Tsubaki Nakashima Co., Ltd. (6464): A Franchise the Bearing Majors Trust, on a Balance Sheet the Lenders Own

Verdicts

Lens	Verdict	Buy below	Most load-bearing items
Buffett	too-hard	null	B61 survive-the-wait leverage — ¥60.2bn reclassified to current, covenant bars any operating loss; B43/B52 volatile earnings + serial impairments; NCAV < 0
Munger	too-hard	null	M40/M42 PE roll-up — ¥15bn goodwill impaired, ¥20.9bn residual goodwill = 56% of equity; M2/M17 covenant survivability on lender goodwill; M18 no moat / negative pricing
Pabrai	pass	null	P1 stressed floor ≤ 0 — pessimistic haircut nets negative equity; P20 covenant-breach-and-waived "Delta Financial fingerprint"; P17 not paid to wait
Li Lu	too-hard	null	L20 can't survive without refinancing at the market's mercy; L1/L3 knowledge bar fails on undefinable trough + losing price war; L21/L35 retained-earnings deficit, book half goodwill
Claude	too-hard	implied ¥15	C46 maturity runway unknowable — MUFG ¥41.5bn due 2026-11-30 > cash; C38/C39 no floor, EPV/haircut NAV negative; C34 priced for full turnaround success

The business

Tsubaki Nakashima makes precision balls — the tiny, exacting steel spheres that sit inside a ball bearing, plus ceramic, tungsten-carbide and glass variants, in over 20,000 types across materials and sizes E2E5. Steel balls are the core product, "used chiefly as a key component in ball bearings, ensuring the quality and reliability of end products such as automobiles and machine tools," and they turn up in the rotating parts of cars, motorcycles, appliances and general machinery E3. Ceramic balls are the designated growth product — lighter, harder, heat- and corrosion-resistant — aimed at wind power, electric vehicles and, notably, semiconductor manufacturing equipment E4. The company's own framing is a 90-year heritage in "precision machining" and a purpose "to move the world with the power of precision machining" E10; its stated edge is short delivery lead times drawn from a deep, wide inventory of ball types E2.

Who pays: the world's bearing majors, above all. AB SKF took ¥14,217M (20.4% of revenue) and Schaeffler ¥7,210M (10.3%) in FY2025 E28, with NTN a third customer that has historically crossed the 10% line F254. Concentration cuts both ways — the filings flag that "a high proportion of the Group's products are sold to a relatively small number of large manufacturers" and that bearing-makers dominate the precision-ball customer base E33. The end-markets are cyclical autos, machine tools and semiconductor equipment E32.

The group runs two reportable segments: Precision Components (98.7% of FY2025 revenue E2) and a small Blower / Real-Estate stub (1.3%, medium and large blowers E6). Its footprint is genuinely global — the parent in Nara plus 20 overseas consolidated subsidiaries manufacturing in the US, Italy, Poland, Slovakia, the Netherlands, Bosnia-Herzegovina, the UK, China, Thailand and India E1E9. In 2025 the company completed a portfolio move: it carved out its ball-screw and ball-way business (including TN Taiwan) into TN Linear Motion Co., Ltd. and transferred it to MinebeaMitsumi, completed 2025-10-03, now reported as discontinued operations E24E25. The buyer is not a related party E63.

The numbers

The five-year operating line is violently volatile: +5,816 / −9,065 / +5,018 / +814 / −22,336 (FY2021 → FY2025) [F6–F10] — two large operating-loss years within five, the most recent catastrophic. FY2025 revenue was ¥69,837M, down 8.0% year on year on European auto weakness and price-driven market-share loss F5E17. The FY2025 loss attributable to owners was −¥27,214M F20, driven by a ¥16,696M impairment (goodwill ¥15,049M booked in other expenses F255, plus ¥1,647M of PP&E on idle/disposal assets in Poland and Thailand F258) and a ¥6,516M inventory write-down F269 — itself ¥4,548M above the ¥1,968M prior-year charge F268. Gross margin collapsed to 4.65% from 13.85% D4D5; FY2025 EBITDA on the company's own definition was negative −¥1,854M F328.

The balance sheet is the story. Equity attributable to owners crashed from ¥61,472M to ¥37,035M [F26→F30] in a single year, an equity ratio of 24.4% F40, and non-controlling interests fell to zero so total equity equals owners' equity F161F163. Retained earnings went to a −¥12,036M deficit F153. Total interest-bearing borrowings are ¥92,844M F302 against ¥34,633M of cash F104, leaving net debt of ¥58,211M D1 — about 3.8× the ¥15.26bn market cap. And ¥20.9bn of acquisition goodwill still sits on the just-impaired Precision Components CGU F261 — roughly 56% of the ¥37,035M equity F30.

