MLCC Sector Deep-Dive — Top 2 With Crowding in Mind

Date: 2026-05-23 (Friday post-close Asia) Universe: Murata 6981.T, TDK 6762.T, Taiyo Yuden 6976.T, Yageo 2327.TW, Walsin 2492.TW, Holy Stone 3026.TW, PDC 6173.TWO, Kingboard 0148.HK, Samsung Electro-Mech 009150.KS, Sakai Chemical 4078.T (upstream) Builds on: mlcc-peer-swarm-2026-05-20 (May 20 swarm Murata #1 / Walsin #2 by risk-adjusted entry math)

1. Verdict — Top 2

1. Murata Manufacturing (6981.T) — the only name where the AI MLCC thesis lines up structurally rather than narratively. Roughly 40% global MLCC share overall, with industry triangulations putting Murata's share of AI-server-grade MLCCs at 50-70% (Banyan Lane uses 60% as the working assumption). The April-2026 industry-leading price hike of 15-25% on AI/auto-grade parts, the ¥150B buyback announced May 2026 retiring ~4% of float, the +89.5% capacitor backlog surge, and the FY27 operating-profit guide of +34.8% on only +7% revenue together describe a margin-mix-driven earnings inflection that is happening now, not a forecast. Crowding score 3/5 in the basket framework — meaningful but the most moderate of the Japanese majors, with the deepest moat carrying the score (Moat 5/5).

2. Kingboard Holdings (0148.HK) — the crowding-aware second pick. Different sub-industry (copper-clad laminate, PCB substrate, chemicals), upstream of MLCC in the AI server bill of materials. Forward P/E ~9.9x against the rest of the basket trading at 26-87x. The basket framework scored Kingboard 19/30 Watch (Inflection 4/5; everything else mid-3 except Moat 3/5 because CCL is less moat-dense than MLCC). It is the only name in the entire universe where consensus mean price target (HK$64.0 from 2 analysts) sits above the current price (HK$58.30), and yfinance flags the consensus as strong_buy. Dividend yield 3.81% versus 0.79-1.5% for Japanese / Taiwanese MLCC peers. Collyer Bridge actively posting on Kingboard Laminates (subsidiary 1888.HK) in the Subscriber Chat as of 23 May — the highest-quality independent confirmation signal in the data set.

These two are deliberately differentiated: TSE Prime Japan vs HKEX, MLCC component manufacturer vs upstream PCB substrate, ~80x forward P/E premium vs ~10x, the basket's #1 score vs #2. They participate in the same AI infrastructure build through entirely different layers of the BoM, with non-overlapping ownership bases and non-overlapping retail-flow vectors.

Honorable mention — TDK (6762.T) as a Japan-MLCC pivot if the structural Murata thesis at ¥7,130 feels too extended: TDK at ¥3,370 still trades just below BofA's PT (¥3,500) and the 17-analyst consensus mean (¥2,869, though the live print is above that). BofA Buy maintained on stable battery + HDD-head AI cycle + auto MLCC mix. The catch is the May 20 swarm's diagnosis still holds — owning TDK for MLCC means owning ATL Apple batteries by accident, because MLCC is only ~10% of group revenue. The cleanest expression of an MLCC view is Murata, not TDK.

Hard-pass list at today's prices: Taiyo Yuden (BofA Underperform PT ¥3,800 vs spot ¥9,102 = 139% above target on cut estimates), Samsung Electro-Mech (basket 13/30 Hard Pass on 9× in 12 months, Crowding 1/5), Yageo / Walsin / Holy Stone / PDC (all limit-up +10% on 23 May with stock now 49-200%+ above sell-side PTs).

2. What Changed vs the May 20 Swarm

The May 20 swarm ranked Murata #1 (best-positioned anchor) and put Walsin #2 (best risk-adjusted on pullback at TWD 200-225). The new data does three things to that picture.

It strengthens Murata. The May 21 Murata-vs-Yageo and Banyan Lane theses both make the same diagnosis on the same evidence: the AI MLCC supply/demand setup is structurally different from a normal MLCC cycle because there are only four qualified suppliers for ultra-high-CV parts (Murata, TDK, Samsung Electro-Mechanics, Kyocera AVX), hyperscalers have locked 60-70% of available high-CV capacity, AEC-Q200 qualification is 12-18 months, and Chinese capacity additions cannot replicate the 220µF+ in 1206 size. None of that was new information on May 20, but the four independent sources arriving at the same conclusion in the same week (BofA, Sino-Pac sell-side desk, Banyan Lane Substack, the global passives basket scoring framework) is the convergence pattern that matters.

