Financebotresearch desk研究台

Discovery › screen

Screen japan tech semis

Date: 2026-07-29 Criteria: Convert the Japan finding from screen-global-vs-us-value-2026-07-29.md into names. Field: Japanese tech, semis, components, automation. Framework: analysis_notes.md §0 (trap test), §1 (health), §3 (valuation) Status: First analysis of any Japanese company by this agency.

Executive Answer

The Japan opportunity and the Japan-technology opportunity are two different trades, and they barely overlap.

The governance/buyback thesis our global screen identified lives in banks, insurers, trading houses, autos and low-ROE conglomerates. It does not live in Japanese semis. Japanese semis are priced as AI winners — because they are — at the top of a record equipment cycle:

Name P/E (ttm) P/B Total shareholder yield Verdict
Advantest 60.2x 28.1x 0.25% 🔴 Priced for perfection
Ibiden 70.8x 7.7x 0.26% 🔴 Great asset, untradeable, absurd multiple
Murata 59.1x 5.1x ~1.1% 🔴
Tokyo Electron 43.7x (Simply Wall St: 59.2x) ~12x 1.8% 🔴 Best business, worst price
Fanuc 39.1x 3.5x 4.1% 🟡 Cash-rich, but 31.5x fwd for 9% ROE

And the genuinely cheap Japanese tech names are cheap for the classic thirty-year reason:

Name P/B ROE Why it's cheap
Omron 1.26x 4% Low returns, 12.9K shares/day ADR
Kyocera 1.44x 4% Cross-holdings = 53% of net assets
Nikon 1.07x −14% Loss-making
Nidec 1.56x 6% Accounting fraud — see below

Exactly one name sits in both circles: Shin-Etsu Chemical. It is the only company in this field that is simultaneously a high-quality business, in a real drawdown, and delivering the buyback thesis at scale.

🚩 Three findings that change the framing

1. Japan's "−8.4% drawdown" is ~100% currency. There is no index-level drawdown to buy.

Our global screen flagged EWJ at −8.4% off high — the largest of any developed market. That is not a selloff.
TOPIX Record closing high 4,101.96 on July 6, 2026 (prior record 4,095.05, June 22)
Nikkei 225 71,962 high in the June 28–July 28 window
USD/JPY ~163.5 (July 30), broke above 163 on July 21 — weakest yen in ~four decades
Yen, trailing 12 months −8.41%

EWJ's −8.4% and the yen's −8.41% are the same number. Japanese equities in local currency are at an all-time high; the USD investor's drawdown is entirely the exchange rate. Anyone treating EWJ's drawdown as a valuation opportunity is misreading an FX chart as an equity chart.

But the refinement matters: the index masks enormous dispersion. Japanese tech has genuinely sold off while domestic/cyclical names carried the index to records — "profit-taking in AI and semiconductor-related names offset by robust buying in domestic demand-oriented and cyclical sectors." Real drawdowns exist at the single-stock level: Murata −40.5%, Ibiden −39.5%, Nidec −36.0%, Renesas −34.2%, Tokyo Electron −33.0%, TDK −30.7%, Shin-Etsu −25.6%, Sony −25.1%.

2. The currency is a live, quantified risk — and it cuts both ways

At USD/JPY 163.5, near a 40-year low, with the Fed at 3.50–3.75% versus the BOJ at 1.00%:

Scenario USD/JPY USD return impact
Bank consensus (several majors) 150–155 +5.5% to +9.0%
Bull yen (BOJ tightens, Fed eases) 145 +12.7%
Consensus range (some forecasters) 167–175 −2.1% to −6.6%

Japan spent ¥11.7T defending the yen and it still broke 163. MUFG's July 2026 view: Fed policy, not BOJ policy, remains the dominant driver, so near-term dollar strength persists.

Honest framing: buying Japanese equities at 163 is buying an asset at a record high in a currency at a 40-year low. The currency is closer to a tailwind option than a headwind here — the asymmetry favours the buyer — but it is an FX bet layered on an equity bet, and the equity leg is not cheap. None of the three names below work if you need the yen to be the reason. If the yen is the thesis, DXJ (hedged) or a currency position is the honest instrument, not a stock.

