Taiyo Yuden Co., Ltd. (TSE: 6976) - Deep Dive Research Report

Prepared 21 June 2026. Fiscal year ends 31 March; "FY2026" in this report means the year ended 31 March 2026 (reported 8 May 2026), and "FY2027" means the year ending 31 March 2027 (the current guidance year).

1. What the company does

Taiyo Yuden makes the tiny, unglamorous components that sit on almost every circuit board in the modern world. Its flagship product is the multilayer ceramic capacitor, or MLCC: a grain-of-sand-sized block of alternating ceramic-dielectric and metal-electrode layers that stores and releases electrical charge, smooths voltage, and filters noise. A modern smartphone contains roughly a thousand of them; a premium electric vehicle contains well over ten thousand; and a single AI-server rack now needs tens of thousands. Taiyo Yuden also makes inductors (coils that resist changes in current and shape power delivery) and high-frequency RF devices (filters that let a radio pick out one signal from the noise). None of these parts cost more than a few cents each. The business works because they ship in the tens of billions, because the highest-performance versions are extraordinarily hard to manufacture, and because a device simply does not function without them.

The company was founded on 23 March 1950 by Hikohachi Sato in Suginami, Tokyo, originally to make ceramic capacitors. In the 1960s it moved into ferrite materials and inductors and built a materials research institute in Gunma Prefecture, a decision that defined the company's identity: Taiyo Yuden describes itself as a firm that "creates products starting from the development of materials." It listed on the Tokyo Stock Exchange's First Section in January 1973. For a generation, the wider public knew it for something else entirely: in 1988 Taiyo Yuden invented the recordable CD (the "That's CD-R" brand), and for years it was a dominant name in optical recording media. That business was commoditised by China and then made obsolete by flash memory and streaming; the company exited optical media in 2015. What remains, and what matters today, is the passive-components business built on the same underlying skill - the chemistry of ceramic powders and the physics of stacking them into ever-thinner, ever-more-numerous layers.

The core value proposition is deceptively simple. An MLCC's capacitance, for a given physical size, is governed by how thin you can make each dielectric layer and how many of them you can stack. Taiyo Yuden's materials heritage lets it formulate ceramic powders and print electrode layers thin enough to pack more capacitance into smaller packages, or to deliver high capacitance in components rated to survive the heat and vibration of an engine bay or the brutal power-cycling of an AI accelerator. That is the whole game: smaller, higher-capacitance, more reliable, in volumes that never stop growing.

President Katsuya Sase frames the company's edge as living in the highest-difficulty corner of the catalogue - "the high value-added area which is our field of expertise" - rather than competing on the lowest-cost commodity parts where Chinese makers now win on price.

A concrete example of the product in action: inside an Nvidia-class AI server, the GPU draws enormous, rapidly fluctuating current. When the chip's load spikes in nanoseconds, the power-delivery network cannot react fast enough on its own, and the voltage would sag and crash the processor. Hundreds of MLCCs placed millimetres from the silicon act as a local charge reservoir, dumping current instantly to hold the voltage steady. As GPUs have grown more power-hungry, the number and quality of these capacitors per server has exploded - which is precisely why Taiyo Yuden, a company most people have never heard of, has become an AI-infrastructure beneficiary.

2. Business segments

Taiyo Yuden reports three operating segments. The revenue mix below is derived from FY2026 results (total net sales ¥355.3bn) and the FY2027 segment guidance disclosed at the 8 May 2026 results conference.

Capacitors (≈71% of sales, predominantly MLCC)

This is the company. Capacitors generated roughly ¥252bn in FY2026 and are guided to ¥282bn in FY2027 (+12%). Almost all of it is MLCC. The segment's core capability is materials science: Taiyo Yuden formulates its own barium-titanate-based dielectric ceramic powders and prints internal nickel electrode layers measured in sub-micron thicknesses, then co-fires the stack into a monolithic block. The difficulty scales non-linearly - each generation of higher capacitance-per-volume requires thinner, more uniform layers with zero defects across thousands of layers, because a single flaw shorts the part. This is process knowledge accumulated over decades and embodied in proprietary equipment; it is the reason only a handful of firms worldwide can make the top-end parts.