It weakens Walsin and the rest of Taiwan on crowding. The May 20 swarm flagged Walsin's pricing posture as a fast-follower not a price-setter, but rated it BUY-on-pullback. The May 21 basket gave Walsin Crowding 1/5 with the explicit note "trading at session high in a limit-up tape on the most recent print." After today's complex-wide +10% limit-up move in Taiwan (Yageo, Walsin, Holy Stone, PDC all limit-up), the original May 20 entry zones (Walsin TWD 200-225, Yageo NT$480-520) are now 30-40% below current prices. Either the cycle is moving so fast that any pullback is shallow, or the Taiwan complex is in late-stage retail-flow chase. Either way, the entry math on the May 20 zones is broken until a pullback materializes.

It introduces Kingboard. The May 20 swarm covered seven pure-passives names but did not include the upstream CCL/PCB layer. The basket PDF and Collyer Bridge's chat both put Kingboard in the field as the crowding-discounted way to play the same AI BoM ramp. At 9.88x forward P/E with a 3.81% yield and 2 analysts with mean PT HK$64, the math is structurally different from the Japanese / Taiwanese MLCC names.

It validates the "avoid" list. The May 20 swarm called Holy Stone parabolic-vs-quality (PASS at TWD 433.5) and PDC asymmetric-to-downside on June 11 earnings. Both ranked Hard Pass in the May 21 basket. Both went limit-up again on May 23 (Holy Stone TWD 476.5, PDC TWD 192.5). The PASS framing has held; if anything, the disconnect with sell-side coverage is wider now.

3. The 23 May Tape — Why Crowding Matters Today

Every passives name moved hard on 23 May (Friday post-close Asia):

Ticker	Name	Close	Day Chg	52w	Fwd P/E	Mean PT	Spot vs PT	Crowding (basket)
6981.T	Murata	¥7,130	+5.99%	2,019-7,226	30.1x	¥5,203 (17 analysts)	+37%	3/5
6762.T	TDK	¥3,370	+7.29%	1,487-3,383	31.5x	¥2,869 (17 analysts)	+17%	n/a
6976.T	Taiyo Yuden	¥9,102	+11.75%	2,257-9,325	48.6x	¥5,739 (16 analysts)	+59%	2/5
2327.TW	Yageo	NT$629	+9.97% (limit-up)	112-629	29.5x	NT$423 (12 analysts)	+49%	2/5
2492.TW	Walsin	TWD 292.5	+9.96% (limit-up)	76-292.5	16.8x	n/a (sparse cov.)	n/a	1/5
3026.TW	Holy Stone	TWD 476.5	+9.92% (limit-up)	75-476.5	37.0x	TWD 160 (1 analyst, stale)	+198%	1/5
6173.TWO	PDC	TWD 192.5	+10.00% (limit-up)	35-192.5	n/a	n/a	n/a	n/a
0148.HK	Kingboard	HK$58.30	+7.07%	21.75-58.80	9.88x	HK$64.00 (2 analysts) — strong_buy	-9% (still below PT)	3/5
009150.KS	Samsung E-M	KRW 1,340,000	+11.30%	116k-1,363k	49.8x	n/a	n/a	1/5
4078.T	Sakai Chemical	¥3,465	+2.97%	2,520-4,265	12.8x	n/a (no coverage)	n/a	not in basket

Sources: yfinance close 2026-05-23; basket scores from "Global MLCC & Passives Basket" 2026-05-21.

The pattern is hard to miss. Six of nine names hit the daily limit or near-limit; every Taiwan name is at the limit-up ceiling at a fresh 52-week high. The only two names not running parabolic are Kingboard (+7%, well below its absolute price ceiling, still under PT) and Sakai Chemical (+3%, the upstream BaTiO3 powder bottleneck flagged in the May 20 swarm). Of the names with meaningful analyst coverage, Kingboard is the only one where consensus PT is still above the current price.

That is the crowding lens in one table. The framework reading goes: stocks trading above their published sell-side PTs with positive earnings revisions are mid-cycle and durable; stocks trading above their PTs while estimates were cut (Taiyo Yuden) are the late-cycle retail-flow chase that the framework's Revision Velocity dimension was designed to surface.