3. Data hazard confirmed and worse than warned

fin.py returns negative enterprise value for essentially every Japanese ADR — EV computed in JPY against a USD market cap:

Ticker Printed EV Printed EV/EBITDA
TOELY −369.79B −0.52
FANUY −699.22B −3.02
SHECY −1.15T −1.29
MRAAY −529.53B −1.06

All EV/EBITDA and P/S figures for Japanese ADRs in this data source are unusable. I hand-computed EV from market cap + debt − cash for the three finalists (below). Additional traps found:

  • SCRYY is SCOR SE, a French reinsurer — not Screen Holdings (7735). Screen has no usable US line. A screen that trusted the ticker would have surfaced a French insurer as a Japanese semi-equipment play.
  • CAJ (Canon) returns no data at all.
  • KYOCY's quote is dislocated: last trade $22.89 but bid $24.46 / ask $24.55 — the quote sits 7% above the last print on 82K shares/day.
  • Disco (6146), Lasertec (6920), Screen (7735), SUMCO (3436), Keyence, Yaskawa, Harmonic Drive, Rohm, Socionext, Tokyo Ohka and JSR have no practically tradeable US line. Much of the most interesting part of this field is simply not reachable from a US retail account.

The mandatory test: cheap on the multiple, or cheap on shareholder returns?

Our own research said Japan's cheapness has migrated into buybacks and dividends. So I computed it directly from cash-flow statements (FY ending 2026-03-31), converting at USD/JPY 163.5.

Name Buyback (¥) Prior yr (¥) Buyback yield Div yield Total Trend
Shin-Etsu ¥500.0B ¥194.0B 4.48% 1.93% 6.41% ⭐ +158%
Kyocera ¥200.0B ¥0.004B 4.07% ~0.2% ~4.3% from zero
Fanuc ¥150.0B ¥150.0B 2.34% 1.73% 4.07% flat
Sony ¥522.1B ¥285.6B 2.39% 0.62% 3.01% +83%
Tokyo Electron ¥114.3B ¥50.1B 0.46% 1.32% 1.78% +128%
TDK ¥0 ¥0 0.00% 1.36% 1.36% none
Advantest ¥0.55B ¥49.6B 0.03% 0.22% 0.25% −99%

The dispersion is the finding. Shin-Etsu returns nearly 5x what TDK returns and 25x what Advantest returns. The governance thesis is real but it is name-specific, not country-specific — buying "Japan" gets you Advantest's 0.25% as readily as Shin-Etsu's 6.4%.

Note Shin-Etsu paid out ¥703B against ¥357B of FCF — a ~197% payout funded by its ¥1.15T net cash pile. That is precisely the "cash-release leg" our global screen said was what remained after the re-rating leg. It is also, by definition, not repeatable indefinitely.

Top 3

1. ⭐ Shin-Etsu Chemical — SHECY (ADR) / 4063 (TSE)
Sub-field Semiconductor materials (silicon wafers) + PVC
Price / Market cap $18.41 · $68.5B
% off 52w high −25.6% ($24.74) · +35% off low ($13.64)

Why it's a gem. It is the only name in this field that passes both §0 tests at once: a genuinely superior business (26% operating margin, 34% gross margin, 11% ROE, 6.09x current ratio, essentially no net debt) that is also delivering the shareholder-return thesis at 6.4% and also in a real 25% drawdown. The market is discounting Shin-Etsu's PVC half — housing-linked and weak — and in doing so is marking down a wafer franchise that sits in a five-player oligopoly controlling ~85% of global 300mm capacity and that just guided 2026 shipments up 15%.

