest only on regional revenue structure and competitive progress in the industry, and cannot be turned into a precise sensitivity calculation.

Third, on qualification cycles, what I can find is indirect evidence reflected in industry trade sources and competitor disclosure rather than a standard cycle figure given by Fujimi itself. So the conclusion that qualification lock-in is strong holds, while exactly how many months it takes can only serve as auxiliary understanding rather than a precise fact.

Fourth, the current share price and market capitalization rest partly on a combination of Japanese market data services and the company's disclosed shares outstanding, rather than raw exchange terminal data. That is enough to support valuation analysis at the research level, but it is not a high-frequency trading basis.

Main sources

This report relies mainly on Fujimi's IR and disclosure documents, including the company's IR stock information page, the results briefing materials for the fiscal year ended March 2026, the earnings release for the fiscal year ended March 2026, corporate history and board materials, and the integrated report together with sales data by application and by region. External industry and demand verification draws mainly on SEMI for global silicon-wafer shipments, WSTS for the 2026 semiconductor market size forecast, Anji Microelectronics' 2025 annual report for CMP consumable value content and advanced-process qualification progress, and Entegris' public materials for horizontal comparison. The market price and valuation snapshot uses public pages from mainstream Japanese market data services and Reuters and Google Finance.

Other tickers mentioned

• ENTG.US: the closest global platform comparable in CMP consumables and contamination control
• 688019.SHG: a Chinese domestic challenger in CMP slurry and wet electronic chemicals, representing domestic-substitution pressure
• 4186.TSE: a Japanese advanced-process materials comparable, reflecting the valuation of high-quality Japanese electronic materials assets
• 4063.TSE: Japan's integrated semiconductor materials leader, used as a broad valuation anchor
• 2330.TW: a key end proxy for advanced logic and AI-related wafer demand
• 005930.KO: an important end proxy for advanced memory and HBM demand
• MU.US: another key end proxy for the global memory cycle and HBM demand

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.