4. What Each PDF Actually Says

BofA Component Signals (19 May) — METI data + Murata/TDK Buy, Taiyo Yuden Underperform

BofA's Masashi Kubota reads the March 2026 METI Japan MLCC data as front-loaded ordering distortion. Production value -9% YoY but +14% MoM; volume +18% YoY but ASP -23% YoY. The volume / ASP divergence is the signature of customers pulling commodity orders forward to hedge anticipated price hikes and material shortages — which is bullish for the cycle's next quarter and bearish for the current quarter's ASP optics. BB ratio for major MLCC makers exceeded 1.3 in Jan-Mar 2026. BofA expects production value April-June onward to improve as the front-load distortion clears and high-end product (AI server, high-voltage MLCC) takes a larger mix share.

The ratings:

• Murata (6981.T) — Buy maintained, PO ¥6,200 (vs spot ¥7,130 today, so spot is now 15% above BofA's own PT). 30x FY3/28 EPS, ~10% above 10-year historical upper range. BofA FY3/28 numbers are above CoE — revenue ¥2,007B (vs CoE for FY3/27 ¥1,960B), OP ¥416B, NP ¥330B, EPS ¥168 — and FY3/29 BofA E: revenue ¥2,510B, OP ¥609B, EPS ¥242.
• TDK (6762.T) — Buy maintained, PO ¥3,500 (vs spot ¥3,370). 25x FY3/28 EPS, ~10% above 22x historical upper range. FY3/29 EPS ¥154 → PO ¥3,500 implies 22.6x trailing. Drivers: stable ATL battery performance + expansion in HDD-related components + MLCC and inductor demand for AI servers from 2H onward.
• Taiyo Yuden (6976.T) — Underperform maintained, PO raised from ¥3,250 to ¥3,800 (vs spot ¥9,102, so PO is 58% below spot). 18x FY3/28 EPS, in line with 10-year average. BofA verbatim: "we believe excessive expectations for MLCC price increases are reflected in the share price, as domestic MLCC makers deny price hikes based solely on supply/demand, and Taiyo Yuden's MLCC operating profit margin remains in the single digits owing to weak productivity ... We see the shares as highly overvalued at current price."

The dispersion within Japan tells you what the analyst desk thinks of the cycle composition. Buy ratings on the two names with structural moat + diversified end-market mix; Underperform on the most MLCC-pure but margin-impaired name. That is the same conclusion the basket framework reaches independently.

Murata vs Yageo (21 May)

This is the cleanest single-piece treatment of why Murata's moat and Yageo's growth are different things. The bifurcation framing is the central claim: high-end strategic MLCCs (high-voltage, polymer electrolytic, DC-link film, ultra-high-CV for AI rails) are 35-40% of market value but capture 60-70% of industry profit, growing 12-17% CAGR, with 24-47 week lead times and 30-80% price premiums. Commodity MLCCs are everything else, -10-20% pricing pressure from Chinese capacity, 12-16 week lead times. Murata is mix-shifting hard into the strategic tier, which is why FY27 OP guide is +34.8% on only +7.1% revenue.

The Yageo MLCC AI story is reframed downward. Yageo's AI rack MLCC exposure, even at a 15-20% share of the NVL72 MLCC BoM, lands at ~$160-290M annually — 3-6% of total Yageo annual revenue. Yageo's record Q1 2026 was driven primarily by the end of the 2-year passive inventory correction, returning strategic pricing power broadly, the tantalum business (which legitimately is AI-driven at >30% of tantalum revenue from AI servers), and general electronics recovery — not the AI GPU rack MLCC BoM directly. The note's verdict: "Yageo's AI story is more accurately a tantalum story than an MLCC story."

The Morgan Stanley Rubin BoM table is worth quoting: NVL72 MLCC content per rack rises from $1,530 (GB300, Blackwell) to $4,320 (VR200, Rubin), +182%. Total rack BoM rises $4.0M → $7.8M, +95%. MLCC is the third-fastest-growing line in the BoM behind memory (+435%) and PCB (+233%). MLCC content nearly triples generation-on-generation despite being only ~0.05% of total rack BoM — pricing power matters because hyperscalers will not lose AI training share over a fraction of a percent of capex.