Metric Value Note
P/E (ttm) 23.9x
P/E (fwd) — ⚠️ not published for this ADR
P/B 2.46x
EV/EBITDA ~11.3x ⚠️ hand-computed ($68.5B − ~$7.1B net cash ≈ $61.4B EV; EBITDA est. ~$5.4B). Printed value −1.29 is garbage.
FCF ¥356.7B (~$2.18B) FCF yield ~3.2%
Revenue CAGR ~5% yoy Low-growth; this is not a growth story
Net income growth +6% yoy
Debt / Assets minimal (D/E ~5.5%) ¥1.15T net cash
ROE / Op margin / Gross margin 11% / 26% / 34%
Div + buyback yield 1.93% + 4.48% = 6.41% Buyback +158% yoy
Payout ratio 41% (dividend only)

Moat — efficient scale + process intangibles. Five firms (Shin-Etsu, SUMCO, GlobalWafers, Siltronic, SK Siltron) hold ~85% of 300mm capacity. Wafer qualification cycles with fabs run years; a defect excursion is catastrophic for the customer, so incumbency is extraordinarily sticky. Capacity additions are capital-heavy and the incumbents have shown genuine discipline in not adding. Shin-Etsu is the #1 player and has the best cost position. The PVC business (Shintech, US Gulf Coast) is a separate cost-advantage moat — lowest-cost producer with captive ethylene.

Where in the cycle — favourable, and this is the key differentiator. Unlike the equipment names, wafers are exiting a downcycle, not peaking. Q1 2026 wafer shipments +13%; SUMCO revised 300mm shipments up 12% (Jan), Shin-Etsu +15% (Feb), GlobalWafers +20% (Mar) — a synchronous upward revision across the oligopoly signalling formal exit from the 2023–24 downcycle. Volume 12.82B → 13.41B sq inches 2025→2026. This is early-to-mid cycle, not peak.

Bear case (adversarial). - The "cheap for thirty years" objection is only half-deflected. Shin-Etsu at 23.9x is not statistically cheap. It has never been a P/B-under-1x Japan trap — it's a quality compounder — but that means you are paying a fair price for a good business, not buying a mispricing. Graham IV computes to $11.38 against a $18.41 price — Graham says 38% overvalued. I weight it low (§3: Graham is for asset-backed value plays, not high-margin oligopolists) but I will not hide it. - The buyback is a cash-pile drawdown, not earnings power. ¥703B returned against ¥357B FCF. When the ¥1.15T net cash normalises, the 6.4% yield reverts toward ~3%. Do not underwrite 6.4% as a run rate. - PVC is genuinely bad and could get worse. Roughly half the company is a housing-cycle chemical business facing high US rates and Chinese capacity. If PVC deteriorates faster than wafers recover, the "hidden semi play" thesis inverts into "chemical company with a semi option." - Mature-node and 200mm are still in a glut. The recovery is bifurcated: AI/leading-edge is strong, mature-node logic is still working off inventory. Shin-Etsu is exposed to both. - China localisation. A national 300mm wafer localisation push is explicitly underway. Wafers are exactly the kind of commoditisable input a state programme can attack over 10 years. - Currency. −25.6% off high is measured in USD and therefore includes the yen's −8.4%. In yen the drawdown is nearer −19%. You are getting less of a discount than the ADR chart suggests.

Tradeability — good, the best of the OTC names. SHECY trades on OTC Pink (unsponsored-style ADR) but with real depth: 413K shares/day 3-month average (999K over 10 days), bid $18.41 / ask $18.43 — a 0.11% spread, ~$7.6M daily dollar volume. That is comfortably tradeable for retail size. ⚠️ OTC status still means: no options, possible depositary fee (typically $0.01–0.05/share, usually netted from the dividend), and no automatic reduction of Japan's 15.315% dividend withholding — expect to reclaim via the foreign tax credit. The local line 4063 would be marginally better on spread and would avoid ADR fees, but the improvement does not justify the account complexity at retail size. Use the ADR.

Fair value & entry. Bogle expected return: 1.93% dividend + ~6% earnings growth + ~2–3% from share-count reduction ≈ 10–11% before any multiple change, which is a credible long-term return. On 20–24x forward earnings of ~$0.85 (assuming ~10% growth off the wafer upcycle): fair value $17–21. Current $18.41 is inside fair value, not below it.

Entry zone: $15.00–16.75. Attractive only on a further ~10% drawdown or a yen rally toward 150. At $18.41 this is a watch, not a buy. Conviction: 6/10. High confidence in the business, low confidence that today's price is a bargain.

