·Dashboard·Research·Work·Archive
/
← wiki
ticker stockpassives-mlcc updated 2026-06-01

6976 — Taiyo Yuden Co., Ltd.

Thesis

Verdict: WATCH / Bear-tilted Hold / Pass at current price (~¥8,287-8,398). Do not initiate at this level. Taiyo Yuden is the highest-beta listed pure-play on the MLCC cycle re-rate — the only major with ~71% MLCC revenue mix, giving the cleanest single-ticker MLCC beta available. The operational story is real: four consecutive quarterly earnings beats since May 2025, utilization rebuilt to 85-90%, capex discipline kicking in (cut 28% to ¥46B FY3/26E, below depreciation for the first time in 4 years), an industry-leading 6-13% May 2026 price hike, and credible AI-server design wins (world-first 1005-size 22µF Aug 2025, 0402-size 22µF Sep 2025).

But the stock has run +267% off the May 2025 low to ¥8,287 — that is +44% ABOVE the 16-analyst mean PT of ¥5,739 (median ¥5,600) — trading 44× forward EPS for a business guiding a 5.2% operating margin with a deteriorating balance sheet (net debt +¥106B in 3 years to ¥86B / 1.6× EBITDA). The market is already pricing a clean cycle handoff back toward FY3/22 peak earnings (EPS ¥433). At ¥8,287 and 44× forward P/E on FY3/26E EPS of ¥187, the stock is priced for the ENTIRE FY3/22-to-FY3/27 trough-to-peak earnings recovery to materialize within ~18 months.

What has to be true to win: the single most important thing is that the consensus PT upgrade cycle catches up to the stock within 6-12 months — either via 4-6 more quarters of beats forcing analysts to mark up FY3/27-FY3/28 EPS, or via the Aug 2026 FY3/27 full-year guide landing meaningfully above consensus. If PTs do not move, the stock retraces to the mean PT around ¥5,700.

Pure cycle conviction: high. Entry conviction at ¥8,287: low. This is a cyclical earnings-recovery / momentum trade, not a compounder thesis — the deep-dive's 5-year lock-up test fails (the right name for the cycle trade; the wrong name for a permanent compounder allocation). The easiest money was made by people who bought at ¥3,000-5,000 between May and Dec 2025; Pink missed that window. The remaining trade is the catch-up plus cycle-peak overshoot. For peer-relative MLCC exposure today, Murata (6981.T) offers a similar end-market thesis with less re-rate risk (~22× forward P/E vs 44×); Samsung Electro-Mechanics (009150.KS) offers cheaper entry. Conditional entry: initiate on a pullback to ¥6,000-6,500 (small), build below ¥5,500.

Snapshot

One-liner: World's #3 multilayer ceramic capacitor (MLCC) maker behind Murata and Samsung Electro-Mechanics (~11-13% global share), the most MLCC-pure of the majors at 71% of revenue — the highest-beta clean listed proxy for the AI-server / EV / smartphone MLCC cycle.

Ticker / exchange: 6976.T (Tokyo Stock Exchange Prime). Legal name Taiyo Yuden Co., Ltd. GICS Technology / Electronic Components. HQ Kyobashi East Building, Tokyo. Founded 23 March 1950 (Hikohachi Sato). ~21,000+ group employees (FY3/24). Reporting currency JPY. OTC ADRs: TYOYY / TYOYF.

Price / valuation snapshot (yfinance, mid-May 2026):

  • Price: ¥8,287 (May 20); ¥8,231 quoted in profile; ¥8,398 May 21 close
  • Market cap: ¥1.036T (~$6.9B at ¥150/USD); profile cites ¥1,028B
  • Enterprise value: ¥1.054T
  • Trailing P/E: 76.9× (76.4× profile) — distorted by FY3/25 EPS trough
  • Forward P/E (FY3/26E): 44.2× (43.9× profile); ~22-25× on consensus FY3/27 of ~¥187
  • EV/EBITDA TTM: 15.4×
  • P/B: 3.0×
  • Dividend yield: 1.14%
  • 52-week range: ¥2,257 - ¥8,399 (stock at top decile; +265-267% off the low)
  • Beta: 1.44
  • FCF yield (FY3/26E): ~0%

Analyst snapshot (16 analysts): Recommendation Buy (mean 2.44 of 5); breakdown Strong Buy 2 / Buy 5 / Hold 7 / Sell 1 / Strong Sell 1. Mean PT ¥5,739, median ¥5,600, high ¥10,000, low ¥2,900. Spot is +44.4% above mean and -17.1% below high — consensus is stale and unusually wide-distributed.

Business

What it makes: The tiny ceramic chips that store charge and filter noise inside every smartphone, EV, and AI server on earth. A multilayer ceramic capacitor is a stack of alternating ceramic dielectric layers and metal electrode layers co-fired into a chip 0.4-3.2mm long; capacitance scales with (dielectric constant × electrode area) ÷ dielectric thickness. The whole industry runs on making layers thinner without shorting and stacking more without cracking.