Global MLCC & Passives Basket (21 May)

The framework explicitly carves out Crowding as a separate dimension from Valuation and Moat. The six dimensions are Moat, Crowding, Valuation, Inflection, Paid Now, Revision Velocity, each scored 1-5 to a 30-point maximum. The basket scores:

Rank	Ticker	Score	Verdict	Moat	Crowd	Val	Infl	Paid	Rev
1	6981.T Murata	21	Buy	5	3	3	4	3	3
2	0148.HK Kingboard	19	Watch	3	3	3	4	3	3
3	2327.TW Yageo	17	Pass	4	2	2	4	3	2
4	2492.TW Walsin	16	Pass	3	1	2	4	3	3
5	6976.T Taiyo Yuden	14	Hard Pass	3	2	2	3	3	1
6	009150.KS Samsung E-M	13	Hard Pass	4	1	1	3	2	2
7	3026.TW Holy Stone	12	Hard Pass	2	1	1	4	2	2

The scoring spread on Crowding alone is the cleanest read on positioning risk. Murata and Kingboard sit at 3/5 — meaningful coverage but not extreme. Yageo and Taiyo Yuden at 2/5 reflect retail-flow saturation and the stock-vs-revisions divergence (Yageo +178% in 12 months on 3-7% earnings revisions; Taiyo Yuden +77% from February while consensus EPS was cut from ¥130 to ¥90). Walsin / Samsung E-M / Holy Stone at 1/5 reflect limit-up tape on the most recent session, 4-9× moves in 12 months, and 4× analyst target dispersion (Samsung E-M Mirae KRW 250,000 vs KB Securities KRW 1,050,000).

The PEG table is worth flagging because Walsin, Holy Stone, PDC, Samsung E-M all carry PEG <1 — but the framework's own note is explicit that "a PEG below 1 in this basket often reflects cycle bounce rather than structural value." PEG is the wrong screen here precisely because the cycle is what most of these names are.

Goldman Sachs Yageo (16 April, Buy NT$346)

GS raised 2026/27/28E earnings 3%/7%/6% post the strong Q1 print, taking the 12-month PT from NT$302 to NT$346 on a 15x average 3Q26-2Q27E PB/ROE (2 STDV above the past 10-year industry average of 11.5x). New EPS NT$19.12 / NT$25.27 / NT$27.87. The framing of the upgrade is that Yageo is the supplier with the broadest product portfolio (MLCC + chip-R + tantalum + magnetics) and AI exposure is real but more cleanly seen in tantalum than in MLCC — same conclusion as the May 21 Murata-vs-Yageo piece. Channel mix from the deck: 1Q26 Distributors 45% / Direct Sales 35% / EMS 20%; product mix Magnetics 25% / Tantalum 24% / MLCC 19% / Resistor 14% / Sensor 13% / Others 5%; application Industrial 29% / Computing & ESs 25% / Automotive 17% / Consumer 13% / Telecom 12% / Aero-Def-Med 4%.

The PT history is the crowding read in chart form. GS PTs from 2023-2024 were NT$625-870 on FY24 forecasts; at the cycle trough the PT collapsed to NT$265 (Oct 2025); the April 2026 upgrade to NT$346 is the recovery, but it sits 45% below the current limit-up close of NT$629. GS sees Yageo as fair to fully valued; the market is paying for a more aggressive cycle than the sell-side will write down.

5. Substack Triangulation

Banyan Lane Capital (6 May) — "The Most Boring Component in AI Is About to Have Its Moment"

A 30-minute Murata long thesis at the time of writing ¥5,138 (ADR MRAAY $15.41). The frame is a three-axis setup (cyclical + secular + idiosyncratic), and Murata is one of the rare names that scores on all three. Cyclical: MLCC industry in early phase of multi-year up-cycle with lead times 24-40 weeks. Secular: AI MLCC content per rack (Blackwell ~441,000 MLCCs, Rubin estimated 600,000+) creating a new product category. Idiosyncratic: Murata-specific 60% AI MLCC share (vs 40% overall), May 2026 ¥150B buyback retiring 4% of float, conservative guidance culture that sets up serial beat-and-raise.

The model output is FY28E EPS ¥307 vs consensus ~¥210 (40-90% above consensus). Base case 22x P/E gets ADR $21.77 / Tokyo ¥7,254 — which Murata is now trading approximately at (¥7,130) after the +39% move from the date of publication. Bull case 28x gets $27.71 / ¥9,235. Asymmetric 32x gets $31.66 / ¥10,500.

The piece quotes Murata president Norio Nakajima saying customer inquiries are running at roughly 2× available supply capacity (Bloomberg, early 2026). The supply-side detail is granular: total AI-capable MLCC industry capacity ~80-110B units/year, of which 50-60% is contracted to non-AI; available for AI ~30-50B units; 2026 AI MLCC demand already 130B+ and rising. No near-term supply response possible. Murata's Izumo plant started shipping April 2026 (25 months from announce). SEMCO Korean expansion is mid-2027.

Cross-reference with the Murata-vs-Yageo PDF's hyperscaler lock-in claim (60-70% of high-CV capacity locked) and the BofA backlog surge data (+89.5%, computers/servers application +28.4%): three independent sources reading the same constraint.

