4462 — Ishihara Chemical Co., Ltd. (石原ケミカル)
🔴 Live ¥2,927 (yf:4462.T) · 2026-09-05 · research written 2026-06-20 your re-engage <¥3,150–¥3,200 (own-DD, on-pullback) — now in zone (-9%)
Thesis
Verdict: NO-BUY at ¥3,730 (2026-06-20). Watchlist with a disciplined limit-buy below ~¥3,150–3,200. Business conviction Medium; buy-now conviction Negative.
Ishihara Chemical is a genuinely good ~120-year-old Kobe specialty-chemicals house whose plating segment (55% of sales, 70% of operating profit) sells the tin/solder and copper wafer-level plating chemistries that form the bump electrodes connecting chips to substrates — real, verified exposure to advanced semiconductor packaging. The balance sheet is a fortress (zero debt, 82% equity ratio, ~¥15bn net cash + securities ≈ 30% of market cap), the tin/tin-alloy plating franchise is a confirmed domestic #1, and on the company's own FY03/27 guidance the stock trades at ~18.5x forward P/E and ~8x EV/EBITDA (~6x ex-cash) — a real discount to JCU (4975) and Uyemura (4966).
The reason it is not a buy here is the gap between the "purest AI exposure" pitch and the verified reality, plus a poor entry. The anchor thesis (Made in Japan, 14 Jun 2026, author long a small position, no near-term catalyst) was written at a ¥43bn cap; the stock has since run ~18% in six days and +101% over twelve months, so the cheap entry is gone. The "AI" label is overstated — the plating segment is multi-end-market (autos, smartphones, PCs, general electronics), the company discloses no semiconductor split, and FY03/26 plating growth actually came from Chinese automotive substrates and RFID tags while semiconductors entered an "adjustment phase" (Q1 FY03/26 plating revenue −22% YoY). FY03/27 is a guided-down earnings year. Governance is the weak leg and moving the wrong way (poison pill being renewed, not retired). And the peer discount is largely justified, not pure mispricing: a ~26%-of-sales industrial-chemicals leg earning ~4% drags the group to a 16.4% operating margin versus JCU's 40.9% / Uyemura's 23.2%, and Ishihara is a self-described late entrant in the high-value wafer-copper node JCU dominates (~70% world share in via-fill copper).
What has to be true to own it: that low-to-mid-teens earnings growth resumes after the FY03/27 depreciation trough, that the plating cycle has turned back up, and that either a re-rating catalyst appears (NVIDIA-beneficiary recognition, pill abolition, buyback resumption) or you bought cheap enough that you don't need one. At ¥3,730 you need the catalyst; at ¥3,000–3,200 you don't.
Portfolio note (Pink-specific): this is a lower-quality, lower-conviction version of value-chain exposure already held at higher quality — Uyemura 4966 (BUY scale-in), MEC 4971, Ibiden 4062, Mitsui 5706. Marginal portfolio value is low and it adds overlap. Size as a starter ≤1% even on a pullback.
Snapshot
Ishihara Chemical Co., Ltd. (石原ケミカル) — TSE Prime: 4462 (4462.T). Kobe-based specialty chemicals, founded April 1900 (Meiji 33), >120 years old; renamed from Ishihara Yakuhin to Ishihara Chemical Oct 2013; listed Osaka 1991, TSE since 2011. Fiscal year ends 31 March; currency JPY. GICS sector Basic Materials, industry Specialty Chemicals. NOT to be confused with Ishihara Sangyo Kaisha (4028, TiO₂/agrochemicals) — a different company; the "40% payout / Vision 2030" target sometimes cross-cited belongs to 4028, not 4462.
A 120-year-old Kobe plating-chemicals house that makes the tin-bump and wafer-copper chemistry behind GPU/HBM packaging — priced like the diversified cyclical chemicals company it actually is, not the AI pure-play the pitch implies.