At year-end ¥60,238M of borrowings breached financial covenants — a "no operating loss in any trailing-12-month measurement" test and a "net assets ≥ 75% of six months prior" test E15 — and was reclassified from non-current into current liabilities (current bonds and borrowings jumped ¥11,356M → ¥71,995M; non-current fell ¥81,294M → ¥20,849M) [F130→F131][F139→F140]F294. The company obtained written waivers of the acceleration right from all relevant lenders and judged no material uncertainty about going concern exists — a 継続企業の前提に関する重要事象等 (material-events) disclosure, but not a formal going-concern-doubt note E13E14. The largest single facility is a MUFG-agent, unsecured term loan of ¥41,522M maturing 2026-11-30 E38F296 — larger than the ¥34,633M of cash on hand F104.

Book value per share is ¥968.15 F50; at ¥399 that is P/B ≈ 0.41× (399 ÷ 968.15). The dividend was cut to zero for FY2025 (from ¥25.00) and is planned at ¥0.00 for FY2026 F92F351E59. Deferred tax assets were written down ¥3,097M → ¥646M F122F123.

Two facts point the other way, and they are why the discount exists. FY2025 operating cash flow was ¥10,519M — the highest in five years F65F201, up from ¥4,873M F64, against capex of only ¥1,872M F211; and Q1 FY2026 returned to profit — operating profit ¥1,127M, net profit attributable to owners ¥308M (versus a ¥558M loss the prior-year quarter) F330F332F318. Both carry heavy caveats developed below: the record OCF was flattered by a ¥10,352M one-off inventory drawdown [F107→F108], and Q1's operating profit includes a ¥1,041M TN Georgia fixed-asset disposal gain, leaving underlying operating profit at roughly ¥86M on revenue still down 2.7% F335F329E66.

The five lenses

Buffett — too-hard

Let me tell you what this company sells before I say a word about the stock. Tsubaki Nakashima makes tiny steel and ceramic balls — the sort that go inside ball bearings, which go inside the axle of your car, the spindle of a lathe, the turbine of a windmill. Twenty thousand types E2, ninety years at it E10, sold to the biggest bearing companies on the planet — SKF takes about twenty cents of every revenue dollar, Schaeffler another ten E28. Can I explain in one paragraph what it sells, who pays, and why they keep paying? On a summary basis, yes — precision balls are mission-critical, qualified in by exacting customers, and the product has been essentially the same for decades E3. So I pass B1.

But I must be honest about the boundaries of my circle. The company's own filings say ceramic-ball execution "may not proceed as anticipated," Chinese makers are undercutting on price, and the EV wave that was supposed to lift ceramics "has not occurred at the anticipated pace" E34E35E19. Chinese and Indian entrants are expected to intensify the price war for five more years E36. That is not a franchise I can evaluate with confidence over a decade.

More immediately, read the filings plainly. The five-year operating record — +5,816 / −9,065 / +5,018 / +814 / −22,336 [F6–F10] — is the opposite of demonstrated consistent earning power (B43): two devastating loss years in five. The FY2025 loss came from a ¥15,049M goodwill impairment F255, a ¥6,516M inventory write-down F269 and ¥1,647M of PP&E impairment F258 — special charges that, twice in five years and on the same CGU, are beginning to look recurring, not one-time (B52). The balance sheet tells the rest: net debt ¥58,211M D1 against ¥37,035M of equity F30, the equity itself down from ¥61,472M a year earlier F29. And here is the dagger on the steering wheel: ¥60,238M of borrowings — more than the entire market cap — breached covenants and was reclassified to current F294E14, surviving only on written waivers from every lender E13. The covenants bar any trailing-12-month operating loss E15; the MUFG facility of ¥41,522M matures this year E38. B61 — survive-the-wait — is the hard stop. One more loss year, or a slow refinancing, and this is a liquidity event with no earnings cushion. There is no asset margin either: NCAV is deeply negative (current assets ¥88,091M minus all liabilities ¥114,623M = −¥26,532M F113F148), and a Dempster-style haircut nets a liquidation deficit [B89][B90].

This is a distressed-debt situation dressed in the clothes of a franchise. To price it I would need to be a distressed-credit analyst, and that is not my game.

What a student should take from this: a business can have a real franchise and still sit outside the circle — not because the business is bad, but because the financial structure has moved the primary risk from "will this earn money" to "will these lenders be repaid." When the dominant question is a lender's question, not an owner's, the Buffett circle does not help you, and a 0.41× book multiple is not a margin of safety when the book was just impaired and NCAV is negative. Verdict: too-hard.

Munger — too-hard

Invert first, always. The question is not how a precision-ball maker at 0.41× book makes me money — it is how it kills me. And this one hands me the murder weapon in its own filings.

Start with what it is: ninety years of grinding steel spheres E10, 20,000 types E2, sold to SKF (20.4%) and Schaeffler (10.3%) E28 — roughly 31% of the top line riding on two bearing customers who buy a commodity input and squeeze it every year. SKF's take already slid from ¥17,352M to ¥14,217M F250F251. Name the moat's mechanism (M18) and you can't. There is no pricing power here; the MD&A tells you the opposite — European auto weakness plus market-share loss to price competition, and in ceramics a knife-fight with Chinese makers E17E35. Gross margin fell 13.85% → 4.65% D5D4. A moat lets you raise price through a downcycle; this company lowered volume AND lost share AND cut price, all at once.