Segments (FY3/26E revenue split, IR Nov-2025, total ¥347.5B):

  • Capacitors (MLCC + aluminum electrolytic): 71.4% (¥248B) — core franchise, ~11-13% global MLCC share
  • Inductors (ferrite + metal): 17.4% (¥60.3B) — power inductors for phones, autos, servers
  • Integrated modules & devices (FBAR/SAW filters, circuit modules): 4.3% (¥14.8B) — Chinese-smartphone demand weak; modules -35.6% / -36% YoY
  • Others (energy devices, sensors, recording media legacy): 7.0% (¥24.4B)

End-market mix (2Q FY3/26 company estimate): Automotive (ADAS, ECU, meter cluster) 29%; IT infrastructure / industrial (servers, base stations, security camera) 23% — the AI-server tailwind; Communication equipment (smartphones) 22%; Information equipment (tablets, PCs, HDD/SSD) 17%; Consumer products (game console, smartwatch, earbuds) 9%. Medium-term plan targeted auto + IT-infra to 50% of sales by FY3/26 — already at 52% (29%+23%), ahead of schedule.

Customers (intentionally opaque): (1) Apple — "major" undisclosed % (est. 10-15%), iPhone/MacBook MLCC alongside Murata, Kyocera; (2) Chinese smartphone OEMs (Huawei, Xiaomi, Oppo, Vivo) — high-end weak, a cited FY3/26 drag; (3) Auto Tier-1s (Bosch, Denso, Continental, Aptiv) — growing, served from Changzhou EV-grade plant; (4) AI-server ODM/EMS chain (Foxconn, Quanta, Inventec) — small but surging; (5) game console / wearable OEMs (Sony, Nintendo). No single customer believed to exceed 15-20%; Apple is the largest single direct customer.

Geography / footprint: ~80%+ overseas revenue, concentrated in Asia ex-Japan. Major sites: Tamamura (Gunma) — flagship high-end MLCC, world-first 1005-size 22µF mass production Aug 2025; Niigata Taiyo Yuden (MLCC, ramping 2024-25); Akita (ferrite + MLCC); Changzhou, China (EV-grade MLCC, $200M initial 2019, phase 1 expanded to ~$600M, production 2023); Malaysia/Sarawak; Korea + Philippines (module/device assembly). All key subs 100% owned; no material JVs of the kind common at Murata/TDK.

Moat: Process IP concentrated in (a) in-house barium titanate powder formulation (doped for Class II X7R/X8R behavior), (b) nickel inner-electrode metallurgy (Ni replaced palladium in the 2000s — the shift that cratered margins for non-leaders), and (c) co-fire process control over 50+ years of learning curve. Plus automotive AEC-Q200 / ISO-TS qualifications (18-24 months to re-qualify a second source) and capital intensity (~¥46B/yr capex). Taiyo Yuden's specific edge is small-form-factor high-cap density — first to ship 1005-size (1.0×0.5mm) at 22µF in Aug 2025, when the industry had been stuck at 10µF in that size for years. Each AI server uses 20,000-30,000 MLCCs (vs ~10,000 standard server, ~1,000 phone, ~10,000-15,000 EV) because every GPU/accelerator voltage rail needs decoupling; AI-server MLCC demand scales with accelerator silicon area × power density × voltage-rail count, all three accelerating.

Business-quality 3-test (deep-dive): (1) 5-year lock-up — No (4-5yr MLCC cycle; buy-and-hold through next down-cycle gives back the gain; Murata passes this far better via diversification). (2) Unique economic engine — Yes, partially (small-form high-cap MLCC process IP, edge durable 5-10yr but not Murata-class). (3) Blank-check disruptor — No (barium titanate process IP + AEC-Q200 cycles mean even unlimited capital takes 5-7 years to enter high-end; Yageo's KEMET acquisition is the proof case — bought position, still trails 7 years later). Verdict: durable mid-tier oligopoly member with cycle-sensitive returns.

Porter's snapshot: Rivalry high (top-3 oligopoly but Yageo/Taiwanese chase mid-range); new entrants low (capital + qualification barriers); buyer power medium-high (Apple/auto Tier-1 scale, Apple second-sources); supplier power low-medium (barium titanate / nickel commoditized, nickel price bites margins); substitutes low for the core high-cap small-size function.

Financials

All figures ¥ billion unless noted. FY3/22 = year ended March 2022 (cyclical peak); FY3/26E = company forecast Nov-2025.

Income statement & margins:

Metric FY3/22 (peak) FY3/23 FY3/24 FY3/25 FY3/26E
Revenue 349.6 319.5 322.6 341.4 347.5
Revenue growth YoY — -8.6% +1.0% +5.9% +1.8%
Gross profit 125.0 87.4 65.5 71.6 n/d
Gross margin 35.7% 27.4% 20.3% 21.0% n/d
Operating profit 68.2 32.0 9.1 10.5 18.0
Operating margin 19.5% 10.0% 2.8% 3.1% 5.2%
Net income 54.4 23.2 8.3 2.3 9.0
EPS basic (¥) 433.5 186.3 66.8 18.7* ~72 (implied)

*FY3/25 EPS depressed by tax + non-operating items; "normalized" ~¥34. Forward EPS consensus per yfinance ¥187 — analysts model a sharper rebound than company guidance. (Note a yfinance discrepancy: trailing EPS shows ¥107.7 ttm, appears to annualize the rapid-recovery quarters, vs company-reported ¥18.7 basic — flagged in profile.)