2. Sony Group — SONY (NYSE) / 6758 (TSE)
Sub-field Diversified tech — image sensors, games, music, pictures
Price / Market cap $22.73 · $133.5B
% off 52w high −25.1% ($30.34)

Why it's a gem. Sony is the only genuinely investable name in this field — NYSE-listed, sponsored, 6.5M shares/day, zero bid-ask spread ($22.76/$22.76), options available. It is also a live governance-reform case study: it completed the partial spin-off of Sony Financial Group (a real conglomerate-simplification, not a press release), nearly doubled its buyback to ¥522B, and is shrinking share count ~1.1%/yr. Image sensors (~50%+ global share in CMOS) are a durable franchise levered to the one thing every phone and every autonomous system needs more of.

Metric Value Note
P/E (ttm) 21.7x ⚠️ distorted — see below
P/E (fwd) 19.1x The number to use
P/B 2.71x
EV/EBITDA ~9–10x ⚠️ hand-checked. Printed EV/EBITDA of 0.02 is nonsense.
FCF ¥1.49T (~$9.1B) FCF yield ~6.8% — the best in this field
Revenue CAGR (3y) 4.4%
OCF CAGR (3y) 83.5% Flattered by a weak 2023 base
Debt / Assets 10.65%
ROE / Op margin 12% / 11%
Div + buyback yield 0.62% + 2.39% = 3.01% Buyback +83% yoy
Shares (dil) CAGR −1.1%/yr
Analyst target $29.38 (+29%)

⚠️ Data flag: Sony's FY2026 reported net income is −¥326.9B and total assets fell from ¥35.29T to ¥15.68T. Both are artifacts of the Sony Financial Group spin-off, not operating deterioration — operating income actually rose to ¥1.55T. Trailing EPS, ttm P/E and "EarnGrowth −57%" are all contaminated. Use forward figures only for Sony.

Moat — intangibles + switching costs, unevenly distributed. The image-sensor business is a real moat: ~50%+ share, decades of process IP, and design-in cycles that lock customers for years. PlayStation is a network-effect moat (installed base + third-party developers). Music is the best asset of all — a catalogue that collects streaming royalties in perpetuity with near-zero incremental cost, structurally advantaged as streaming penetration grows. Pictures is the weakest link and the least defensible.

Bear case (adversarial). - Not cheap enough to be the reason to enter Japan — the same conclusion our global screen reached at $23.27, and it still holds at $22.73. 19.1x forward for 4.4% revenue growth is a fair price, not a gift. - Conglomerate discount is deserved and may be permanent. Games, sensors, music, movies and (until recently) insurance share no operating logic. The spin-off helped; four unrelated businesses remain. - Sensor concentration risk. Apple is an enormous customer. Samsung's sensor business is closing the gap, and Chinese sensor makers are moving up from the low end. - Games is capital-hungry and hit-driven, with Microsoft/Game Pass structurally attacking the console economics. - 3.0% total shareholder yield does not clear the governance bar. Sony is a decent capital returner, not a great one. If the thesis is "buy Japan's cash release," Sony is a weak expression of it. - Currency: −25.1% off high in USD; nearer −18% in yen.

Tradeability — excellent, and uniquely so. NYSE (NYQ), sponsored ADR, 6.48M shares/day 3-month average, $0.00 spread, ~$147M daily dollar volume, listed options. No liquidity, spread, or execution constraint whatsoever. This is the only name in the field where tradeability is a non-issue, and that is worth a point of conviction on its own.

Fair value & entry. At 20–23x forward EPS: fair value $24–28, corroborated by the $29.38 street target. Current $22.73 is modestly below fair value.

Entry zone: $19.50–21.50. Add aggressively below $20. Conviction: 6.5/10. The highest-quality access in the field; a fair-not-cheap price.

3. TDK Corporation — TTDKY (ADR) / 6762 (TSE)
Sub-field Electronic components — HDD heads, MLCC, batteries, sensors
Price / Market cap $18.53 · $35.2B
% off 52w high −30.7% ($26.73)

Why it's a gem. TDK is the cheapest growing quality name in the field and the purest Japanese expression of the AI-storage thesis this agency already holds through US names. It dominates HDD magnetic recording heads, and nearline HDD head volumes are guided up ~50% next fiscal year on AI data-centre demand — after already rising 14%. It posted record FY2026 results (revenue ¥2.50T, +13.6%; operating profit ¥272.4B, +21.5%) and carries a strong_buy with a $26.45 target, +43% — the largest implied upside in the field — while trading at 24.1x forward against Advantest's 60x.