Now the kill paths (M1), and the ugly part is that they are already in motion. The price war in the one "growth" product where the EV wave did not arrive E19E35 — that is the textile-loom lesson in real time (everyone installs the same equipment, the savings leak to customers). Customer concentration cutting the wrong way. And, decisive, the balance sheet: net debt ¥58,211M D1 against a ¥15.3bn market cap — 3.8×; total interest-bearing debt ¥92,844M F302; equity ground from ¥53,335M to ¥37,035M F26F30; retained earnings a −¥12,036M deficit F153. At year-end the company breached the net-assets and no-operating-loss covenants on ¥60,238M F294E15, got written waivers from every lender E14, and faces a ¥41,522M MUFG maturity on 2026-11-30 with the refinancing conversation still open E38E13. Run M17: if the quote halved, could company and holder sit still? No — the lenders hold an annually re-tested acceleration trigger tied to reporting an operating loss E15, on a company that has printed two operating-loss years in five [F6–F10].

I will weigh the other side fairly. The FY2025 loss is largely non-cash — operating cash flow was +¥10,519M F65, best in five years, because the −¥22,336M loss was stuffed with a ¥16,696M impairment F209 and a ¥6,516M write-down F269. Ex-KKR CEO E52, activist Arcus at 5% E44, the clean Minebea carve-out E24, a genuinely clean governance shell E46E47E51E62. Q1 even printed a profit F330. It is not worthless. But it is too hard, for three reasons, each a checklist item. M40 (raisins and turds): to underwrite this at 0.41× book I must know which of ~20 overseas children is the wonderful core and which are value-destroyers — the ledger shows TN Tennessee with negative equity of −¥10,150M and a ¥3,878M loss E7 — and I cannot separate the segment capital, because the Precision Components segment is one 98.7% lump that just lost ¥22,501M F225E2. M42/M53 (roll-up allocation): the ¥15,049M goodwill just impaired F255 is the tuition bill of a PE roll-up, and ¥20,898M of goodwill — 56% of equity F261F30 — STILL sits on the CGU that just failed its own test (FY2024 headroom was a thin ¥6,224M F264), tested at a 10.6% discount rate and 2.3% terminal growth I have no way to defend F266F267. M18/M26 (no moat, negative pricing): the mechanism that would make cheapness matter is simply absent. The mid-term plan wants ¥87bn revenue and ¥10bn operating profit by 2029 E11 — a wish wearing a spreadsheet, off a base of two loss years in five. Cheapness is not the hurdle; the best alternative is, and the best alternative to a leveraged distressed roll-up with more goodwill to burn is almost anything, including cash. Too hard. Next.

What a student should take from this: cheapness against book is the bait, not the thesis — 0.41× means nothing when a fifth of that book vanished in one year and ¥20.9bn of impairment-prone goodwill (56% of equity) still sits on the failing segment F261F30. The decisive question in a leveraged turnaround is never "how cheap," it is "can it survive its own covenants," and here the lenders hold a trigger that fires on a single operating-loss year E15. When you cannot separate the wonderful core from the value-destroyers inside a 20-subsidiary roll-up (M40) and the numbers you'd need are not in the record, the honest move is the too-hard basket. Verdict: too-hard.

Pabrai — pass

Let me tell you what I'm looking at, and then why I'm walking away — because the walking-away is the whole lesson. Tsubaki Nakashima makes precision balls — over 20,000 kinds, in steel, ceramic, tungsten carbide, glass E2E5 — the guts of ball bearings, 98.7% of revenue E2, sold to SKF at 20.4% and Schaeffler at 10.3% E28. Boring, slow-changing industrial at 0.41× book. A ball is a ball. Heads I win, tails I don't lose much, right? Wrong. And I want to show you the arithmetic, because this is where Dhandho earns its keep.

The first question is never "how cheap." It is P1: what is the realistic worst case, and what fraction of my money does it destroy? I do not read the floor off the equity line; I build it from marked-down assets minus every liability. So watch. Total assets ¥151,658M F127, total liabilities ¥114,623M F148, equity ¥37,035M F163 — and because minorities went to zero F161, that IS the book the 0.41× is measured against. Now I mark it like a pessimist. Cash ¥34,633M F104 whole. Receivables ¥18,587M F106 haircut ~15%. Inventory ¥25,726M F108 — and they just wrote it down ¥6,516M and are scrapping US and ceramic stock E18F269 — cut in half. Goodwill ¥21,277M F119 to zero; reality already agrees, because they impaired ¥15,049M of it this very year F255E21. PP&E ¥36,224M F115 — specialized ball-grinding machinery, some just written down as idle F258E22 — a distressed sale fetches maybe 40 cents. Add it up: ~¥83bn of gross asset value against ¥114.6bn of liabilities. The stressed equity is negative — roughly minus ¥31bn. There is no floor.

