Showdown: AP Memory (6531.TW) vs Nippon Chemical (4092.T)
The AI-passives supply chain, two rungs apart. As of 2026-06-01; prices verified live. Companion to the full deep-dives on each.
Full write-ups: AP Memory (6531) · Nippon Chemical (4092).
Read this first: this is not a peer comparison
These two are not competitors and they are not comparable on financials. They are two different layers of the same AI-passives supply chain, and you picked them deliberately to span it:
- AP Memory (6531.TW) is a fabless Taiwanese IC designer selling silicon capacitors - the new passive component that delivers clean power inside the most advanced AI accelerator packages, in the three places where ceramic capacitors physically cannot (above ~100 MHz, ultra-thin, embedded-in-substrate). Asset-light, ~46% gross margin, net cash, ~$5.9B cap.
- Nippon Chemical (4092.T) is a 133-year-old Japanese chemicals house making barium-titanate (BaTiO3) dielectric powder - the raw material that goes into the multilayer ceramic capacitors (MLCCs) sold by Murata, TDK, and Samsung. Asset-heavy, ~4% operating margin, net debt, ~$303M micro-cap.
So this compares the component vs the raw material, one fabless design house against one upstream-materials micro-cap nineteen times smaller. The unifying thesis is the AI-passives content explosion: an AI rack's capacitor content is rising fast, and both rungs benefit. The honest tension worth holding in mind: silicon caps are, at the bleeding edge, a mild substitute for the highest-frequency MLCC content - so owning both is very slightly internally hedged. In practice they are far more complementary than competing, because silicon caps remain a tiny niche while MLCC (and the BaTiO3 underneath it) is the volume.
At-a-glance
| AP Memory (6531.TW) | Nippon Chemical (4092.T) | |
|---|---|---|
| What you buy | Silicon-capacitor IC (fabless) | BaTiO3 dielectric powder (chemicals) |
| Supply-chain rung | Component into AI packages | Raw material into MLCCs |
| Price | NT$1,125 | ¥5,570 (+14.4% on the day, stop-high) |
| Market cap | NT$183B (~$5.86B) | ¥48.3B (~$303M) |
| Net cash / (debt) | +NT$11.5B (~$368M) | −¥6.3B (net debt) |
| 52-week move | +320% off low; −5.5% off high | +185% off low; at the 52w high |
| YTD | +145% | +95% (+43% in the last week) |
| RSI | 54.7 (neutral, cooled) | 71 (overbought) |
| Gross / operating margin | 46% / 28% | 20% / 3.9% |
| Revenue growth | +115% YoY | +3.4% full-year (+22% is Q4-only) |
| ROE | 13.3% | 6.0% (below cost of equity) |
| FCF margin | ~49% | <1% (heavy Tokuyama capex) |
| Dividend yield | 0.59% | 2.46% |
| Analyst coverage | 2 (mean PT NT$1,075; MS NT$666) | Zero |
| Insider | 36.6% (founder Chen) | 16.6% (Tanahashi family) |
Business model: opposite in almost every dimension
AP Memory owns IP and outsources all manufacturing to Powerchip, so capex is under 1% of revenue and the fab capital sits on someone else's balance sheet. That is what produces 46% gross margins, ~49% FCF margins, and a fortress net-cash position on a sub-$6B company. Its growth is explosive and recent: S-SiCap did NT$570M in Q1 2026, up 767% year on year, already ~27% of revenue, and it is the first qualified supplier into Intel's EMIB-T socket.
Nippon Chemical is the mirror image: it owns plants (the new Tokuyama BaTiO3 line), runs a 4% operating margin, generates almost no free cash flow because the buildout eats it, and earns a 6% ROE that sits below its cost of capital. Its "growth" is a cyclical recovery plus a one-off-inflated net line; full-year revenue grew 3.4%. The moat is a 12-24 month qualification lock-in on tier-one MLCC powder and, since 1 April 2026, a named JV with TDK. The business is structurally lower-quality, and it is cheap for reasons.
The two theses, and why each is "now"
AP Memory is the direct, leading-edge AI bet: every new 2.5D/3D accelerator package carries silicon-cap content, Intel is embedding silicon caps from 2027, and AP Memory holds a real first-qualified position. The catalyst is immediate (Q1 already inflected; Q2 prints 30 July).
Nippon Chemical is the indirect, lagged AI bet: Murata raised high-end MLCC prices 15-35% in April, capacity is sold out, and the dielectric powder underneath should reprice with a one-to-two quarter lag. But that ASP catch-up is an FY3/27-H2 event that has to show up in the financials, and management's own powder guide decelerates to +4.6% next year. The market just front-ran the theme via a micro-float squeeze.