Valuation snapshot (2026-06-20, live ¥3,730):
| Metric | Value |
|---|---|
| Share price | ¥3,730 |
| Market cap | ¥51.0bn (~$340M) |
| Enterprise value | ¥35.7bn |
| P/E (TTM, FY03/26 EPS ¥217.33) | 17.2x |
| Forward P/E (FY03/27 guide EPS ¥201.97) | 18.5x — yfinance shows 26.5x; not reconcilable, use 18.5x |
| EV/EBITDA | 8.0x (~6x ex-cash) |
| P/B | 2.06x |
| EV/Revenue | 1.5x |
| ROE | 12.7% |
| Operating margin (group, FY03/26) | 16.4% |
| Dividend yield | 1.36% (DPS ¥44 FY26, ¥48 guided FY27; payout ~20%) |
| Net cash + securities | ~¥15bn (~30% of cap); zero interest-bearing debt |
| Beta | 0.63 |
| 52-week price change | +101% |
| Shares outstanding | ~13.66M (down from 15.67M FY22 via buybacks) |
Business
Four reported segments (FY03/25, ¥mn sales / operating profit):
- Metal Surface Treatment Agents & Equipment (plating) — 13,056 / 2,637 (20.2% margin), 55% of sales, ~70% of segment OP (77.6% of consolidated OP after adjustments). The earnings engine. Lead-free tin/tin-alloy plating = confirmed #1 domestic Japan share (the load-bearing, independently corroborated franchise — Fisco/kabutan repeat it). Wafer-bump plating forms the bump electrodes joining chip to package substrate in flip-chip assembly; wafer-copper plating is a smaller, growth-aspirational line where Ishihara is a late entrant vs JCU/Uyemura. The 50%-wafer-bump / 10%-wafer-copper internal split is MIJ's estimate — not disclosed. Segment margin was only 13.7% in FY03/24; the 20%+ level is a recent AI-cycle peak, not steady-state.
- Automotive Chemicals (UNICON brand) — 3,705 / 837 (22.6% margin), ~16% of sales. Non-cyclical aftermarket chemistry (A/C-unit cleaning liquids, etc.) sold OEM to Japanese auto dealers/repair/body shops/gas stations. More profitable than the MIJ thesis stated (~22–26% vs ~20%). The "A/C cleaning = 60% of segment" figure is unverified.
- Industrial Chemicals — 6,033 / 253 (~4% margin), ~26% of sales. The structural margin drag that distinguishes Ishihara from pure-play peers. Original/legacy business.
- Electronic Materials — 835 / 7 (~breakeven), ~4% of sales. The "free option": copper nano-ink and sintered copper bonding paste for power devices (room-temperature photonic sintering, vertically integrated). Company's stated "fifth business pillar." Not commercialised; rigorous power-device qualification cycles. Caution: some sintering-spec and silver-paste-superiority claims circulating online belong to other "Ishihara" entities — treat unsourced specs skeptically.
Moat: R&D/know-how and protected IP in a niche too small to attract large entrants but too specialised for new ones to catch up; rigorous customer qualification that creates high switching costs; very-Japanese close customer relationships and after-sales service. Plausible and consistent with the margins, but largely single-sourced to the company. Ishihara's own "high share in HBM/CPU" belief is not independently verifiable — even the company says it struggles to get share data.