And that isn't even a stress case — it's the printed balance sheet. Net debt ¥58,211M D1 is 3.8× the market cap. When you buy this at ¥399 you are not buying a cheap asset; you are buying the thin, subordinated tranche of a leveraged balance sheet and calling it a bargain because the residual happens to be positive on the accounting page today. Now leverage — P20, the single item that has cost investors the most money, my own Horsehead scar included. This isn't leverage that might bite; it already bit. At year-end the company breached its financial covenants — a no-operating-loss test and a net-assets-≥75% test E15 — on ¥60,238M, yanked into current liabilities F287F294; current bonds-and-borrowings went ¥11,356M → ¥71,995M in one year F131. It survives only because every lender signed a written waiver of the acceleration right E14, while negotiating a refinance of the ¥41,522M MUFG term loan due 2026-11-30 E38. The company itself files this as a 重要事象 material-events note — one notch below a formal going-concern doubt E13. This is precisely the Delta Financial fingerprint P20 and P23 warn about: solvent on paper, alive only at a lender's goodwill. The dividend is zero for FY2025 and planned zero for FY2026 E59F351 — P17 fails: value is not accruing to me while I wait; it's being conserved to appease banks.

Yes, there are turnaround tells I respect — ex-KKR CEO E52, Arcus at 5.03% E44, the Minebea sale for ¥2,048M of delevering proceeds F283E26, Q1 back in the black F330. But that Q1 profit of ¥1,127M includes a ¥1,041M one-time disposal gain F335 — underlying operating profit about ¥86M. That's a story, and stories are what P50 and P52 exist to reject. There is no version of "tails, I don't lose much" here. This is a pass, and it isn't close.

What a student should take from this: an equity cushion is not an asset floor — write that on your hand. When net debt is 3.8× market cap D1 and the "book" you're buying at 0.41× is a ¥37bn residual squeezed between ¥152bn of assets and ¥115bn of liabilities F127F148F163, a modest haircut turns the floor negative and you own the most junior slice of a distressed capital structure. Leverage plus a covenant breach met only by lender waivers (P20/P23) sends it to pass however low the multiple, because upside never rescues a bad downside. Verdict: pass.

Li Lu — too-hard

Let me begin where I always begin: not with the price, but with whether I could honestly claim to know this business's next ten years better than the people who already own it. If I cannot, it does not matter that it trades at four-tenths of book. A cheap thing I cannot predict is not an opportunity; it is a temptation.

The business is, in one sense, easy to admire — ninety years of precision machining E10, 20,000 ball types E2E5, the humble but essential heart of a bearing E3, sold to the great bearing houses (SKF ¥14,217M, Schaeffler another ten percent E28). When I look for the single factor that decides where value accrues (L14), I find it here: it is the ball's quality-and-cost position against a rising tide of Chinese and Indian makers E35E36. And on that decisive factor the filings tell me the company is losing — management writes that revenue fell on "market-share decline caused by price competition" E17, and in ceramics the Chinese makers have intensified a price war E35. This is not a moat holding under attack (L38); it is a moat being crossed.

Now the knowledge bar, which gates everything (L1). Three variables decide the ten-year outcome: the auto/industrial cycle E32, pricing power against low-cost entrants E34E36, and — uniquely, urgently — whether the balance sheet survives at all. Two are unknowable a decade forward, and the third is a coin I will not call. The operating line reads +5,816 / −9,065 / +5,018 / +814 / −22,336 [F6–F10]; ROE ran +7.3 / −17.6 / −2.5 / +1.6 / −55.3 [F41–F45]. What is the worst case — trough revenue, trough margin, an earnings floor I can defend (L2, L3)? This company has no stable trough; its downturns gap through the floor. But the item that ends the analysis is L20 — can it survive a severe downturn without issuing equity or refinancing at the market's mercy? The answer is a plain no, and it is not close. Net debt ¥58,211M D1 is ~3.8× the market value of the equity; ¥60,238M of borrowings breached covenants and was reclassified to current F294E14; the company is a going concern today for one reason only — every lender signed a written waiver E14 — and the MUFG term loan of ¥41,522M comes due 2026-11-30 E38. That the disclosure is a "material events" note rather than a formal going-concern doubt E13 is a legal nicety; the economic fact is that outside parties, not the operating business, presently decide whether the equity lives.

Trace the retained yen (L21). Capital went into acquisitions that became goodwill — ¥36,274M F118 — of which ¥15,049M was written off this year F255, leaving ¥20,898M still inside a ¥37,035M equity base F261F30; more than half the remaining book is goodwill from deals now impairing. Retained earnings went negative, ¥15,486M → −¥12,036M F152F153; one subsidiary, TN Tennessee, carries −¥10,150M equity E7. Book value per share melts rather than compounds (L35): ¥1,315 → ¥1,261 → ¥1,362 → ¥1,543 → ¥968 [F46–F50], and the ¥968 is inflated by residual goodwill. I will be fair to what is genuinely sound: the structure is clean — no controlling shareholder, no parent E51, trivial cross-holdings E58, no related-party transactions E62, a three-committee board with four independents E46E47, an activist at 5% E44, PE-schooled management running a real restructuring E24. So my too-hard is not a governance verdict. It is the knowledge bar and the leverage. Q1 showed a return to profit — but ¥1,041M of the ¥1,127M was a one-off disposal gain F335, underlying about ¥86M, and revenue still fell E66. That is a sale of the furniture, not durable earning power.