Cash flow & balance sheet:

Metric FY3/22 FY3/23 FY3/24 FY3/25 FY3/26E
Operating cash flow 67.3 39.5 51.1 33.9 n/d
Capex (50.6) (60.4) (82.8) (63.5) (46.0)
Free cash flow 16.7 -20.9 -31.7 -29.6 est. ~flat
Net debt (19.6) net cash 11.7 46.1 86.2 ~
Net debt / EBITDA net cash 0.2× 0.9× 1.6× ~1.2×
ROE ~18% ~7% ~2.5% 0.7% ~3% (guide)

Balance-sheet detail (deep-dive): Cash ¥93B→¥87B→¥103B→¥78B FY3/22-25; total debt ¥73B→¥99B→¥149B→¥164B; equity ¥300B→¥318B→¥330B→¥319B. Debt added ¥91-106B over 3 years to fund the FY3/23-25 capex peak; equity declined slightly in FY3/25 (FX + earnings drag). At 1.6× net debt/EBITDA in a cyclical low-quality moment, leverage is uncomfortable but not dangerous — FY3/27 normalized EBITDA of ¥80-100B would drop it to ~1.0×. Another down-cycle before deleveraging completes would force a capital raise or dividend cut.

Second-derivative / earnings-momentum (8-quarter surprise series, yfinance) — the table that matters most:

Earnings date EPS est (¥) EPS actual (¥) Surprise %
Nov 7 2024 (Q2 FY3/25) 36.75 -22.01 -159.9%
Feb 7 2025 (Q3 FY3/25) 26.21 36.88 +40.7%
May 8 2025 (Q4 FY3/25) -23.09 -46.82 -102.8% (the trough)
Aug 5 2025 (Q1 FY3/26) 10.29 -7.02 -168.2% (last miss)
Nov 5 2025 (Q2 FY3/26) 40.19 51.39 +27.9%
Feb 6 2026 (Q3 FY3/26) 33.06 56.70 +71.5%
May 7 2026 (Q4 FY3/26) 12.00 17.42 +45.1%
Aug 5 2026 (Q1 FY3/27) 36.74 — (upcoming)

May 2025 was the absolute earnings trough. Since then 3 of the last 4 prints beat by 28-72%, magnitude increasing — the exact earnings-revision pattern that drives multi-quarter momentum trades. The Aug 5 2026 print is the critical confirmation event.

Quarterly revenue acceleration (profile + Nov 2025 IR): Q4 FY3/25 ~¥85B / ~1% OP margin (trough); Q1 FY3/26 ~¥85B / ~3.5%; Q2 FY3/26 ¥92.8B (+9% QoQ) / OP ¥5.9B / 6.4% (+87% QoQ on utilization); Q3 FY3/26 ~¥95B+ / ~7%; Q4 FY3/26 ~¥95B / ~7%. Second derivative positive and steady — revenue from low-single to double-digit; OP margin ~1% to ~7% in 4 quarters. The ¥18B OP company guide implies actuals overshoot to ~¥22-25B.

Core Four: organic revenue growth FY3/26E +1.8% on flat volume + price recovery, accelerating into FY3/27; margins 3.1% trough → 5.2% guide → consensus rebuild to 8-12% by FY3/28 vs 19.5% peak; capital intensity capex/revenue 24% peak (FY3/24) → 13% (FY3/26E), depreciation now exceeds capex; capital deployment funded capex with debt, no buybacks since FY3/22, dividend stable. ROIC fell from ~15-18% (FY3/22) to ~1-3% (FY3/25 trough); WACC ~6-7% — currently destroying value (ROIC < WACC), inflecting neutral FY3/27, value-creating by FY3/28 if the cycle holds. Gross margin 23.1% TTM / 23% per checklist (recovery from 21.0% trough, below 35.7% peak). Book-to-bill 0.89 company-wide in 2Q FY3/26, down from 1.04-1.07 a year earlier — the cycle's only soft data point.

Industry landscape

Global MLCC TAM ~$15B in 2024, ~7-9% CAGR through 2030; AI-server is the highest-growth sub-segment with content per server roughly 2-3× standard. Share: Murata (6981.T) ~33-35% (#1, diversified, MLCC ~50% of revenue); Samsung Electro-Mechanics (009150.KS) ~20-22% (#2, captive Samsung + merchant); Taiyo Yuden (6976.T) ~11-13% (#3, most MLCC-pure); Yageo (2327.TW) ~12-13% (KEMET tantalum + commodity); TDK (6762.T) ~5%; Walsin (2492.TW) / Holy Stone (6957.TW / 3026.TW) ~3-5% each (commodity tier). Industry runs a 4-5 year cycle; the 2025-2026 up-cycle is driven by a four-year capex underbuild colliding with AI-server BOM explosion + EV/smartphone bottoming + yen weakness. High-end (AI-server, auto-grade) MLCC structurally undersupplied through at least 2027; commodity tier more balanced (Yageo/Walsin can refill). Upstream bottleneck flagged: Sakai Chemical (4078.T) sub-micron barium titanate powder (4% of Taiyo Yuden's mcap, under-priced if AI MLCC accelerates); Toho Titanium (5727.T) qualified Ni powder a softer second flag.