The thesis also notes a substitution risk in the long term — embedded deep-trench capacitors (eDTC) and HD-MiM on-chip integration could eventually displace AI MLCC content. Banyan Lane flags this as a real long-term threat that is not a near-term one. It is the right risk to track.

Korea Invest Insights (14, 17, 22 May) — Samsung Electro-Mechanics

Three pieces explaining the SEMCO 90-day double from KRW ~400k to KRW 1,340,000. The May 17 piece's framing: this is not a generic AI-tailwind rally — FC-BGA entered Nvidia's Vera/Rubin platform and Q1 2026 crossed KRW 3T in revenue for the first time with OP +40% YoY and Package Solutions (FC-BGA) revenue +45% YoY. The May 22 piece focuses on the AI Package Power Bottleneck and a ₩1.5 trillion contract.

This is useful context for understanding why Samsung E-M ran 9× in 12 months despite the basket's 13/30 Hard Pass score: the company has real AI substrate exposure beyond MLCC. But the basket's read on positioning is the relevant lens — Crowding 1/5, Valuation 1/5, NTM P/E 82x, analyst target dispersion 4× (KRW 250,000 to KRW 1,050,000). The stock is priced for the bull case to land perfectly. Real-business is not the same as risk-adjusted-investable.

Collyer Bridge Subscriber Chat (21-23 May)

Active threads posted by @collyer this week:

• "MLCCs" (5 replies, posted 21 May)
• "MS Rubin note" (21 replies, posted 21 May) — Morgan Stanley NVL72 BoM context
• "CTR Holdings" (8 replies, posted 22 May) — flagged for follow-up; HK-listed PCB-adjacent
• "Rigaku" (27 replies, posted 22 May)
• "Kingboard Laminates" (6 replies, posted 23 May 5:19 PM)

The signal density on Kingboard / MLCC / Rubin BoM in the Subscriber Chat over a single week is the qualitative confirmation that what is on the page of the basket PDF is also what the analyst Pink trusts is actively discussing with paying subscribers. Collyer Bridge is APAC-focused, sector-specialist, and tends to surface ideas before sell-side ratings catch up (the March 2026 "hottest idea in Taiwan" post is the pattern). That he is posting Kingboard Laminates (1888.HK, the publicly-listed CCL subsidiary of Kingboard Holdings 0148.HK) the day Pink picked up the file is the strongest independent corroboration of the basket PDF's #2 ranking.

(Note: the Subscriber Chat thread bodies were not fully extracted by opencli — the platform's React-virtualized panel does not yield well to the eval pattern that works for posts. The thread titles + timestamps + reply counts visible in the chat inbox sidebar were extracted; thread contents were not. Flagging because the chat is a primary corroboration source and Pink may want to read the actual messages in-browser.)

6. Crowding Teardown — Per Name

Murata (6981.T) — Moat is the deepest in the basket. ~40% overall MLCC share; 50-70% high-end AI MLCC share. Crowding moderate (3/5) — broad sell-side coverage (17 analysts) but consensus mean PT ¥5,203 is now 27% below spot ¥7,130, so the price is ahead of the published sell-side view. The Banyan Lane Substack thesis priced in fast (¥5,138 on 6 May → ¥7,130 today, +39% in 17 days). Insider activity, options skew, and short interest data not pulled in this report — flagging as the gap to close before sizing aggressively. The right framing is: structural thesis intact, valuation gap to sell-side targets has closed, the cleanest entry is on the next pullback rather than today's print.

TDK (6762.T) — Buy-rated by BofA at ¥3,500, mean consensus ¥2,869. Spot ¥3,370 is just below BofA PT and ~17% above mean consensus. Crowding lower than Murata in absolute price-vs-PT terms because the published Buy PT is still above spot. The catch on TDK is what the May 20 swarm flagged: MLCC is only ~10% of group revenue, ATL battery is 55% of revenue / 89% of segment OP. Owning TDK for MLCC means owning Apple-battery and HDD-head cycles by default. ISS QualityScore 1 across all four governance pillars (rare for any Japanese mega-cap). FY3/27 BofA E OP +10.8%, FY3/28 +12.6%, FY3/29 +12.5% — steady but not the structural inflection Murata is offering.

Taiyo Yuden (6976.T) — Crowding 2/5 in the basket; Revision Velocity 1/5 (estimates cut from ¥130 to ¥90 over trailing 90 days while stock rallied 77% from February). BofA Underperform with PT ¥3,800. Spot ¥9,102 is 139% above PT. The basket framework calls this "the rubric's worst Revision Velocity scenario." Of every name here this is the cleanest short candidate on technical disconnect with fundamentals; not a long.