Metric Value Note
P/E (ttm) / (fwd) 29.4x / 24.1x
P/B 2.63x
EV/EBITDA ~10.6x ⚠️ hand-computed ($35.2B + $3.8B debt − $5.2B cash ≈ $33.8B EV; EBITDA est. ~$3.2B). Printed −0.38 is garbage.
FCF ¥209.1B (~$1.28B) FCF yield ~3.6%
Revenue CAGR (3y) 4.7% FY26 alone +13.6% — accelerating
Net income CAGR (3y) 19.7% Best in the field
OCF CAGR (3y) 24.5%
Debt / Assets 13.95% Net cash ¥227B
ROE / Op margin / Gross margin 10% / 5%* / 31% *snapshot op margin conflicts with statements (10.9% actual) — flagged
Div + buyback yield 1.36% + 0.00% = 1.36% ⚠️ Fails the shareholder-return test outright
Shares (dil) CAGR 0.0% No buyback, no dilution
Analyst target $26.45 (+43%) strong_buy

Moat — process intangibles + cost advantage. HDD magnetic heads are a two-and-a-half player market with brutal physics; TDK's head business is the leader and has already begun MAMR mass production with HAMR ~2 years out. Head technology gates areal density, which gates the entire nearline HDD roadmap — a genuine bottleneck position. MLCCs are a scale-and-yield game where TDK is top-three. The small-cell battery business (ATL) is #1 in smartphone Li-ion. Each moat is real but none is as wide as Shin-Etsu's wafer position.

Bear case (adversarial). - It flatly fails the test this screen was built around. Zero buyback, 1.36% total yield, flat share count. If Japan's opportunity has migrated into shareholder returns, TDK is not participating. It is a Japanese-domiciled growth stock, and it must be justified on growth alone. - The "cheap for thirty years" objection lands hard here. TDK has been a perennially cheap-looking Japanese component maker for decades precisely because it earns ~10% ROE on a capital-intensive base and returns almost nothing. Nothing in the FY2026 results changes the capital-allocation culture — capex was ¥298.6B against ¥60.7B of dividends. Management's instinct is to build, not to return. - HDD is a structurally declining medium. The AI-nearline reprieve is real but it is a reprieve. QLC NAND is attacking the nearline tier from above and the cost-per-bit crossover is a matter of when. A 50% volume guide is a cyclical spike inside a secular decline. - MLCC and batteries are commodity-adjacent with Chinese competition (CATL, Sunlord) compressing pricing. - Graham IV: $9.99 vs $18.53 — 85% overvalued on Graham. Low weight (§3), but the lowest Graham score of the three. - Tradeability is the worst of the three (below). - Currency: −30.7% in USD, nearer −24% in yen.

Tradeability — usable but the weakest of the three. OTC Pink, 283K shares/day, ~$5.2M daily dollar volume, and bid $18.37 / ask $18.49 — a 0.65% spread, five to six times wider than SHECY's and infinitely wider than SONY's. Round-tripping costs ~1.3% before commission. ⚠️ Unsponsored-style OTC line: no options, depositary fee likely, 15.315% Japanese withholding. Use limit orders only — never a market order. For a position of any size, the local line 6762 is materially better here, which is not true for Shin-Etsu. This is a real constraint, not a footnote.

Fair value & entry. At 22–26x forward: fair value $21–26. The street's $26.45 sits at the top of that. Current $18.53 is the clearest discount-to-fair-value of the three — which is the entire argument for it.

Entry zone: $15.50–17.50. Conviction: 5.5/10. Best upside, worst capital allocation, worst execution. Owning it means explicitly abandoning the governance thesis and buying the AI-storage cycle instead.