See sector page: passives-mlcc

Management

Overall grade: B (institutional-grade Japan-standard). Competent execution-focused team, intact governance, clean recent credibility tape, no red flags identified — but no exceptional alignment either, and capital-allocation timing is the standout concern. The thesis stands or falls on the cycle and the entry point, not on the people.

Leadership: Katsuya Sase (President, CEO & Representative Director; age 61, born 1964; in seat since June 2023; career insider, prior EVP under Tosaka, placed to execute MTP 2025 not transform). Shoichi Tosaka (Chairperson & Director; prior President; healthy CEO-to-Chair transition, not a founder-Chair-controlling pattern; not Sato-family). Tomomitsu Fukuda (Senior Executive Operating Officer & Director, Chief of Management Planning HQ; age 61; de-facto CFO-equivalent, investor-facing, runs earnings calls/IR). Toshiyuki Watanabe (Executive Operating Officer & Director, Chief of Sales HQ; age 63; owns Apple / auto Tier-1 / AI-server EMS relationships). Shinji Masuyama (EVP & Director). Plus Shunji Murai, Mitsuo Takagi, Susumu Higuchi, Iwao Fujikawa (operating officers); Koshi Anai (GM of Accounting Dept, functional finance head, below board level). All career lifers; bench depth 8+ senior officers lowers key-person risk. No named CFO board title — normal for Japanese large-caps (function split Fukuda + Anai); flag, not a red flag.

Ownership / alignment: Insider 1.67% (yfinance) — low absolute (~¥17.3B / ~$115M across the insider class) but normal for Japan large-cap; no founder dynasty (Sato lineage historical, not operational). Institutional 61.8% — heavily indexed. Insider transactions L6M: zero buys / zero sells per yfinance — but this feed is unreliable for Japan (FIEA Art. 27-23 only triggers at 5%+; sub-threshold exec trades report via TDnet/yuho and do not feed Yahoo). Honest read: we don't know whether insiders transacted; confirming needs a TDnet search (not run). If genuinely zero, mild negative — management doesn't consider the stock cheap enough to add personally.

Compensation: CEO Sase est. ¥150-400M total (~$1-2.7M, roughly half US peers; vs Murata CEO Nakajima ~¥250M, Samsung E-M ~$5-7M, Vishay ~$4-6M). Mix ~60-70% fixed / 20-30% short-term / 10-20% long-term. SBC negligible (<0.1% of revenue). Long-term incentive uses restricted stock (3-yr vesting) + performance-linked units tied to ROIC, operating profit, and TSR vs TOPIX Electronic Components index — the sector-relative TSR hurdle is a meaningful alignment signal (forces beating the sector, not just riding beta). Soft green. Cannot verify hurdle calibration (stretch vs layup) or grant timing without the yuho; a substantive mid-2024 grant at the ¥2,000-3,500 trough would now be a 3-4× unrealized gain and a strong confidence signal — worth checking.

Capital allocation (C+): Bad timing the standout concern. The only meaningful buyback in 5 years was ¥5B at the FY3/22 cycle peak (~¥5,500) — approximately the worst possible timing (stock then fell to ¥2,257, -60%). Capex peaked at 25.7% of revenue in FY3/24 — the same year revenue contracted (classic over-build into a down-cycle); FY3/26E cut to 13.2% is the right correction. Did NOT buy back at the ¥2,257 FY3/25 trough when valuation was distressed — the honest counter is the net-debt build (+¥106B) likely left no headroom, making the FY3/22 peak buyback the original sin. Dividend held flat through the FY3/25 earnings collapse (¥8.1B→¥10.6B→¥11.2B→¥11.2B; ~¥65→¥85→¥90→¥90/share) — modest positive, no panic-cut. Zero equity dilution (share count flat 124.7M-125.4M, no ATM/shelf/convertibles). No material M&A in 5 years — a positive read vs Yageo's KEMET (2020) / Chilisin (2021) roll-up stress. Grade below Murata's B+, above Yageo's C-.

Guidance credibility (Moderate-to-high, ~75-85%): Erratic during the cycle inflection (3 of 4 trough prints missed catastrophically Nov 2024-Aug 2025), transitioning toward conservative/sandbagger in recovery (3 consecutive beats of 28-72% Nov 2025-May 2026). This is the standard cyclical forecasting-difficulty pattern, not a credibility problem. Operational commitments delivered: OP raised ¥16B→¥18B (actual tracking ¥22-25B); capex ¥46B on plan; auto+IT-infra mix hit 52% ahead of schedule; 1005-22µF launched on time. The one to watch: the May 2026 6-13% price-hike claim — if Yageo/Walsin don't follow within ~90 days it becomes a soft/unverifiable claim. Treat Aug 2026 FY3/27 guide as a moderate floor (likely sandbagged).