Yageo (2327.TW) — Crowding 2/5, Revision Velocity 2/5. The fundamentals are genuinely strong (Q1 2026 EPS +44.4% YoY, 25.2% OPM new record, AI 14-15% of revenue with tantalum being the legitimate AI driver). The problem is purely positioning: stock NT$629 vs GS PT NT$346 = 82% above the most recent sell-side target. Yageo +178% in 12 months on +44% Q1 EPS and 3-7% sell-side EPS revisions. The framework is screaming that the multiple expansion ran ahead of the earnings revision. The Murata-vs-Yageo PDF reframes the AI exposure (tantalum, not MLCC), which suggests the multiple should compress as the market realizes the cleaner moat is upstream at Murata.

Walsin (2492.TW) — Limit-up at TWD 292.5. Sparse sell-side coverage; the May 21 basket noted "consensus PT TWD 600 was computed against an earlier reference price of TWD 147, implying the published targets have not been refreshed for the recent rally." Crowding 1/5. The May 20 swarm's "best risk-adjusted on pullback" framing is correct in principle but is now waiting on a 25-35% pullback to TWD 200-225 to be a clean entry. With Taiwan limit-ups today, the pullback is not on the immediate calendar.

Holy Stone (3026.TW) — Limit-up at TWD 476.5. One analyst, stale PT TWD 160 (198% below spot). Tang-family control, mediocre underlying business per the May 20 swarm. The cycle exposure is real but priced for perfection at 65x trailing P/E on a mid-tier distributor-hybrid. Hard Pass at any reasonable cycle exit valuation.

PDC 6173.TWO — Limit-up at TWD 192.5. Real upstream-of-MLCC asset (top-2 ceramic dielectric powder maker globally, oligopoly with Sakai and Nippon Chemical) bolted onto a sub-scale finished MLCC business. June 11 Q1 2026 earnings is the catalyst. The May 20 swarm called it asymmetric to downside given 53x trailing P/E; the +10% limit-up today doesn't change that read. Worth a hard look post-print if it pulls back.

Kingboard (0148.HK) — Spot HK$58.30 vs mean PT HK$64.00 (2 analysts, both strong_buy). Forward P/E 9.88x. Dividend yield 3.81%. 52-week range HK$21.75-58.80 — moved +168% from the 52-week low, +7% today, still ~1% below the 52-week high. The basket flags structural risks honestly: HK/China commercial property segment write-down risk, Cheung-family-control HoldCo discount has persisted for decades, mid-spec CCL competition from Chinese capacity, thin English-language sell-side coverage. The mid-spec CCL competition is the live risk; high-spec CCL for AI server PCBs is the up-cycle. Four announced CCL price hikes through April 2026 and Resonac / Mitsubishi Gas Chemical signaling 30%+ on bonding sheets and copper foil resin sheets confirm the upstream margin support. The HoldCo discount is the price you pay for the family-control structure; the dividend is the offset.

Samsung Electro-Mechanics (009150.KS) — KRW 1,340,000, +11% today. FC-BGA business is real and AI-leveraged. But Crowding 1/5, Valuation 1/5, 4× analyst PT dispersion, 9× in 12 months. The basket and Korea Invest Insights both make the same point at different volumes: real business, broken entry math.

Sakai Chemical (4078.T) — ¥3,465, +3% today (the only name in the universe not running parabolic). Upstream BaTiO3 dielectric powder, no analyst coverage, no consensus PT to disagree with. PEG not available; trailing P/E ~22x, forward P/E ~13x, dividend yield 4.62%, beta 0.346 (genuinely defensive). The May 20 swarm flagged this as "the cleanest non-consensus way to play the MLCC cycle." It is — but it is also genuinely under-researched and needs a dedicated deep-dive before sizing. Flagging for follow-up. Not in the top-2 because the analytical basis isn't there yet.