Within capacitors, the highest-value end markets are now AI servers (the fastest-growing slice), automotive (high-reliability, high-temperature parts), and IT infrastructure/industrial equipment. Combined, IT-infrastructure/industrial and automotive reached 56% of total company sales in FY2026, up year on year - management's explicit strategy of tilting the mix toward high-margin applications and away from price-eroding consumer commodity parts. Competitively the segment sits third globally behind Murata and Samsung Electro-Mechanics. It is simultaneously the margin engine, the cash cow, and now the growth bet - an unusual situation created by the AI-server demand surge.

Inductors (≈18% of sales)

Inductors generated roughly ¥64bn in FY2026, guided to ¥65bn in FY2027 (+1.1%, essentially flat). This segment traces directly to the 1960s ferrite-core heritage. Taiyo Yuden makes multilayer chip inductors and metal-alloy power inductors - the coils that regulate current in power-supply circuits and DC-DC converters. The core capability again rests on materials (ferrite and metal-alloy magnetic powders) plus fine-coil-winding and multilayer fabrication. End markets overlap heavily with capacitors: smartphones, automotive, and increasingly the power-delivery stages of AI servers, where metal power inductors handle the high currents feeding the GPU. The company has continued launching smaller, higher-performance metal power inductors (a new line was announced in early 2026). It is a respectable number-two-or-three global position in chip inductors, competing chiefly with Murata and TDK. Strategically it is a complementary growth lever that lets Taiyo Yuden sell into the same power-delivery design slots as its capacitors.

Other / Devices (≈11% of sales)

The "other" segment - roughly ¥39bn in FY2026, guided down to ¥37bn in FY2027 (-5.7%) - houses RF/communication devices (SAW and FBAR filters), modules, and miscellaneous components. This is the problem child and the restructuring story. SAW/FBAR filters go into smartphones to separate radio bands; demand, particularly from Chinese smartphone makers, fell, and the business was loss-making. Over FY2025-FY2026 management explicitly restructured the communication-device business - scaling capacity down to match demand and taking restructuring charges - to stop it dragging group profitability. Its competitive position here is weak relative to specialists like Murata, Qorvo, and Broadcom in RF filters, which is precisely why management chose to shrink rather than fight. Strategically this segment is being managed for cash and profitability, not growth; the declining guidance is deliberate.

3. Products and business detail

Multilayer ceramic capacitors (MLCC). The catalogue spans the full grid of case sizes (from the sub-millimetre 008004/0201 metric parts up to larger 3216/3225 packages), capacitance values, voltage ratings, and temperature grades. The technically demanding lines are: high-capacitance small-case parts (cramming microfarads of capacitance into a package the size of a sand grain, used near AI-server and smartphone processors); automotive-grade parts qualified to AEC-Q200 and rated for high temperature and high reliability; and embedded MLCCs designed to be buried inside the circuit-board substrate itself rather than soldered on top - a frontier product line that Niigata Taiyo Yuden has been ramping into mass production, including a world-first 1005-size 22µF embedded part. The hardest parts to make are the high-capacitance, small-case, high-reliability variants, because thinning the dielectric while maintaining defect-free uniformity across thousands of layers is where the process knowledge concentrates.

Inductors. Multilayer chip inductors, wire-wound inductors, and metal power inductors. The metal-alloy power inductors are the strategically important line - they handle the large currents in modern power-delivery circuits and are increasingly designed in alongside MLCCs in AI-server power stages. The company continues to push miniaturisation (smaller footprint at equal or higher performance), announcing new metal power inductors in early 2026.

RF / high-frequency devices. SAW (surface acoustic wave) and FBAR (film bulk acoustic resonator) filters and duplexers for smartphone radio front-ends, plus chip antennas and noise-suppression components. This is the restructured, shrinking line.

Other passives and modules. Varistors, NTC thermistors, piezoelectric components, and power-supply circuit modules round out the catalogue.