Governance (green): June 2024 transition to Audit & Supervisory Committee structure (TSE Prime-preferred, positive). Board ~12, ~42% independent: internal directors Sase/Tosaka/Fukuda/Watanabe/Masuyama; 3 outside independent (Masashi Hiraiwa, Seiichi Koike, Emiko Hamada — Hamada adds gender diversity per ISS/Glass Lewis pressure); internal audit & supervisory Kazuyuki Oshima, Toshimitsu Honda; outside audit & supervisory Hajime Yoshitake, Tomomi Fujita (independent). No dual-class, no poison pill. Cross-shareholding unwind in progress per TSE/METI initiative. AGM re-election ≥95%. No activist (net debt rather than cash makes it an unattractive target — also low activist pressure on capital returns). Litigation/enforcement: none disclosed.

Disclosure caveats (honest-incomplete): Yuho (annual securities report, EDINET) not pulled live; shell/cross-holding and related-party scan not run; base-rate probability of material self-dealing at a TSE Prime large-cap is low (~5-10%) but "no red flags" should not be read as "fully cleared." Outstanding sub-30-min follow-ups: CFO title confirmation, long-term-incentive grant date/price, TDnet insider-transaction search, cross-shareholding (mochiai) detail, related-party section. Would I trust Sase and the board with capital? Yes, conditionally — they will collect cycle earnings competently, won't transform the business, won't deploy buybacks aggressively at the next trough.

Catalysts & risks

Bull / tailwinds:

  • AI-server MLCC content — largest single driver; hyperscaler capex (AMZN/MSFT/META/GOOG) up ~50% YoY into 2026, content per server up 2-3× vs 2024 gen as Hopper→Blackwell→Rubin add memory channels and power rails; 20-30k MLCCs per AI server vs 10k standard; durable 3-5yr (Blackwell, Rubin, post-Rubin)
  • MLCC pricing-power return — oligopoly + supply discipline + the May 2026 6-13% hike Taiyo Yuden LED on commodity/mid-cap. Because pricing power normally returns to the high-end first and the commodity tail last, leading a commodity-tier hike is the canonical "cycle confirmed" signal (5-15% ASP through FY3/27)
  • Yen weakness — each ¥1 weaker = ¥1.2-1.5B OP tailwind (¥147-150/USD vs ¥130-140 prior)
  • EV electrification — 8-10× ICE MLCC content per vehicle; steady mid-teens auto MLCC volume growth
  • Capex/D&A normalization — ¥17B+ FCF swing FY3/25→FY3/27 to repair the balance sheet
  • Smartphone bottoming — iPhone 17 cycle + Chinese OEM mid-tier recovery

What sell-side is missing (deep-dive): the 0.89 book-to-bill (2Q FY3/26) was the bear handle, but (1) BB is a turn-on indicator not a level indicator — the monotone 0.7→0.8→0.85→0.89 trajectory matters more than the level vs 1.0; (2) mix matters — AI-server BB is almost certainly >1.0, the drag is consumer modules (FBAR/SAW) flagged for restructuring. Consensus (median ¥5,600) anchors to a 2022-2024 mid-cycle frame, not a 2027-2028 super-cycle frame; the upgrade arc has not run.

Near-term catalysts (0-12mo): (1) Aug 5 2026 Q1 FY3/27 print + full-year guide — the central event; a 5th consecutive beat + above-consensus guide forces the next PT upgrade round; (2) Murata / Samsung E-M concurrent prints — sector-wide BB and pricing read; (3) second-round MLCC price-hike announcement — would confirm pricing returned to the high-end tier; (4) iPhone 17 build-order data (Q3 2026). Medium-term (1-3yr): FY3/27-28 OP margin trajectory to 10-15%; next AI-server density milestones (47µF) at Tamamura tech days; buyback announcement once net debt drops below 1.0× EBITDA. Leading indicators: Murata monthly order book, Yageo monthly revenue, Mouser/Digi-Key MLCC lead times, hyperscaler capex revisions.

Bear / risks:

  • Stock priced for perfection — high at current price; +44% above mean PT; closable only via earnings beats or a pullback
  • Cycle relapses before consensus upgrades land — medium; macro-dependent
  • AI-capex revision in 2H 2026 — low near-term, medium 2027+; the largest risk to the thesis
  • Yen strengthens to ¥130-140/USD on Fed cuts / BoJ tightening — medium; 10-15% earnings hit, 15-25% stock reaction
  • MLCC price compression resumes (Yageo/Walsin commodity capacity refill) — medium-high in commodity tier, low in high-end; mix-dependent
  • Chinese high-end smartphone weakness — already materializing (modules -36% YoY absorbed)
  • Apple second-sourcing — medium ongoing; margin not unit
  • Murata gains share at high-end — medium ongoing
  • Geopolitical (Changzhou plant) — low-medium; diversified footprint
  • Net debt deteriorates further if cycle stalls — low if cycle holds

Dilution risk: low (green flag) — share count flat, no ATM/shelf/convertibles; investors take credit risk not dilution risk. Key-person risk: low — Sase ~3yr in seat, Tosaka chair continuity, deep bench.