What a student should take from this: cheapness is the last question, never the first — and it cannot rescue a business you cannot predict or one whose survival is decided by its lenders rather than its operations. When retained earnings have gone negative and half of remaining book is goodwill from deals now impairing, you are not looking at a dollar selling for fifty cents; you are looking at a melting asset whose true book is unknowable. And distinguish an operational recovery from a sale of assets: a "profit" that is a disposal gain against still-falling revenue is furniture being sold, not a franchise rebuilt. Verdict: too-hard.

Claude — too-hard, implied buy below ¥15

I registered, figures-blind, a below-median prior and named the hypotheses that would have to be confirmed to lift it: a real maturity runway (H1), a cash-generative core at trough (H2), a cleansing rather than serial impairment (H3), and an equity not about to be diluted by its own rescue (H5). The ledger confirms none cleanly, overturns one in the constructive direction, and leaves the load-bearing one genuinely unknowable.

Start with the arithmetic that decides everything, because it is unusual. Net debt is ¥58,211M D1 against a market cap of ¥15,263M — 3.8×. The disciplined move (C33, C38) is to value the whole enterprise downside-first and net the debt last. Capitalize a defensible normalized operating profit at a fair-to-generous 6× EV/EBIT for a low-return cyclical: at the FY2026 forecast OP of ¥2,500M F346 the enterprise is worth ~¥15,000M and the equity is worth roughly −¥43bn; it does not turn positive until normalized OP approaches ¥10,000M — precisely the FY2029 mid-term-plan target E11, a 4× rise. Put the other way, at ¥399 the market is capitalizing more operating profit than the plan's terminal-year goal. The equity is not cheap at 0.41× book F50; it is a leveraged call option on a turnaround, priced as if the option is already substantially in the money (C34).

Now the reverse-checks. H2 (cash-generative at trough) partially confirms, and it is the real news — stripping the ¥16,696M impairment F259 and the ¥4,548M incremental inventory write-down [F269 vs F268] from the −¥22,336M loss F10 leaves a clean core OP of roughly −¥1,092M; the collapse was writedown-driven, not a cash-operating implosion. But "the core loses ~¥1bn at trough" is not "the core earns its keep," and the reported OCF of ¥10,519M F65 is an illusion for underwriting — ¥10,352M is a one-off inventory release [F107→F108] — so ex-working-capital, operating cash was roughly negative. H3 (cleansing impairment) I cannot confirm — ¥20,898M of goodwill still sits on the same CGU that just failed F261, 56% of a ¥37,035M equity F30, tested at 10.6% / 2.3% F266F267, with the risk factor itself saying a further decline drives more impairment E37; a second write-down is the base case, not a tail. H5 (dilution) — the convertible is not the threat: strike ¥676 E41, out of the money at ¥399; the live risk is the refinancing carrying an equity kicker, which routes entirely through H1.

H1 — the maturity runway — is the unknowable that forces too-hard (C46). ¥60,238M breached covenants and was reclassified to current F294E14E15; all lenders granted written waivers E13E14 — but waivers are a decision the lenders can revisit at the next test date, not a cure. The concrete wall is the ¥41,522M MUFG term loan maturing 2026-11-30 E38F296; cash of ¥34,633M F104 does not cover it. The filings disclose that refinancing discussions have commenced E13 — but disclose no committed undrawn facilities, no signed refinancing, no terms. Whether the runway is covered is not resolvable from the archive; it is a lender decision I cannot observe. And the downside floor answers the standing question: haircut liquidation NAV (C39) is −¥678/share, staying negative from PP&E at 40% to 80%; EPV of equity (C38) is −¥40bn to −¥56bn at every defensible EBIT. There is no asset floor; a stressed haircut takes the equity through zero. Arcus at 5.03% E44 and a professional ex-KKR board [E46–E52], plus the completed Minebea deleveraging F283E26, are why this is not a pass — but an activist and a good board do not change the arithmetic that the equity is positive only if the plan largely lands.

Why too-hard and not pass? Pass would assert the short wins at any price; the implied buy-below arithmetic shows a price at which the downside owner-yield could clear a hurdle if the Nissan-style decisive variable — here the refinancing — were knowable. It is not. I publish an implied buy-below of ¥15 (C44) — the probability-weighted going-concern equity, each scenario floored at zero — so the record is falsifiable, and I hold it loosely: its honest content is "far below the tape; effectively no floor," not a price to the yen.