Killed — and why

Name Verdict
Nidec (NJDCY/6594) 🔴 DISQUALIFIED. Screened as the cheapest name in the field (10.7x fwd, 1.56x P/B) — and it is the textbook §0 value trap. At least 1,000 instances of improper accounting; ¥160.7B cumulative hit to net assets; restating back to FY2022; FY2026 results postponed; TSE designated it a "security on special alert" (Oct 28, 2026); executive liability probe underway; investigators cite a "culture of silence." Uninvestable. That the screen's cheapest multiple belongs to the fraud is the cleanest §0 illustration this run produced.
Tokyo Electron (TOELY/8035) 🔴 The best business in the field at the worst price. 43.7x here, 59.2x per Simply Wall St vs a 30.2x JP semi industry average. SEMI: equipment sales at a record $139B in 2026 → $156B in 2027, with 2027 the likely absolute peak. §0: a high multiple on peak earnings is the trap, not the bargain. 1.8% total yield. Revisit below $110.
Advantest (ATEYY/6857) 🔴 60.2x P/E, 28.1x P/B, and it cut buybacks 99% (¥49.6B → ¥0.55B). Priced for perfection, returning nothing, only −15% off high.
Ibiden (IBIDF/4062) 🔴 The one that hurts. Genuine AI bottleneck — ABF/FC-BGA substrate duopoly, ¥500B ($3.3B) capex over FY2026–28 to expand capacity ~2.5x, −39.5% off high. But 70.8x P/E, 7.7x P/B, and 314 shares/day with a 5.3% bid-ask spread ($88.19/$92.89). Effectively untradeable. Great asset, no access, absurd price.
Kyocera (KYOCY/6971) 🔴 The purest governance story — Oasis holds 4.3% and has filed proposals to remove Chairman Yamaguchi; cross-holdings are 53% of net assets vs ISS's 20% threshold; KDDI stake (15.29%, ~¥1.7T) being monetised (¥124.8B sold June 2026); ¥200B buyback from a standing start; ROE target 4.3% → 10% by 2031. But the trade is already on: only −8.4% off high, +116% off the low, 35.8x P/E on 4% ROE, and the mean analyst target is $14.90 against a $22.89 price — 35% downside. You are late. Quote is also dislocated (bid $24.46 vs last $22.89) on 82K shares/day.
Fanuc (FANUY/6971→6954) 🟡 Closest to a fourth pick. Huge net cash, 4.07% total yield, 6.89x current ratio, good ADR liquidity (531K/day, tight). But 31.5x forward for 9% ROE and 11% revenue growth, buyback flat yoy, only −23% off high. Not cheap.
Murata (MRAAY/6981) 🔴 59.1x P/E despite −40.5% off high — the drawdown was from a bubble, not to a bargain.
Renesas (RNECY/6723) 🔴 Negative EPS, D/E 47, net margin −1%, no dividend coverage. Automotive/MCU exposure into a soft auto cycle.
Omron (OMRNY/6645) 🔴 P/B 1.26 with 4% ROE — the thirty-year trap in one line. 12.9K shares/day.
Nikon (NINOY/7731) 🔴 −14% ROE, negative operating margin, payout ratio 195%. Dividend is being paid out of nothing.
Hitachi (HTHIY/6501) 🟡 strong_buy, 13.5x fwd, real restructuring — but as our global screen noted, Hitachi Energy makes it partly an AI-grid derivative, double-counting exposure already owned. Only −14.5% off high.
Fujitsu (FJTSY/6702) 🟡 21.9x/22.1x, ROE 16%, −22% off high. Domestic IT services with limited AI leverage. No forward discount.
Sumco, Disco, Lasertec, Screen, Keyence, Yaskawa, Rohm, Socionext, Shinko, Tokyo Ohka, JSR, Harmonic Drive ⚪ No practically tradeable US line. Excluded on access, not on merit — several (Disco, Lasertec) are among the most interesting assets in global semis.

Verdict on the field

The field is mostly a pass, and the reasons are specific rather than generic.

  1. The premise didn't survive contact. We came looking for Japanese tech sold off alongside the −24% SOX drawdown and trading at Japan's discount. What exists instead is a barbell: the AI-levered names (Advantest 60x, Ibiden 71x, Murata 59x, Tokyo Electron 44–59x) are priced as global AI winners with no Japan discount at all, while the names carrying a real Japan discount (Omron, Kyocera, Nikon, Nidec) carry it because of 4% ROE, negative ROE, or accounting fraud. The discount, where it exists, is earned.