Bear case / downside: any one of (a) hyperscaler AI-capex guide-down 2H 2026, (b) Murata BB <0.9 sustained, or (c) yen to ¥130/USD re-rates the stock to ¥4,500-5,500 (analyst median range). Downside target ¥4,500 = -46% from spot. Thesis breaker: two consecutive quarterly misses + negative AI-capex inflection + concurrent Murata/Samsung E-M revenue guide-downs — would invalidate the 2025-2027 cycle thesis; probability 15-20% over 18 months.

Valuation / DCF

Verdict: no margin of safety at ¥8,287-8,398; stock above every reasonable fair-value framework. Every multiple sits above its 5-year median AND above the peer average — by ~2× on most metrics.

Multiple Current 5-yr median Peer avg (Murata / Samsung E-M / Yageo / TDK)
P/E (trailing) 76.9× ~20× ~16×
Forward P/E (FY3/26E) 44.2× ~18× ~16×
EV/EBITDA 15.4× ~9× ~8×
P/B 3.0× ~1.8× ~2.0×
EV/Revenue 3.0× ~1.2× ~2.0×

Sector comps (deep-dive): Murata 6981.T ~22× P/E / ~10× EV-EBITDA / 14-16% OP margin / lower cycle beta; Samsung E-M 009150.KS ~10-12× / ~5-6× / ~6-8% / mid beta; Taiyo Yuden 6976.T 44× / 15.4× / 5.2%E (3.1% trough) / highest beta; Yageo 2327.TW ~16× / ~9× / ~12% / mid-high; TDK 6762.T ~16× / ~8× / ~12% / very low; Walsin/Holy Stone ~12-15× / 7-10× / 8-12% / high. Taiyo Yuden is the most expensive listed MLCC pure-play on every multiple — partly justified by the 71% MLCC mix (only clean large-cap MLCC beta in one ticker), partly a function of how far it has run on consensus catch-up, partly thin-float dynamics (¥1T mcap, 62% institutional and most of that passive).

Implied-expectations decomposition: at ¥8,287 / ¥8,398 and 44× forward P/E on FY3/26E EPS of ¥187, the multiple compresses to a cycle-appropriate 18-20× peak P/E only if EPS reaches ¥430-450 — i.e. the FY3/22 peak (¥433). The stock is priced for the full FY3/22-to-FY3/27 trough-to-peak earnings recovery to materialize within ~18 months. Not impossible (cycle + price hikes + AI mix can compound fast) but requires near-perfect execution and zero new macro shock.

Simple DCF cross-check (checklist): 3-year revenue CAGR 4% (conservative cycle recovery), terminal OP margin 12% (mid-cycle), 8% WACC, mid-cycle FCF ~¥40-50B/yr by FY3/28 → fair-value range ¥5,500-7,500. The stock trades above this range.

Price targets (16 analysts): mean ¥5,739 (-31% from spot), median ¥5,600, high ¥10,000 (+21% at high), low ¥2,900. Spot +44.4% above mean, -17.1% below high. Implied 12m return at mean PT -31%; at high PT +21%. Consensus is stale: the stock ran +267% off the low while the mean PT barely budged. Either PTs get marked up substantially (bull) or the stock retraces toward the mean (bear); the Mizuho Buy upgrade + 12.5% Nov earnings revision suggest the upgrade cycle is starting, not finished.

Scenarios: Bear ¥4,500-5,500 (cycle relapse / AI-capex revision / yen rally) = -46% to -34%. Base — retrace toward mean PT ~¥5,700 if PTs don't move. Bull — PT mark-ups + cycle-peak overshoot toward / past the ¥10,000 high PT on continued beats. Conviction: Low at ¥8,287; Medium-High if entered below ¥6,000.

Decision log

2026-05-20/21 — /profile, /deep-dive, /mgmt-dd, /checklist (peer-swarm, MLCC ideas vs Yageo 2327.TW, Walsin 2492.TW, Holy Stone 3026.TW, TDK 6762.T, Murata 6981.T, Samsung E-M 009150.KS):

  • Deep-dive verdict: Bear-tilted Hold / Pass at current price (¥8,287). Pure cycle conviction high; entry conviction low. Conditional entry: pullback to ¥6,000-6,500 (initiate ~50% target, high conviction); ¥5,000-5,500 (build to full, very high); new high >¥9,000 on an Aug 2026 beat — wait, don't chase. Suggest 2-4% weight at full conviction; pair with Murata for blended MLCC exposure. Trim triggers approach ¥9,500-10,000. Stop/re-eval: two consecutive misses, Murata/Samsung E-M concurrent revenue guide-down >5%, yen through ¥140/USD, AI-capex revision from any top-4 hyperscaler.