7. The Top-2 Argument — Why Murata + Kingboard

The argument for Murata is the convergence:

• BofA Buy PO ¥6,200 (May 19)
• The Sino-Pac / sell-side Murata-vs-Yageo desk note (May 21) concluding Murata's moat is specific and defensible
• The Global Passives Basket framework (May 21) putting Murata at 21/30 Buy
• Banyan Lane Capital Substack thesis (May 6) modeling FY28E EPS ¥307
• The May 20 peer swarm ranking Murata #1
• The +89.5% capacitor backlog surge and customer inquiries running 2× supply per Murata's CEO
• Hyperscaler lock-in of 60-70% of available high-CV capacity
• 4-supplier oligopoly with 12-18 month AEC-Q200 qualification timelines

Five independent sources reaching the same conclusion through different analytical paths is unusual. The crowding read on Murata is moderate-not-extreme: it is the largest-coverage name in the basket (17 analysts) and has rallied +39% since the Banyan Lane thesis, putting it 27-37% above mean sell-side PTs. But the basket's Crowding 3/5 reflects that this is a Japanese mega-cap with broad institutional ownership rather than a retail-flow vehicle. The crowding shows up as "fair to slightly stretched on consensus," not as "limit-up on retail momentum."

The argument for Kingboard is differentiation + valuation + signal:

• Different sub-industry from MLCC (CCL/PCB/chemicals); participates in the same AI BoM ramp through PCB substrate price hikes
• 9.88x forward P/E vs basket median ~30-40x — the only sub-15x P/E name in the field other than Sakai
• Mean PT HK$64.00 above spot HK$58.30 — the only name in the universe where consensus is still ahead of price
• Analyst consensus strong_buy
• 3.81% dividend yield — real Paid Now component vs Murata's 0.98%
• Collyer Bridge actively posting on Kingboard Laminates (subsidiary 1888.HK) in Subscriber Chat as of 23 May
• Basket framework Watch 19/30, Inflection 4/5 (four announced CCL price hikes through April 2026)

The differentiation from Murata is the structural point. Both participate in the AI infrastructure build, but through entirely different layers of the BoM, with non-overlapping ownership bases and non-overlapping retail-flow vectors. Owning Murata + Kingboard is genuinely two ideas rather than two expressions of the same idea. That is the correct structural answer to a crowding-aware top-2 question.

8. Entry Framework

Murata (6981.T) — BUY (scale-in)

• Spot ¥7,130 (close 23 May, +6% on day, near 52-week high ¥7,226)
• Mean sell-side PT ¥5,203 (17 analysts, 27% below spot); BofA PT ¥6,200 (15% below spot)
• Banyan Lane base-case 22x P/E target ¥7,254 — at this level today
• Banyan Lane bull-case 28x target ¥9,235 (+30% from here)
• Conservative entry: ¥5,500-6,000 on pullback (full position). Mid-conviction: ¥6,000-6,500 (partial / dollar-cost). Today's print: monitoring, not chasing the +6% close.
• Time horizon for the structural thesis: 18-24 months
• The bull case requires Murata to keep ~60% AI MLCC share through SEMCO's mid-2027 Korea ramp; downside scenario at 50% share weakens the thesis materially per Banyan Lane
• Tracking metrics: capacitor backlog (currently +89.5% YoY), AI MLCC ASP (target up 30-40% in 2026 per TrendForce), Murata Izumo plant ramp curve, FY27 Q1 print expected late July 2026 (Murata Q1 FY27)

Kingboard (0148.HK) — STARTER NOW + ADD ON PULLBACK

• Spot HK$58.30 (close 23 May, +7% on day, near 52-week high HK$58.80)
• Mean sell-side PT HK$64.00 (2 analysts, strong_buy, 10% above spot) — the only above-spot PT in the universe
• Forward P/E 9.88x, dividend yield 3.81%
• Starter position justified by sub-10x forward P/E and still-above-PT analyst consensus
• Add on pullback to HK$50-55 zone (5-15% below spot)
• Position cap reflects the structural risks (family-control HoldCo discount, property segment write-down risk, thin coverage)
• Tracking metrics: high-spec CCL pricing (four hikes through April 2026), Kingboard Laminates 1888.HK reporting (subsidiary, more pure-play CCL), Cheung family disclosure changes, property segment write-downs

9. What I Could Be Wrong About

The Murata thesis depends on AI MLCC share holding at ~60%. Samsung Electro-Mechanics is the credible competitive threat (the May 21 PDF flagged the CL31X227MRKNNW# and CL32X337MSVN4S# products — 220µF and 330µF in 1206/1210 sizes — as direct competition for Murata equivalent products). Worth tracking SEMCO's design-win momentum with hyperscalers over the next 2-3 quarters; the Banyan Lane downside scenario at 50% share weakens the FY28 EPS materially. Also, embedded deep-trench capacitors (eDTC) and HD-MiM on-chip integration are the long-term technological substitute. Not a near-term thesis killer but the right multi-year risk to track.