Valuation: priced-for-perfection vs value-priced-but-squeezed
| AP Memory | Nippon Chemical | |
|---|---|---|
| Trailing P/E | ~114-121x (TTM) | 16.8x |
| Forward P/E | ~63-70x (not the 43.6x screens show) | 14.5x |
| P/B | 16.1x | 0.96x (≈ book) |
| EV/EBITDA | very high | 7.95x |
| Price vs fair value | above all credible PTs; MS NT$666 ≈ −41% | ran +43%/week; fair value ~¥3,500-4,500 |
The single most important valuation fact on AP Memory: the "cheap 43.6x forward" that screeners show rests on a corrupted yfinance EPS of NT$25.80 that no primary source supports. Morgan Stanley models ~NT$14, local sell-side NT$16-18, so the real forward multiple is ~63-70x, and MS's price target of NT$666 sits roughly 41% below spot. AP Memory is the better business, but it is genuinely expensive on any honest number, and it has already front-run a 2027 ramp.
Nippon Chemical is the opposite problem. At 0.96x book and 14.5x forward it is priced like value, and on a pullback it would be. But the cheapness has been largely arbitraged away by the +43%-in-a-week run, the move is a confirmed thematic squeeze on an 8.7M-share float (kabutan attributes back-to-back stop-highs to "barium-titanate MLCC speculation"; Nomura Asset Management filed a 5.44% stake on 21 May; Sakai ran +44% in sympathy), and underneath sits a 3.9%-margin, sub-cost-of-capital business. You are buying a real asset at a momentary squeeze price.
Composite scorecard
| Dimension | Weight | AP Memory | Nippon Chemical |
|---|---|---|---|
| Business quality | 20% | 4.0 | 2.5 |
| Financial health | 15% | 4.5 | 3.0 |
| Growth | 20% | 4.5 | 3.0 |
| Valuation | 20% | 2.0 | 3.5 |
| Quality & capital allocation | 10% | 3.5 | 3.0 |
| Risk (inverted) | 10% | 2.5 | 2.0 |
| Technical timing | 5% | 3.0 | 1.5 |
| Weighted | 100% | 3.53 | 2.83 |
AP Memory wins on everything except valuation, where Nippon Chemical is cheaper. The gap is business quality and financial health: a net-cash, 46%-gross-margin, founder-aligned franchise with a real AI design win, versus a net-debt, 4%-operating-margin, sub-CoE micro-cap riding a lagged catch-up that is not yet in its own guidance.
Verdict: both PASS today, for opposite reasons
AP Memory - PASS at NT$1,125 (Medium conviction). The best business of the two and the cleaner thesis, but ~63-70x a realistic forward number, above every credible target, with single-design concentration on one Google/MediaTek inference program. The momentum has cooled (RSI 54) but the price has not. WATCH NT$650-800, BUY NT$400-550. This is a good company you want to own at a price you should not pay.
Nippon Chemical - PASS / do-not-chase at ¥5,570 (Low conviction). Do not buy a micro-cap at a stop-high, at its 52-week high, RSI 71, up 43% in a week on a sentiment theme with no fundamental trigger. The asymmetry that existed at ¥3,905 (10.1x forward, near book, with TDK optionality) has been consumed. Watchlist for a retrace toward ¥3,500-4,000, or for the Q1 FY3/27 print in early August to prove the powder demand is real.
If you can only buy one (on a pullback)
It depends on what you want, and neither is a buy today. For quality plus thesis purity, AP Memory is the answer: a genuinely good, asset-light, net-cash AI-component franchise with direct, immediate exposure - you are simply waiting for the price. For deep value plus contrarian optionality with limited downside-to-book, Nippon Chemical: a 0.96x-book micro-cap with a real TDK anchor and a depreciated growth plant, where the AI catch-up is a lagged call option rather than a current fact. Higher quality and cleaner thesis points to AP Memory; cheaper, more contrarian, more idiosyncratic points to Nippon Chemical.
How they fit the AI-passives theme
These two actually do span the chain cleanly: AP Memory is the leading-edge component riding the package-architecture shift, Nippon Chemical the upstream material riding the MLCC content-and-price cycle. Owned together on pullbacks they are a coherent two-rung expression of "AI needs vastly more, and vastly better, passive power delivery." Owned at today's prices they are a momentum chase. The discipline is identical for both: the theme is right, the entry is wrong, wait.
Companion deep-dives: ~/claude/output/deep-dive/6531-tw-deep-dive.md, ~/claude/output/deep-dive/4092-t-deep-dive.md. Both were adversarially fact-checked; AP Memory's forward multiple and Nippon Chemical's segment/share/JV/TAM figures were corrected against primary and independent sources. Sources: yfinance live (2026-06-01); AP Memory Q1 2026 call + TrendForce + Digitimes + Morgan Stanley; Nippon Chemical FY3/26 results + kabutan + EDINET; prior vault profiles, silicon-cap primer, MLCC sector note, and the 6531-vs-6981-vs-009150 showdown.
Topics
- green-finance