  • Checklist verdict: WATCH (do not buy at ¥8,398). Pre-buy scorecard 5 yes / 1 partial / 2 no / 1 n/a / 1 yes-but-flagged — the two firm "no" answers (valuation, technicals) are the disqualifiers. FundamentEdge hard rules: 1 FAIL (revenue-growth primacy — cyclical ~+1.8% grower, FY3/26E ¥347.5B basically flat vs FY3/22 ¥349.6B; cannot be valued like a compounder), 3 PASS (second derivative decisively positive; valuation-is-not-a-thesis applied; quality-can-be-a-risk applied), 1 PASS-with-caveat (estimate revisions positive but stock has outrun them). Behavioral audit: 3 of 6 traps active (FOMO, confirmation bias, recency bias) — pause warranted. Technicals (May 21 close ¥8,398): RSI 76.3 overbought, +57% above 50d MA (¥5,360), ~+100% above 200d MA (~¥4,200), at 52w high with open air above, 5-day volume +64% above 50-day avg — classic late-stage / blow-off; do not chase. Conditional buy: ¥6,500-6,700 pullback (50% of target), ¥5,500 (100%), Aug 5 2026 beat + above-consensus FY3/27 guide AND stock <¥9,000 (25% starter). Hard target ¥10,000-11,000 trim 50%; stop break of ¥6,500 (exit half) / ¥5,500 (full exit). Expected holding period 12-18 months. Target size 1.5-2.5% at full conviction; 0% at ¥8,398.

  • Mgmt-dd verdict: Overall grade B (institutional-grade Japan-standard); management not a reason to add conviction nor to subtract it. Capital-allocation timing the one real concern. Would trust the board with capital, conditionally.

  • Portfolio note: Pink already holds Yageo (2327.TW) per recent swarm research — adding 6976 stacks correlated MLCC exposure; either pair-trade (long 6976 / short Yageo on cross-sectional value) or size down. For the swarm orchestrator: Taiyo Yuden ranks behind Murata (better-priced MLCC pure-play) and likely behind Walsin (cautious participate at TWD 262); it is the highest-beta but worst-entry-priced of the swarm at today's prices.

  • SemiAnalysis cross-check (all four runs): no SA coverage of Taiyo Yuden / 6976 / MLCC in the local mirror; incidental mentions only (Intel 2022, advanced-packaging 2022, Samsung cultural-issues 2022). No contradiction to flag.

Open follow-ups: dedicated MLCC /primer after the peer swarm closes (currently absent from industry-log); Sakai Chemical (4078.T) upstream-bottleneck /profile; TDnet insider-filing confirmation; CFO identity / yuho FY3/25 read (CFO title, long-term-incentive grant date & price, cross-shareholding detail, related-party section).

Sources

Fragments folded in (consolidated 2026-05-30):

  • 6976-t-deep-dive.md (deep-dive, 2026-05-20)
  • 6976-profile.md (company profile, 2026-05-20)
  • 6976-t-checklist.md (pre-buy checklist, 2026-05-20/21)
  • 6976-t-mgmt-dd.md (management due diligence, 2026-05-20)

An empty *-filings.md placeholder in the folder was intentionally NOT merged.

Primary IR:

Third-party:

Data:

  • yfinance pulls May 20-21 2026 — live price, multiples, ownership, 8-quarter earnings-surprise series, 4-year financials, technicals (RSI, 50d/200d MA, volume)
  • SemiAnalysis mirror at ~/Dropbox/Wafflebun/KB/wiki/semianalysis/ — no MLCC-dedicated coverage; no contradiction

Adjacent primers (~/.claude/industry-log.md): AI Server PCB / HDI (2026-05-03), Cu Wiring / Advanced Packaging Interconnect (2026-05-12), Memory Sector (2026-04-26), SiP/OSAT (2026-04), Glass Substrate (2026-05).

Consolidation queue (merged 2026-05-30)

These four research fragments were folded into this canonical page and stay live pending Pink's archive confirm:

  • [ ] 6976-t-deep-dive.md
  • [ ] 6976-profile.md
  • [ ] 6976-t-checklist.md
  • [ ] 6976-t-mgmt-dd.md

Source updates (auto-maintained)

Drop/2. Frontend (Jun 1, 26) - GS Memory

Goldman Sachs does not cover Taiyo Yuden directly; the report covers memory (Samsung, Hynix, Kioxia) and argues AI server demand will remain in structural undersupply through 2028, with server DRAM share rising to ~58% by 2027 and HBM TAM reaching $168B in 2028.

Relevant to your thesis: The AI server capex durability GS documents (hyperscaler commitments, scaling cluster demand) directly supports the IT-infrastructure end-market (23% of Taiyo Yuden revenue, 20,000-30,000 MLCCs per AI server) that underpins the MLCC cycle bull case.

Source: dropfile://2. Frontend/Memory/GS Memory.pdf

Intake (May 23, 26) - mops-diligence-2026-05-23

Taiyo Yuden is mentioned as one of the global MLCC majors to which Prosperity Dielectrics (PDC) supplies barium titanate dielectric powder externally, alongside the PSA-group diligence context for Taiwan peers.

Relevant to your thesis: Confirms Taiyo Yuden sources barium titanate powder from external suppliers including PDC, consistent with the wiki's "barium titanate powder formulation" moat framing — though Taiyo Yuden's in-house formulation capability is the actual moat claim, not PDC dependency.

Source: intakefile://mops-diligence-2026-05-23.md

Intake (May 21, 26) - 2327.TW-filings

Yageo (2327.TW) filings show Q1 2026 revenue +23% YoY with gross margin at 38.1% and operating margin at 25.2% — substantially higher than Taiyo Yuden's guided 5.2% operating margin — and the Shibaura Electronics acquisition (world #1 NTC thermistor) closed Oct 2025 at 87.3% acceptance after a 54%-over-opening-bid bidding war.