What a student should take from this: when net debt is a large multiple of market cap, the equity is a call option and the buy-below must be computed downside-first — capitalize a defensible operating profit, subtract the debt last, and check whether the equity is positive before the base case; if it isn't, the price is paying for the plan. And distinguish "the loss was non-cash" (often true and comforting) from "the operator earns its keep at trough" (a separate, harder claim); a covenant waiver buys time, not solvency, so when the refinancing terms aren't disclosed, the correct output is too-hard, not an optimistic estimate. Verdict: too-hard (implied buy below ¥15).

Synthesis

Where the five lenses agree

The facts are not in dispute, and all five lenses read them the same way. Tsubaki Nakashima is a genuine niche precision-ball franchise — 20,000+ ball types E2, ninety years of precision machining E10, the qualified-in supplier the bearing majors keep buying from (SKF ¥14,217M, Schaeffler ¥7,210M E28) — sitting on an acutely distressed, over-leveraged balance sheet. The distress is specific and unanimous: a covenant breach on ¥60,238M F294E14 met not with a cure but with written lender waivers E13; net debt of ¥58,211M, ~3.8× market cap D1; a maturity wall — the unsecured MUFG term loan of ¥41,522M due 2026-11-30, larger than the ¥34,633M of cash E38F104; a book that is ~half acquisition goodwill on a just-impaired CGU (¥20,898M residual, 56% of ¥37,035M equity F261F30); a violently volatile operating record (+5,816 / −9,065 / +5,018 / +814 / −22,336 [F6–F10]); and a dividend suspended E59F351. The near-unanimous read: survival is a lender/refinancing decision that cannot be underwritten from the filings. Four lenses go too-hard (Buffett, Munger, Li Lu, Claude); Pabrai passes because the stressed downside has no floor. There is no bull among the five.

Where the lenses diverge

The divergence is thin. It is not too-hard-versus-buy; it is too-hard versus pass — a distinction about whether you cannot judge it or can judge it and decline.

Buffett puts it as a survival failure he cannot price: "B61 — survive-the-wait — is the hard stop … the covenants bar any trailing-12-month operating loss E15; the MUFG facility matures this year E38 … I would need to be a distressed-credit analyst." Munger arrives from inversion and lands on the same wall: "run M17 — if the quote halved, could company and holder sit still? No — the lenders hold an annually re-tested acceleration trigger tied to reporting an operating loss E15." Li Lu gates on the knowledge bar and the leverage together: "L20 — can it survive a severe downturn without refinancing at the market's mercy? The answer is a plain no … outside parties, not the operating business, presently decide whether the equity lives." Claude names the variable and prices the escape hatch shut: "C46 … whether the runway is covered is not resolvable from the archive; it is a lender decision I cannot observe," and the equity is "a leveraged call option … priced as if the option is already substantially in the money."

Pabrai does not disagree with any of that — he simply reaches a decision rather than a suspension. For him the survival question resolves the moment you build the floor: "a pessimistic haircut … nets to negative stressed equity (≈ −¥31bn, marking ¥151,658M of assets F127 down against ¥114,623M of liabilities F148) … there is no floor" (P1), and the covenant is not "unknowable," it is the recognized Delta Financial fingerprint — "solvent on paper, alive only at a lender's goodwill" (P20) — and you are "not paid to wait" while cash is conserved for banks (P17). He can judge it well enough to say no without reaching too-hard. The whole of the disagreement: can't-know-if-it-survives (Buffett B61, Munger M17, Li Lu L20, Claude C46) versus know-there's-no-floor (Pabrai P1/P20). On the ledger both routes end at "no action."

The red team's challenge (and our answer)

Because four of five lenses agreed on too-hard, the consensus was put to a dedicated adversary (red-team.md), and its case is the record's second-strongest bull case — strong enough to state plainly and by name.

The red team's four strongest points. (1) The catastrophe was almost entirely non-cash — and the same year threw off record operating cash. The −¥22,336M operating loss F177 contains a ¥16,696M impairment F259 (of which ¥15,049M is pure-accounting goodwill F255) plus a ¥6,516M inventory write-down F269; strip both and the underlying operating line is positive, +¥876M [F177 + F259 + F269]. Meanwhile FY2025 operating cash flow was ¥10,519M — the highest in five years F65, for free cash flow of ≈¥8,647M [F201 − F211] against a ¥15.26bn market cap: "the equity is priced at under 2× the FCF the business produced in its worst reported year." (2) A genuine, hard-to-replicate franchise — the Precision Components business is 98.7% of revenue E2, selling over 20,000 types of quality-critical balls E2 to bearing majors who "do not casually re-source a component that determines the noise and life of their own products," and "that the largest customer … still directs ¥14bn a year through this supplier E28 is the strongest possible evidence the franchise is intact." So the goodwill impairment is "an accounting artifact, not an economic verdict." (3) The distress is being actively, credibly, professionally managed — a relationship-bank roll, not a cliff: all lenders granted written waivers on unsecured facilities E14E13F296 ("an unsecured lender that re-affirms an unsecured position is voting for the going concern with the strongest instrument it has"); the Minebea divestiture brought ¥2,048M and removed a loss-maker E24E26; ex-KKR operators run it under independent-majority governance E52E47; activist Arcus holds 5.03% E44 and the convertible is 41% out of the money at ¥676 E41, so dilution is not imminent. (4) A cleansed book and asymmetry — every yen of the most impairment-prone asset written off "cannot impair again," and "if even half the mid-term plan lands (¥5bn OP)," a business generating ¥8bn+ of FCF "does not trade at 0.41× book," a 3–5× re-rate. The red team's verdict: WATCH, buy-below ¥340 (≈0.35× the post-impairment book of ¥968.15 F50), "escalating to an outright buy the moment the FY2026 refinancing is termed out and a second consecutive profitable quarter confirms the operational turn."