Manufacturing. Production is spread across Japan and Asia: Japan (including Gunma, Niigata, and other domestic sites for the highest-value and newest products), plus Singapore, South Korea, China, the Philippines, Taiwan, and Malaysia. The vertically integrated model - making its own ceramic powders and key process equipment in-house - is a structural feature, not an outsourced supply chain. Capacity utilisation ran above 90% through early 2026 as AI-server demand tightened supply. The company's stated mid-term plan (covering roughly 2026-2030) targets MLCC capacity growth of about 10% per year, with management signalling it may raise that to 15% per year if hyperscaler demand justifies it.

Geographies and milestones. Selling globally for decades, the company's notable milestones trace its evolution: world's first axial leaded ceramic capacitor (1976); world's first CD-R (1988); exit from optical media (2015); and, in the current era, the embedded-MLCC mass-production ramp and the high-capacitance AI-server parts that have driven order books past ¥100bn in a single quarter for the first time in about five years (the January-March 2026 quarter).

4. Customers

Taiyo Yuden does not sell to consumers; it sells to the companies that build electronic devices, plus a large distribution channel that resells to thousands of smaller manufacturers. The customer base breaks into four buckets: smartphone makers and their chipset suppliers; automotive (Tier-1 suppliers and OEMs); IT infrastructure (server and networking OEMs, and increasingly the AI-server supply chain feeding the hyperscalers); and broad industrial/consumer electronics via distributors.

Smartphones account for roughly 30-35% of revenue, with exposure skewed toward Apple (reached through the Broadcom and Qualcomm chipset supply chains rather than always directly), Samsung, and the Oppo/Vivo ecosystem, with comparatively less direct business with Chinese OEMs. The combined IT-infrastructure/industrial and automotive bucket reached 56% of sales in FY2026 and is the strategic centre of gravity.

The buying decision differs by customer type. For automotive and industrial, the decider is a design/quality engineer at a Tier-1 or OEM, and the criteria are reliability data, qualification (AEC-Q200 for auto), supply assurance, and a multi-quarter to multi-year design-in cycle. Once a part is qualified into an automotive platform, it tends to stay for the life of that platform - high switching cost, because requalifying a different supplier's part is expensive and risky. For AI servers and high-end computing, the decider is the power-integrity and board-design engineering team at the server ODM/OEM or accelerator vendor, and the criterion is whether the part can hold voltage under extreme transient loads; here the binding constraint right now is simply availability of qualified high-capacitance parts, which favours incumbents with capacity. For commodity consumer parts sold through distribution, the decision is more price-driven and switching costs are low - which is exactly the segment Taiyo Yuden is de-emphasising.

Why customers choose Taiyo Yuden: proven materials and process for the difficult high-capacitance and high-reliability parts, supply reliability, and a position as one of only a few global sources qualified for the top end. Concentration risk is moderate rather than acute - the smartphone exposure (and Apple within it) is the single biggest dependency, but the deliberate mix shift toward automotive and IT infrastructure is reducing it. Contract structure is largely a mix of design-win-driven recurring volume (sticky, multi-year, for auto and infrastructure) and shorter-cycle/distribution business; the book-to-bill ratio of 1.25 for all products (1.31 for capacitors) reported in early 2026 signals demand running ahead of shipments.

5. Competitive landscape

The MLCC industry is a tight oligopoly. The top five - Murata, Samsung Electro-Mechanics, Taiyo Yuden, TDK, and Yageo - control roughly 68-72% of global MLCC revenue. Taiyo Yuden sits a clear third by overall MLCC revenue, but the hierarchy shifts by sub-market: in the highest-end AI-server MLCCs, Murata and Samsung Electro-Mechanics together hold over 80% share, leaving Taiyo Yuden as a smaller but fast-growing participant trying to win incremental share as the supply shortage forces customers to qualify additional sources.