  2. The governance thesis is real but lives elsewhere. Record buybacks (¥16.2T Jan–May 2026, +34% yoy), cross-shareholding unwinds committed through 2030, and a Corporate Governance Code revision landing this summer are all genuine. But the buyback yields inside this field span 0.03% to 4.48% — a 150x range. Japanese tech is largely not where the cash release is happening; banks, insurers, trading houses and autos are. Our global screen's own pick, Toyota, remains a better expression of the Japan thesis than anything in this field.

  3. There is no index-level entry. TOPIX is at a record high. EWJ's −8.4% is the yen, to the decimal. Anyone buying "Japan because it's down" is buying a currency chart.

  4. Access is a first-order constraint here, not a footnote. Roughly half the genuinely interesting assets in this field — Disco, Lasertec, Screen, SUMCO, Keyence, Shinko — have no tradeable US line, and the single best AI-bottleneck story (Ibiden) trades 314 shares a day at a 5.3% spread. A US retail investor cannot actually buy the best of Japanese semis.

  5. Three names survive, and none is a table-pounder. All three are fairly valued rather than cheap; all three sit inside or barely below my fair-value ranges; and all three would be genuinely attractive roughly 10–15% lower. This is a watchlist outcome, not a buy outcome.

Is this time different? Partially, and I'd resist the stronger claim. The governance reform is measurably real — the buyback numbers are not narrative, and the cross-holding unwinds are contractual. But the reform has already been capitalised: P/B at 1.7x is a post-2008 high, the Prime market has re-rated 35–40% since 2022, and TOPIX is at a record. The re-rating leg is over. What remains is slower, name-specific cash release — which means Japan is now a stock-picker's market, not an allocation. That is exactly the conclusion our global screen reached, and this field confirms it from the bottom up.

🎯 The one name worth a full /analyze

Shin-Etsu Chemical (SHECY / 4063)

It is the only name in the field that passes both §0 questions simultaneously — good business and a defensible answer on price — and the only one where the shareholder-return thesis, the drawdown, the cycle position, and tradeability all point the same direction:

  • 6.41% total shareholder yield, buyback up 158% yoy, funded by a ¥1.15T net cash pile that reform pressure is explicitly targeting
  • Early-cycle, not peak-cycle — the one semi-linked name here where the §0 peak-earnings trap does not apply, with three of the five global wafer oligopolists synchronously revising 2026 shipments up
  • Real 25.6% drawdown with 11% ROE, 26% operating margins and essentially no net debt
  • Genuinely tradeable — 413K shares/day at a 0.11% spread

The open questions an /analyze must resolve, and they are material: (a) how much of the company is really PVC versus semiconductors, and what the housing cycle does to the former; (b) what the buyback run-rate is once the excess cash normalises — the 6.4% is not repeatable; (c) whether China's 300mm localisation push is a 10-year threat or a 3-year one; and (d) whether a US-based investor should hold this unhedged with the yen at a 40-year low.

Runner-up for the /analyze slot: SONY — not because it's cheaper, but because it's the only name in this entire field an investor can trade without friction, and that has real option value.

Recommendations

Add to Watchlist — 👀 On Deck (none are in-zone today):

Ticker Local Sleeve Entry zone Conviction Rationale
SHECY 4063 ⚡ AI & Infrastructure Capex $15.00–16.75 6/10 Wafer oligopolist, 6.4% total yield, early-cycle. Fair at $18.41, not cheap.
SONY 6758 🏛 Evergreen Compounders $19.50–21.50 6.5/10 Only frictionless access in the field; sensors + music are durable; spin-off is real reform.
TTDKY 6762 🔧 Re-Rating Plays $15.50–17.50 5.5/10 +43% target, AI-storage levered — but zero buyback and a 0.65% spread. Limit orders only.

Do not add: Nidec (fraud — permanent exclusion), Tokyo Electron / Advantest / Murata (peak-cycle multiples), Ibiden (untradeable), Kyocera (trade already run, target 35% below spot), Omron / Nikon (value traps), Renesas (loss-making).