Our answer — conceding what is right. The red team's central facts are real, and they are exactly why the discount exists — we do not wave them away. The non-cash loss and the positive OCF are genuine: FY2025 was an impairment year, not a cash-operating implosion, and Claude's own forensic read (H2) confirmed the clean core was near breakeven, not collapsing. The franchise is genuine: every lens explicitly credited the SKF/Schaeffler relationship E28 and the 20,000-type inventory position E2 — no lens declined on terminal-value or franchise-death grounds; Li Lu and Buffett both said so in as many words. The "franchise + non-cash loss + positive OCF" trio is precisely the reason a real business trades at 0.41× book. Where the consensus holds anyway — and where it is close — turns on four things the red team's own honest self-audit half-concedes:

• The record OCF was flattered by a one-off inventory drawdown. ¥10,352M of the ¥10,519M OCF is a working-capital release [F107→F108] — much of it the very inventory that was written down E18 — that does not repeat; ex-release, operating cash was roughly negative. This is Claude's C16/C47 and it is the direct answer to "under 2× FCF": the sustainable FCF is materially lower than the headline ¥8.6bn, and the red team's own vulnerability #3 admits "the record ¥10,519M OCF was helped by a ¥10,352M inventory drawdown … working-capital release that does not repeat."
• The clean core is thin, and revenue is still falling. Underlying operating profit ex-impairment is only ~+¥876M on ¥69,837M of revenue F165 — a fraction of one percent — and even the red team's own vulnerability #2 concedes "normalized OP is only ~¥0.9–2.9bn … a low-single-digit margin … the ¥10bn OP mid-term target implies a margin expansion this business has not demonstrated." Revenue fell 8.0% in FY2025 E17 and 2.7% in Q1 despite a yen tailwind E66; the top line has not turned. Q1's ¥1,127M is ~92% a disposal gain F335. This is Munger's M18/M26 and Buffett's B43.
• Cash does not cover the wall, and the waivers are revocable. ¥34,633M of cash F104 does not cover the ¥41,522M MUFG maturity E38; the waivers are point-in-time and do not bind future periods, and the same "no operating loss" test E15 must be re-passed every reporting date — a single further operating-loss year re-arms the covenant. The red team concedes this exactly: "the waivers are point-in-time … a second bad year re-arms the trap." This is Claude's C46 and Li Lu's L20: the refinancing is not merely "a normal relationship roll" you can assume — it is the undisclosed term sheet on which the entire equity rests.
• Net debt ~3.8× cap means a stressed haircut takes the equity negative. Pabrai's P1 and Claude's C39 both compute it: goodwill to zero (as reality already did F255), inventory halved amid active scrapping F269, PP&E at a distressed 40% — and the residual equity goes through zero to ≈ −¥31bn (Pabrai) / −¥678/share (Claude). The red team's vulnerability #1 concedes "a debt stub ~3.8× cap that can wipe the equity if a second loss year re-arms the covenant."

So the honest reconciliation is that the consensus and the adversary agree about the facts — the loss was non-cash, the OCF was real but one-off-flattered, the franchise is intact, the distress is professionally managed — and split only on the weight. The red team judges the franchise-plus-management-plus-executed-waivers sufficient to make survival "observable" and demand a wider margin (¥340, ≈0.35× the post-impairment book of ¥968.15 F50) rather than decline; the four too-hard lenses judge the undisclosed refinancing term sheet — cash short of the maturity, waivers revocable, one loss year from re-breach — sufficient to withhold a price entirely, because survival remains a lender decision (Buffett B61 / Munger M17 / Li Lu L20 / Claude C46) with no floor for the residual equity (Pabrai P1). It is genuinely close. But the record ships both priced thresholds (below) so a future review can score which reading was right.

Self-distance note

The Claude lens holds one of the five verdicts compared above and also wrote this synthesis; the reconciled figure table all five lenses consumed, and the evidence ledger, were built by a (Claude-driven) dual-blind extraction (two independent passes per ledger, reconciled against page-delimited source text). The consensus red team that challenged the four too-hard verdicts is likewise Claude-authored. Read the synthesis — and the by-name engagement above — with that concentration of authorship in mind: the adversary and the answerer share a model.

Prediction-vs-actual

VOID. This was an autonomous headless cycle; the human blind prediction is voided (void: no-human-prediction, never forged). No prediction-vs-actual scoring applies to this study.