Against Murata, the market leader, Taiyo Yuden generally loses on scale and on the very highest-capacitance small-case parts where Murata sets the benchmark; it competes by being a credible second/third qualified source and by leaning on its materials skill in specific niches. Against Samsung Electro-Mechanics, the cost-competitive number two, Taiyo Yuden is at a scale and price disadvantage in volume parts. Against TDK and Kyocera, the competition is more balanced and product-mix dependent. The structural shift to watch is at the bottom: Chinese makers (led by Fenghua and others) have reached roughly 10% global MLCC share and are climbing the value chain, which steadily commoditises the low end and validates Taiyo Yuden's retreat from it.

Barriers to entry are genuinely high at the top end - the combination of proprietary ceramic-materials chemistry, sub-micron multilayer process control, in-house equipment, decades of automotive reliability data, and the capital intensity of capacity all protect the incumbents. But the barrier is much lower for commodity parts, where Chinese entrants are proving entry is possible with enough state-backed capital and time. Taiyo Yuden is strong where materials difficulty is the gate (high-capacitance, high-reliability, embedded) and exposed where the part is a commodity and price is the gate.

Market caps are approximate peer-size references as of June 2026 and move with the market; they are not precise and should be re-checked.

6. Industry

Demand for MLCCs and inductors is driven by the electronification of everything: more electronic content per device, and more devices. The historical workhorses were smartphones and consumer electronics; the growth drivers now are automotive (each EV uses many multiples of the MLCC content of an internal-combustion car, for power electronics, ADAS, and batteries) and, most dramatically, AI servers.

The AI-server dynamic is the single most important industry story for Taiyo Yuden. As GPU power draw and transient loads have escalated, MLCC content per server rack has surged. Trade-press figures attributed to SemiAnalysis estimates illustrate the slope: MLCC value per rack has risen from roughly $3,000 in the Nvidia H100 era to about $12,000 for GB200, with later platforms (Rubin-class) estimated at $22,000 and beyond, while piece-count per rack has climbed from around 15,000 to over 90,000. Whether or not the precise numbers hold, the direction is unambiguous and is pulling the whole high-end MLCC market into a simultaneous volume-and-price up-cycle - the AI server demand "strains capacity," producing the shortage conditions and price increases seen across the industry in 2025-2026.

The overall MLCC market is large (multi-billion-dollar, with the top five sharing ~70% of revenue) and growing, with the high-end/AI-server and automotive sub-segments growing far faster than the commodity base. Within the global supply chain, the Japanese and Korean makers occupy the high-reliability, high-capacitance tiers; Taiwan (Yageo) and China occupy the volume/standard tiers, with China substituting imports at the low end and climbing.

Regulation is light in the direct sense (no product-approval regime like pharma), but qualification standards function as a quasi-regulatory gate: automotive AEC-Q200 qualification and customer-specific reliability testing create the lock-in. Cyclicality is real and historically severe - the industry runs through inventory cycles tied to smartphones and consumer electronics, and Taiyo Yuden itself spent 2023-2024 working through a demand trough before the recent recovery. The current tailwinds are AI-server build-out and automotive content growth; the headwinds are the commodity-end price erosion from Chinese supply and the ever-present risk of a consumer-electronics downturn coinciding with a capacity-addition wave.

7. Growth triggers

Drawn from the company's earnings-release conferences across the six most recent reporting periods (quarterly, year ended March; the latest is the 8 May 2026 full-year FY2026 conference).

• AI-server MLCC demand ramping hard, with multi-year acceleration ahead. Management points to AI-server MLCC sales growing about 30% year on year in the current plan, with the slope expected to accelerate to roughly 80-85% growth in FY2027-FY2028 (FY2026 results conference, 8 May 2026; reiterated in mid-2026 commentary).

• Order book inflection - first quarter above ¥100bn in ~5 years. Orders in the January-March 2026 quarter exceeded ¥100bn for the first time in about five years, with a book-to-bill ratio of 1.25 for all products and 1.31 for capacitors (FY2026 results conference, 8 May 2026).

"The combined composition of sales for use in IT infrastructure/industrial equipment and automotive increased to 56%" - management framing the deliberate mix shift toward high value-added applications (FY2026 results conference, 8 May 2026).