Portfolio fit: all three would be first Japan exposure in either account. Note SHECY and TTDKY both add to the AI/semiconductor cluster the portfolio is already heavy in — per the concentration map, they are correlated with existing ASML/KLAC/SNDK-type exposure rather than diversifying it. SONY is the only one of the three that genuinely diversifies, which argues for it despite the thinner discount.

Suggested next step: /analyze SHECY — and if a Japan allocation rather than a Japan name is what's wanted, the honest instrument is DXJ (yen-hedged) rather than EWJ, given the yen sits at a 40-year low and EWJ's entire drawdown is currency.

Data gaps & caveats

  • ⚠️ All EV/EBITDA and P/S figures for Japanese ADRs from fin.py are unusable (JPY EV vs USD market cap → negative EV). The three EV/EBITDA figures above are hand-computed and approximate — EBITDA was estimated from operating income plus an assumed D&A, since D&A is not broken out in the bundle. Treat them as ±20%.
  • ⚠️ No forward P/E published for SHECY, TOELY, ATEYY, MRAAY, IBIDF, KYOCY(unreliable) — forward multiples unavailable for much of the field.
  • ⚠️ Sony's trailing figures are contaminated by the Sony Financial Group spin-off (reported net income −¥326.9B, assets 35.29T→15.68T). Forward figures only.
  • ⚠️ fin.py market cap for ADRs = shares outstanding × ADR price, which is wrong wherever the ADR ratio ≠ 1:1. I verified 1:1 consistency for SHECY, KYOCY, TTDKY and SONY by reconciling ADR EPS against local EPS at USD/JPY 163.5, but did not verify it for every name in the killed list.
  • ⚠️ Recommendation: none for most of the field — no analyst consensus or price target is available for SHECY, TOELY, ATEYY, MRAAY, FANUY, KYOCY(target $14.90 present but with none consensus). The +43% TDK target and +29% Sony target are the only two I'd lean on, and only lightly.
  • ⚠️ ADR depositary fees not verified per-name. I have flagged them as a general OTC risk (typically $0.01–0.05/share/yr) but did not confirm the specific fee schedule for SHECY or TTDKY.
  • ⚠️ Japanese dividend withholding is 15.315%. Not automatically reduced on OTC ADR lines; recovery via US foreign tax credit assumed but not verified for this user's brokers.
  • ⚠️ Only ~4 annual periods available from fin.py, so all CAGRs above are 3-year, not the framework's preferred 5–8 year. Escalating to roic.ai is required before any real conviction — this is a material framework shortfall for a first look at an unfamiliar market.
  • ⚠️ % off 52w high is measured in USD and therefore embeds the yen's −8.41%. Local-currency drawdowns are ~6–8 points shallower throughout.

Screen run 2026-07-29. Prices live as of 2026-07-29/30. USD/JPY 163.5. Related: screen-global-vs-us-value-2026-07-29.md · screen-tech-crash-hidden-value-2026-07-29.md · compare-ASML-KLAC-LRCX-2026-07-15.md

Sources: Japan record buybacks ¥16.2T Jan–May 2026 · Japanese Corporate Governance Code reforms · Cross-shareholding unwind through 2030 / Honda stake sales · TOPIX record high 4,101.96, July 6 2026 · Nikkei 225 2026 levels · USD/JPY breaks 163, 4-decade low, July 2026 · Yen outlook / BOJ vs Fed 2026 · Tokyo Electron valuation 59.2x vs 30.2x industry · SEMI: record $139B equipment sales 2026 · SEMI: $156B in 2027 · Silicon wafer oligopoly ~85% of 300mm; 2026 shipment upgrades · Shin-Etsu ¥250B buyback authorisation, Apr 28 2026 · Shin-Etsu ¥197.3B ToSTNeT-3 repurchase, July 30 2026 · TDK FY2026 record results, nearline HDD heads +50% guide · TDK FY2026 full-year briefing (PDF) · Ibiden ¥500B AI substrate capex · Kyocera: Oasis proposals to remove chairman · Kyocera KDDI stake trim ¥124.8bn · Nidec ¥160.7bn accounting scandal, results delayed · Nidec third-party committee / TSE special alert · Nidec $1.6bn charges