Financials
Verified actuals (¥bn, FY ends March):
| FY22 | FY23 | FY24 | FY25 | FY26 | FY27e (guide) | |
|---|---|---|---|---|---|---|
| Revenue | 19.04 | 20.35 | 20.71 | 23.63 | 23.45 | 25.6 (+9.2%) |
| Operating profit | 2.36 | 2.14 | 2.33 | 3.40 | 3.84 | 3.76 (−2.1%) |
| Net income | 2.05 | 1.68 | 1.91 | 2.47 | 2.97 | 2.76 (−7.0%) |
| EPS (¥) | 130.0 | 110.3 | 127.6 | ~177 | 217.33 | 201.97 |
| DPS (¥) | — | — | 36 | 40 | 44 | 48 |
The balance sheet is a fortress: zero interest-bearing debt every year FY22–26, an 82.4% equity ratio, ¥9.2bn cash and deposits at FY26, and enough investment securities on top to reconcile to ~¥15bn net liquid via the EV bridge (cap ¥51.0bn less EV ¥35.7bn). Growth is real but the headline "~20% 10-year EPS CAGR" is trough-anchored — it holds only from the FY2016 cyclical low (a −21% year, EPS ¥36.33) to FY2026, at 19.6%. Over other windows it deflates fast: 11.0% for FY2014–24, 14.1% over twelve years, 13.7% over the recent four, and SimplyWall.st pegs five-year earnings growth near 10%. The durable rate is low-to-mid teens, helped about a point by buybacks. Those buybacks are lumpy rather than programmatic — one ¥3.1bn burst in FY25 cut the count below 15.15M shares, then FY26 was essentially nothing (~¥81k) — on a conservative ~20–23% payout with seven straight years of dividend increases.
Crucially, FCF is not capex-depressed in the actuals: FY26 free cash flow was ~¥3.15bn on ¥3.27bn operating cash flow with only ¥0.67bn of capex (~3% of sales). The capex ramp and its depreciation drag are forward (FY27 and beyond), which is precisely what guides FY03/27 operating profit down −2.1% on +9.2% revenue. ROE is ~12.7% (confirmed); the 15% ROIC headline could not be independently verified.
Competitive position & peers
Ishihara is the cheapest and lowest-quality of the three listed Japanese plating-chemistry names, and has lagged the rally:
| 4462 Ishihara | 4966 Uyemura | 4975 JCU | |
|---|---|---|---|
| Market cap | ¥51bn | ¥434bn | ¥184bn |
| P/E (TTM / fwd) | 17.2x / 18.5x | 31x / 34x | 20x / 26x |
| EV/EBITDA | 8.0x | 15.5x | 11.6x |
| P/B | 2.06x | 3.73x | 3.37x |
| Group op margin | 16.4% | 23.2% | 40.9% |
| ROE | 12.7% | 12.5% | 17.7% |
| Net cash | ~¥15bn | ¥57bn | ¥24bn |
| 12-mo price | +101% | +205% | +132% |
| Value-chain node | wafer-level tin-bump (niche #1 JP), late in Cu | FCBGA substrate surface finish (ENIG/ENEPIG) | via-fill Cu near-monopoly (~70% world) |
The discount is partly value, partly justified — both, not a binary. Justified: structurally lower margin (industrial-chem drag), weaker position in the highest-value copper node, sub-scale (¥51bn), illiquid, thin/English-free IR (~26 meetings/yr), poison pill, unreliable MT plans. Value: zero debt, ~30%-of-cap net cash, real tin-niche leadership, mid-teens growth, and it has lagged peers with room to re-rate. Global wafer-level plating-chem leaders (DuPont, Element Solutions/MacDermid, Atotech/MKS) do not include Ishihara — its "heart of GPU/CPU/HBM" framing is real but as a niche domestic tin-bump supplier, not a top-tier copper player.
Governance, ownership & risks
Governance is the weak leg. The takeover-defence poison pill is being renewed, not abolished — adopted 2008, renewed at every triennial AGM since, and resubmitted as the "New Response Policy" for a vote at the 25 June 2026 AGM, extended to ~2029 with a 20% trigger and the backing of all four Audit & Supervisory Committee members. That cuts against Japan's reform tide and reads as an entrenchment signal. There is no founding-family control: the CEO is Akihiko Fujimoto (not an Ishihara) and directors hold ~1.3% combined, with the register dominated by a stable-shareholder lattice (supplier/employee association 8.54%, Japan Custody/Master Trust, Nippon Life 5.55%, SMBC 4.92%, Mizuho). The cross-shareholding unwind is confirmed but modest (13 listed holdings, ~¥1.5bn book; policy to reduce "as a principle"). Customer concentration is a real, under-discussed risk — AMPOC Far-East at 13.8% and Nippon Steel at 11.1% of FY24 sales, about a quarter of revenue from two names. The rest of the bear list runs to cyclicality and inventory swings (FY03/26 saw a correction; Q1 plating −22%), raw-material exposure (Cu/Sn/Ag/Ni, directional and not company-quantified), historically unreliable medium-term plans (the FY03/26 OP target of ¥4.6bn set in 2023 was missed at ¥3.84bn), illiquidity and no near-term catalyst, electronic-materials capital-allocation risk, and M&A appetite. Financial health rates A; management quality is mixed (B−/C+) — operationally competent on the semiconductor pivot, but entrenchment-leaning and an under-distributor.