• Capacity expansion stepping up. The mid-term plan (2026-2030) targets MLCC capacity growth of about 10% per year, which management has signalled could be raised to 15% per year depending on hyperscaler demand (mid-2026 commentary, President Sase).

• Embedded-MLCC mass production at Niigata. The world-first 1005-size high-capacitance embedded MLCC moving into mass production opens a new product category buried inside board substrates (FY2026 results conference, 8 May 2026; repeated theme).

• Selective price increases on mid/low-capacitance parts. Price increases of roughly 6-13% on mid-to-low-capacitance consumer and automotive MLCCs from May 2026, a shift after years of price declines (FY2026 results conference, 8 May 2026 / mid-2026 commentary).

• Margin tailwind from communication-device restructuring completing. The shrinking of the loss-making SAW/FBAR communication-device business is guided to keep lifting group profitability (FY2025 results conference, May 2025; FY2026 conferences).

• New metal power inductors for power delivery. Newly launched smaller, higher-performance metal power inductors target the same AI-server and automotive power-delivery design slots as the capacitors (early-2026 product announcement).

• FY2027 guidance steps up across the board. Guidance of net sales ¥384bn (+8.1%), operating profit ¥30bn (+50%), and net profit ¥18bn (+21.6%), with MLCC sales guided to ¥282bn (+12%) (FY2026 results conference, 8 May 2026).

8. Key risks

AI-server demand is real but concentrated and early-cycle. The most important upside driver is also a risk: AI-server MLCC demand depends on a small number of hyperscalers continuing to build out at the current pace. If hyperscaler capex slows, digests, or the AI accelerator roadmap stretches out, the 80-85% growth trajectory management is pointing to could compress quickly. Taiyo Yuden is a smaller share player in this exact sub-market (Murata and Samsung Electro-Mechanics hold 80%+), so it is a price-and-allocation taker rather than a setter, and could see its incremental volume squeezed if the leaders add capacity aggressively.

Capacity-addition timing. Management is raising capex to grow MLCC capacity 10-15% per year while competitors (Murata announced ¥80bn of incremental capex over two years) do the same. The classic passive-components failure mode is the whole industry adding capacity into a demand peak, then meeting a consumer-electronics downturn with a glut - which crushes utilisation, prices, and margins simultaneously. The industry has done this before; 2023-2024 was the most recent trough.

Commodity-end price erosion from China. Chinese makers have reached ~10% global share and keep climbing the value chain. Taiyo Yuden's defence is to retreat upward into high value-added parts, which works only as long as the difficult tiers stay difficult. If Chinese process capability advances faster than expected, the band of "protected" high-margin product narrows.

Smartphone and Apple concentration. Smartphones are still ~30-35% of revenue with meaningful Apple exposure through the chipset supply chain. A weak smartphone cycle, an Apple inventory correction, or design losses would hit a large slice of the base. The Chinese-smartphone weakness already forced the communication-device restructuring.

Management's own restructuring language - rebuilding the communication-device business, "scaling down to match demand conditions" and taking restructuring charges - is an admission that part of the portfolio was sized for demand that did not arrive. The risk is that the same can recur elsewhere if a demand forecast proves too optimistic.

Currency. Reported results benefited materially from yen depreciation (the yen weakened from ~¥149 to ~¥159 per US dollar over FY2026). A sharp yen reversal would reverse a chunk of the reported sales and profit tailwind even if unit volumes hold. This is a genuine, company-specific sensitivity given the export-heavy revenue base, not a generic caveat.

Cyclicality / inventory whip. Passive components amplify end-market swings because of channel inventory: when end demand turns, distributors and OEMs cut orders below consumption to destock, and book-to-bill overshoots downward. The current 1.25 book-to-bill is a peak-ish signal that mechanically cannot persist indefinitely.

9. Walk the talk

The six reporting periods used here are: Q3 FY2025 (Oct-Dec 2024, reported Feb 2025); full-year FY2025 (ended Mar 2025, reported May 2025); Q1 FY2026 (Apr-Jun 2025); Q2 FY2026 (Jul-Sep 2025, reported 6 Nov 2025); Q3 FY2026 (Oct-Dec 2025, reported Feb 2026); and full-year FY2026 (ended Mar 2026, reported 8 May 2026). The most recent is within 90 days of today.