The Kingboard thesis assumes the CCL/PCB cycle pricing actions hold. Four hikes through April 2026 is the data; whether mid-spec CCL Chinese capacity additions undercut the price discipline is the live risk. The basket framework's Moat 3/5 reflects this — CCL is more commoditized than MLCC at the mid-spec tier. If Chinese mid-spec capacity arrives faster than expected, Kingboard's blended ASP compresses and the 9.88x forward P/E starts to look like a value trap rather than a discount.

The crowding lens itself can mislead. Stocks at limit-up are at limit-up because earnings revisions are catching up in real time; if the next earnings prints show the revisions accelerating faster than the price moves, the "crowded" tape can keep going. Yageo, Walsin, Holy Stone going limit-up today may be the next leg of the same cycle, not the top. The framework's Revision Velocity dimension is designed to surface this — when estimates are being revised up faster than the stock has moved, that's positive Revision Velocity, even if absolute crowding is high. The basket scored Murata Revisions 3/5 and Yageo Revisions 2/5 — meaning Murata's estimates are at least keeping pace with the price while Yageo's are lagging. That asymmetry is part of the case for Murata over Yageo.

Finally, the entire complex is +5-12% on 23 May. The Asia desk closes for the weekend and Asian investors will absorb the basket PDF, the BofA report, and the Substack-driven retail flow. Monday's open could be either a continuation move or a reversal. The "today" prices in this note may be 5-10% different by next session. Entry math should be re-checked Monday.

10. SemiAnalysis Cross-Check

SA mirror at ~/Dropbox/Wafflebun/KB/wiki/semianalysis/ searched for MLCC / capacitor / Murata / Yageo / Kingboard. No dedicated MLCC piece found in the mirror's 2020-2025 archive — MLCC content appears as supply-chain context in broader pieces (semiconductor roundup 2102022.md, advanced packaging part 2 review, energizing AI power delivery competition 2023). The 2023 power-delivery piece is the most relevant SA reference and is consistent with the AI MLCC content per rack thesis (more aggressive decoupling required at higher GPU current draws). No SA contradiction with the conclusions above; the SA take is implicit (MLCC content scales with compute density) rather than explicit (no specific MLCC stock thesis published).

11. Open Loops

• Pull Collyer Bridge "MLCCs" and "Kingboard Laminates" thread bodies in-browser (opencli React-virtualized chat panel limitation, see Section 5)
• CTR Holdings (8 replies, posted by Collyer Bridge 22 May) — separate name flagged for follow-up
• Sakai Chemical (4078.T) — upstream BaTiO3 powder, +3% today, 12.8x fwd P/E, 4.62% yield, no analyst coverage. Worth a dedicated /profile or /deep-dive run. The May 20 swarm flagged but didn't research.
• The Banyan Lane substitution-risk note on eDTC + HD-MiM warrants a multi-year tracking watch — not a near-term thesis risk, but the right thing to monitor over 24-36 months
• Murata FY27 Q1 earnings late July 2026 — first print testing whether the Apr 2026 price hikes flow through to reported financials with the assumed 1-2 quarter lag
• Kingboard Laminates 1888.HK as a more pure-play subsidiary alternative to 0148.HK (avoids the HoldCo property / chemicals overhang)

12. Sources

• BofA Securities, "Component Signals: MLCC production: Maintain Buy on Murata and TDK on solid earnings support," 19 May 2026 (in-tray)
• "Global MLCC & Passives Basket — Comparative Analysis," 21 May 2026 (in-tray)
• "Murata (6981.T) vs Yageo (2327.TW)," 21 May 2026 (in-tray)
• Goldman Sachs Equity Research, "Yageo Corp. (2327.TW): Profitability driven by solid AI demand will be a new norm; Buy, with new TP of NT$346," 16 April 2026 (in-tray)
• Banyan Lane Capital LLC, "The Most Boring Component in AI Is About to Have Its Moment," banyanlanecapital.substack.com, 6 May 2026
• Collyer Bridge Subscriber Chat, threads on MLCCs / MS Rubin note / Kingboard Laminates / Rigaku / CTR Holdings, 21-23 May 2026
• Korea Invest Insights, three pieces on Samsung Electro-Mechanics, 14-22 May 2026
• yfinance for live prices, analyst consensus, dividend yields (close 2026-05-23 Asia session)
• Prior wiki: mlcc-peer-swarm-2026-05-20, 6981/6981-t-deep-dive, 2327/2327-deep-dive, 6762/6762-deep-dive, 6976/6976-t-deep-dive, 3026/3026-tw-deep-dive, 6173/6173-two-deep-dive