Valuation & scenarios (18–36 months, ~FY03/29)
The forward multiple is the key correction: 18.5x forward / 8x EV/EBITDA, not the 26.5x yfinance scare number — so the objection is catalyst and overlap, not price. The MT-plan math (EBITDA CAGR ~12% to FY03/29 implies EV/EBITDA ~4.8x) should be haircut given the plan's poor track record.
The bear (~¥2,600, −30%) is a cycle reversal or destock with copper traction failing and margin stuck near 16%, de-rating to 12–13x — though the balance sheet caps the downside, so it is not a zero. The base (~¥3,900–4,300, +5–15%) is mid-single-digit revenue compounding at a 16–17% margin, EPS recovering to ~¥240–260 by FY29 on a 16–17x multiple — roughly fairly valued today, where you earn the growth rather than a re-rating. The bull (~¥6,000–6,500, +60–75%) needs wafer-copper to scale, margins to expand toward peers, and the market to tag it an "NVIDIA beneficiary" alongside an abolished pill and resumed buybacks, lifting EPS past ¥300 at 20–22x — the catalyst the author admits does not yet exist.
Asymmetry is real (≈ −30 / +70) but the upside leg is catalyst-dependent and you'd be buying after an 18% six-day pop into a down-earnings year.
Decision log
2026-06-20 — Full deep-research + adversarial-verify run (11-agent workflow), live ¥3,730. Verdict NO-BUY at ¥3,730; watchlist limit-buy <¥3,150–3,200; starter ≤1% even then. Anchor: Made in Japan (14 Jun 2026, author long small, no catalyst) — thesis spine verified (55/70 segment split, tin #1 JP share, fortress balance sheet, FY25 OP +46% / FY26 +13%) but four load-bearing claims dented: (1) "70% of OP is AI" overstated — plating is multi-market, semi share undisclosed, plating −22% YoY in Q1 FY26; (2) "~20% 10-yr EPS CAGR" trough-cherry-picked, durable rate low-to-mid teens; (3) no family steward + pill being renewed (25 Jun AGM); (4) discount largely justified (16.4% group margin vs JCU 40.9% / Uyemura 23.2%; late entrant in wafer copper). Key positive correction: forward P/E is 18.5x not 26.5x. What flips it to buy-now: pill abolished at the AGM, buyback/payout step-up, evidence wafer-copper is winning share vs JCU, or a clean print confirming the plating cycle re-accelerated. Triggers to revisit lower: ¥3,000–3,200 limit.
Sources
- Made in Japan, "A quick note on a 120-year-old Japanese Chemical Company at the heart of the GPUs CPUs and Memory," 2026-06-14 (paid; accessed via Pink's session) — anchor thesis, author long
- Ishihara Chemical FY03/25 & FY03/26 kessan-tanshin; company site unicon.co.jp (product/IR pages); EDINET 有価証券報告書; 2026-05-15 AGM/takeover-defence disclosure
- IR Bank (E00804), Monex scouter, kabutan, Fisco, Kobe Keizai, kabutore — financials, segment, ownership, guidance triangulation
- yfinance (4462.T, 4966.T, 4975.T), pulled 2026-06-20
- STF Research (stfbutnou.substack.com) — no Ishihara post; their adjacent Uyemura coverage is folded into 4966
- Vault cross-refs: 4966 (Uyemura — peer, AI-substrate incumbent), 4975 (JCU — peer, via-fill Cu), cu-wiring-resin-primer, advanced-packaging, ai-infrastructure