Starting at the FY2025 results conference (May 2025), management guided two things: that operating income would rise sharply in FY2026 (around +50%+) driven by better MLCC utilisation, and that the communication-device business would be restructured to stop dragging profit. The set-up was a recovery story off a depressed base, and management framed it conservatively - third-party coverage at the time repeatedly characterised Taiyo Yuden's operating-income guidance as conservative relative to the MLCC sales momentum.

Through the FY2026 quarters, the recovery delivered, and then some. By Q2 FY2026 (reported 6 November 2025), net sales were up about 9% and operating profit up about 87%, well ahead of plan, driven by improved capacity utilisation and steadier selling prices, and management raised the full-year forecast upward. The restructuring of the communication-device business proceeded as promised, taking modest restructuring charges and removing a profit drag. By the time the full-year FY2026 results landed on 8 May 2026, the company reported operating profit up about 91% and net profit up more than fivefold, with the order book crossing ¥100bn in a quarter for the first time in roughly five years and book-to-bill at 1.25.

The pattern across these six periods is of management that guided a conservative recovery and then beat it as the AI-server demand wave arrived faster than its own plan assumed. The promise to restructure communication devices was kept and the promised margin benefit showed up. The promise of MLCC-driven operating-income recovery was not just kept but exceeded. The honest qualifier is that a meaningful part of the beat came from two factors outside management's control - the AI-server demand surge and the yen's depreciation from ~¥149 to ~¥159 - rather than purely from execution. That makes this look like consistently conservative guidance meeting a favourable external cycle, which is a credible-but-helped track record rather than pure operational outperformance.

Net read: this is conservative, do-what-they-say management whose recent results were flattered by a favourable demand-and-currency backdrop. The FY2027 +50% operating-profit guide is the next test of whether the beat continues or the cycle does the work.

10. Shareholder friendliness index

Dividends. Taiyo Yuden has paid a flat ¥90 per share annually (¥45 interim + ¥45 year-end) for FY2023, FY2024, FY2025, and FY2026, after raising it from ¥80 in FY2021-FY2022. The dividend was held flat right through the 2023-2024 earnings trough rather than cut, which is the notable point: in a year when net profit was a fraction of its prior level, the payout ratio rose well above the company's normal band (reported around 75-76%), meaning management chose to protect the dividend through the down-cycle rather than tie it mechanically to depressed earnings. As earnings recovered sharply in FY2026, the ratio normalises. (Source: company dividend disclosures; stockanalysis.com dividend history.)

Buybacks and dilution. Taiyo Yuden executed a share-repurchase programme during the FY2025 period - press coverage describes a buyback in the order of ~5% of shares and roughly ¥8-10bn returned - so the company does retire stock when cash flow allows, and management has indicated that if free cash flow comfortably exceeds the ~¥40bn level it would consider additional returns (higher dividend or accelerated buyback). Over the three-year window the share count has been broadly stable to modestly shrinking, with no signs of meaningful option-driven dilution. MoatMap's database recorded zero buyback transactions for 6976.T in the trailing ~90 days (since 23 March 2026); that is the last-90-day window only and is consistent with the FY2025 programme having completed earlier - it does not contradict the older repurchase activity described above, which comes from annual-report/press sources for the prior periods.

Verdict: Returns Capital (moderate). A dividend defended through the trough plus episodic buybacks marks this as genuinely shareholder-friendly, though the payout is steady rather than aggressively growing, and the bigger capital priority right now is reinvestment in MLCC capacity.

11. Insider activities

Source for recent transactions: MoatMap cross-market disclosure database (Japan 5%-rule large-shareholder reports / EDINET). Note: the last MoatMap scrape was 2026-06-19 23:00 UTC, ~41 hours before this report, so the most recent filings may be incomplete - flagged as stale.