Relevant to your thesis: Yageo's 25% operating margins vs Taiyo Yuden's 5.2% guidance reinforces the peer-relative valuation case that Yageo (or Murata) may offer better cycle exposure per unit of margin risk than 6976 at ¥8,287.

Source: intakefile://2327.TW-filings.md

Drop/Bottleneck (May 24, 26) - 东莞证券_MLCC行业深度报告：供需矛盾加剧，高阶MLCC价格有望上扬_260325 (1)

Dongguan Securities (March 2026) notes Taiyo Yuden's delivery times and prices are rising alongside Murata and Samsung Electro-Mechanics, with high-end MLCC supply-demand imbalance intensifying and Murata citing order inquiries at twice current capacity.

Relevant to your thesis: Confirms the May 2026 price hike thesis and the AI-server MLCC demand surge underpinning the bull case, while the full-capacity / limited new supply backdrop supports the cycle-re-rate narrative — though at ¥8,287 this is already priced in.

Source: dropfile://Bottleneck/MLCC/东莞证券_MLCC行业深度报告：供需矛盾加剧，高阶MLCC价格有望上扬_260325 (1).pdf

Drop/Bottleneck (May 21, 26) - global_passives_basket_comparison

The Bottleneck basket (21 May 2026) scores Taiyo Yuden 14/30 (Hard Pass), with Revision Velocity at 1/6 — estimates cut from ¥130 to ¥90 over the trailing 90 days while the stock rallied ~77% from the Feb 2026 print, and the consensus PT (~¥3,860) sitting far below market.

Relevant to your thesis: The negative revision velocity directly challenges the wiki's core bull condition — that analyst PT upgrades catch up to the stock — and the ¥3,860 consensus PT is materially below even the wiki's bear-case entry level of ¥5,700.

Source: dropfile://Bottleneck/MLCC/global_passives_basket_comparison.pdf

Drop/Bottleneck (May 24, 26) - J.P. Morgan-MLCC Industry：Growing likelihood of tight supply...

J.P. Morgan upgraded Taiyo Yuden from Neutral to Overweight (April 3, 2026), raising its price target by applying 2017–18 and 2020–21 cycle-peak multiples, on the thesis that MLCC utilization approaching 90% signals tightening supply/demand from mid-2026, with AI server MLCC demand (GB300 up ~50–60% vs GB200) preventing price cuts even to high-bargaining-power smartphone customers.

Relevant to your thesis: The JPM upgrade is the catalyst that could close the 44%-above-mean-PT gap — if analysts absorb the peak-multiple framework, consensus PTs should revise sharply upward toward the stock, validating the "catch-up PT cycle" as the single most important condition for bulls to win.

Source: dropfile://Bottleneck/MLCC/J.P. Morgan-MLCC Industry：Growing likelihood of tight supply demand； Murata Manufacturing and Taiyo Yuden up to Overweight-260403.pdf

Drop/Bottleneck (May 24, 26) - 2026 04 Capacitor Dossier

The 2026 Capacitor Dossier notes the Kyocera AVX achieving 47µF in 0402-size (1.0×0.5mm) MLCCs — a 2.1× density improvement — while Taiyo Yuden is cited among Japanese/Korean suppliers pursuing capacity expansion focused on high-margin automotive and AI segments amid China's commodity-pricing pressure.

Relevant to your thesis: The Kyocera AVX 0402/47µF data point directly challenges Taiyo Yuden's moat narrative — the wiki credits Taiyo Yuden as first to 0402/22µF (Sep 2025), but Kyocera AVX has now shipped 47µF in the same form factor, compressing that miniaturization lead faster than the thesis assumed.

Source: dropfile://Bottleneck/MLCC/2026 04 Capacitor Dossier.pdf

Pages that link here (23)
009150 — Samsung Electro-Mechanics2327 — Yageo Corporation2492 — Walsin Technology Corporation3026 — Holy Stone Enterprise Co., Ltd.3026 — Holy Stone Enterprise Co., Ltd.3436 — SUMCO Corporation4078 — Sakai Chemical Industry Co., Ltd.4092 — Nippon Chemical Industrial Co., Ltd.4092 — Nippon Chemical Industrial Co., Ltd.6173 — Prosperity Dielectrics Co., Ltd.6762 — TDK Corporation6762 — TDK Corporation6981 — Murata Manufacturing Co., Ltd.MLCC Peer Swarm — Best-Positioned Into the AI/EV/Smartphone Up-CycleMLCC Peer Swarm — Delta vs 2026-05-20Stock Showdown — Santec ([[6777/6777|6777]]) vs. JEM ([[6855/6855|6855]]) vs. Anritsu ([[6754/6754|6754]]) vs. [[6976/6976|Taiyo Yuden]] ([[6976/6976|6976]]) vs. Dexerials (4980)Passives & Components (MLCC)Passives & Components (MLCC)Semiconductor Capital Equipment & TestSemiconductors (cross-cutting / geopolitics)2026-05-23-mlcc-sector2026-05-25-mlcc-peer-swarmDongguan Securities: MLCC Industry In-Depth Report
Open in Obsidian