Japan's disclosure regime for this company surfaces 5%-rule substantial-shareholder reports (大量保有報告書), not director/officer open-market dealing. Every transaction in the last 12 months is from a financial institution, not from company management - and the character of the filer matters enormously for reading the signal.

Reading the buys. None of these are conviction buys by company insiders. The large Nomura Securities entries swinging between roughly 18% and 22% of shares outstanding are explicitly held "as inventory for the securities business and for cumulative-investment operations" - market-making and dealing inventory, plus stock-lending. A securities firm's 5%-report holding crossing back and forth over the threshold reflects its trading book, not a view on Taiyo Yuden's fundamentals. The Mizuho entries are similarly flagged as dealing (short-term trading), derivatives-related, and temporary lending positions. These are mechanical, not informative.

Reading the sells. SMTAM (Sumitomo Mitsui Trust Asset Management) and M&G Investment Management are asset managers reducing fund positions - ordinary portfolio rebalancing, "reason not disclosed" beyond the standard investment-management purpose stated in the filing. There is no estate, founder, or insider-distribution story here.

Net assessment. There is no genuine management-insider signal in this window - not a single director or officer open-market purchase or sale appears. The entire dataset is institutional 5%-rule reporting dominated by a single securities dealer's inventory book (Nomura, 8 of 13 filings), which mechanically counts as both the biggest "buyer" and the biggest "seller." Stripping out the dealing noise, the only directional content is modest trimming by two long-only asset managers, which is a weak-to-neutral signal at most. Net read: neutral - no informative insider signal. Note also that this venue (Japan EDINET/5%-rule) does not routinely surface director/officer dealing the way US Form 4 does, so the absence of management buys is a feature of the disclosure regime, not evidence that management is avoiding the stock.

12. Scenarios

Bull case. The AI-server build-out proves to be a multi-year structural up-cycle rather than a spike. MLCC content per rack keeps climbing with each GPU generation, the high-end shortage persists, and Taiyo Yuden - having raised capacity growth toward 15% per year and qualified more of its high-capacitance parts into the hyperscaler supply chain - captures incremental share precisely because Murata and Samsung Electro-Mechanics cannot supply all the demand alone. The 80-85% AI-server MLCC growth management points to for FY2027-FY2028 materialises, automotive content growth runs alongside it, and the deliberate mix shift to high value-added parts holds margins even as commodity prices erode. The communication-device drag is gone, the embedded-MLCC line opens a genuinely new category, and the company moves from third-place laggard to a structurally more profitable, better-mixed business. Capital returns step up as free cash flow clears the ¥40bn bar.

Base case. Management delivers roughly what it guided. FY2027 lands somewhere near the ¥384bn sales / ¥30bn operating-profit plan, MLCC grows low-double-digits led by AI servers and automotive, inductors are roughly flat, and the "other" segment keeps shrinking by design. AI-server demand grows fast but Taiyo Yuden remains the number-three participant in that sub-market, taking solid incremental volume without overturning the Murata/Samsung duopoly. Capacity additions are absorbed by demand without a glut. The dividend stays around ¥90 with the occasional buyback. The business is meaningfully more profitable than its 2023-2024 trough, the mix is healthier, but it remains a cyclical components maker whose fortunes track the smartphone, auto, and now AI-server cycles.

Bear case. The AI-server demand wave turns out to be partly an inventory build. Hyperscalers digest, the accelerator roadmap stretches, and orders that crossed ¥100bn in a quarter reverse as book-to-bill falls back through 1.0. The whole industry - Taiyo Yuden, Murata, Samsung Electro-Mechanics - has been adding capacity into the peak, so utilisation drops just as a soft smartphone cycle and an automotive inventory correction arrive together, and prices fall across both the commodity and mid-tiers where Chinese supply is most aggressive. The yen strengthens, erasing a chunk of the recent reported tailwind. As the smaller, third-place high-end player, Taiyo Yuden gets squeezed on allocation and price by the two leaders, the protective band of "too-hard-for-China" product narrows, and the company finds itself back in a margin trough having spent heavily on capacity - the classic passive-components down-leg, only this time after a much larger capex commitment.