Verdict accounting (fixed ex-ante)

• A buy-below-¥X verdict is price-falsifiable against the unadjusted stamp.
• pass / watch / too-hard are recorded but unscored in any future review.
• The original verdict counts at its original stamp regardless of later corrections.
• On a stock split, reverse split, or consolidation, the buy-below threshold restates mechanically by the announced ratio (corporate-action disclosure cited); the stamp itself never restates.
• Two priced thresholds are recorded for this study but were not issued as lens buy-below verdicts: the red team's falsifiable ¥340 watch-anchor and Claude's implied ¥15. Neither is a scored verdict; both are recorded as the priced thresholds a future review reads against (see "What would change our minds").

Red team

A consensus red-team ran because four of five lenses agreed on too-hard. The adversary's strongest points, verbatim-faithful and cited, with the synthesis's engagement:

• "The catastrophe was almost entirely non-cash — and the same year threw off record operating cash." Strip the ¥16,696M impairment F259 (¥15,049M pure goodwill F255) and the ¥6,516M inventory write-down F269 from the −¥22,336M operating loss F177 and the underlying line is positive, +¥876M; FY2025 OCF was ¥10,519M, the highest in five years F65, FCF ≈¥8,647M [F201 − F211] — "the equity is priced at under 2× the FCF the business produced in its worst reported year."
  • Engagement (conceded, then bounded): the non-cash loss and the positive OCF are real, and this defeats the naïve "another write-down is coming next" reflex. But ¥10,352M of the OCF is a one-off inventory release [F107→F108] — the very inventory written down E18 — so sustainable FCF is far below the headline (Claude C16/C47); the clean core is only ~+¥876M on ¥69,837M revenue F165, and the red team's own vulnerability #3 concedes the release "does not repeat."
• "This is a genuine, hard-to-replicate franchise — the reason the impairment is an accounting artifact and not an economic verdict." 98.7% of revenue E2, 20,000+ ball types E2, SKF still routing ¥14,217M / 20.4% E28 through the supplier, 90-year precision-machining heritage E10, ceramic optionality in wind/EV/semiconductor equipment E4E19.
  • Engagement (accepted): the consensus agrees — every lens credited the franchise, and no lens declined on franchise-death or terminal-value grounds. This point is correct; the decline is customer-survival-of-the-issuer via the balance sheet, not the durability of the ball business.
• "The distress is being actively, credibly, professionally managed — this is a relationship-bank roll, not a cliff." All lenders gave written waivers on unsecured facilities E14E13F296; Minebea divestiture brought ¥2,048M and removed a loss-maker E24E26; ex-KKR operators E52 under independent-majority governance E47; activist Arcus 5.03% E44; convertible 41% out of the money at ¥676 E41.
  • Engagement (conceded, then bounded): all real, and it is why this is not a mere liquidation. But the waivers are point-in-time and revocable, the "no operating loss" test E15 must be re-passed every reporting date — a single further loss year re-arms it — and ¥34,633M of cash F104 does not cover the ¥41,522M MUFG maturity E38; the refinancing term sheet is undisclosed E13. Survival remains a lender decision (Buffett B61, Munger M17, Li Lu L20, Claude C46). The red team concedes: "a second bad year re-arms the trap."
• "The impairment cleansed the book and de-risked the future … the asymmetry." Goodwill on the CGU was cut ¥35,895M → ¥20,898M F260F261; "every yen … that has now been written off is a yen that cannot impair again," and "if even half the mid-term plan lands (¥5bn OP)" a ¥8bn+-FCF business "does not trade at 0.41× book" — a 3–5× re-rate toward the FY2029 ¥10bn-OP plan E11.
  • Engagement (conceded, then bounded): the write-off does reduce recurrence risk. But ¥20,898M of goodwill still sits on the same CGU that just failed its test F261 — 56% of equity F30 — with the risk factor flagging further impairment if the decline persists E37 (Munger M42, Claude C39/H3); and net debt ~3.8× cap D1 means a stressed haircut takes the residual equity negative (Pabrai P1: ≈ −¥31bn; Claude C38/C39: −¥678/share). The asymmetry is real only above the balance sheet; below it there is no floor.

The red team's own verdict — "Watch, with a buy-below at ¥340 … escalating to an outright buy the moment the FY2026 refinancing is termed out and a second consecutive profitable quarter confirms the operational turn" — rests on the claim that the decisive variable is "unusually well-lit": executed waivers, a cash-covered relationship maturity, a knowable franchise. The consensus holds because, on the red team's own test (is the refinancing observable enough that the survival question stops mattering?), the answer is no: cash is short of the single maturity F104E38, the waivers are revocable and one loss year from re-arming E15, and the term sheet is undisclosed E13 — so the unknowable is not neutralized, and beneath the balance sheet a stressed haircut leaves no floor for the residual equity D1. The disagreement is one of weight, not fact, and it is close; the record ships both priced thresholds so it stays falsifiable.

What would change our minds

No lens issued a buy-below or watch with a falsifier: line — the four too-hard verdicts carry no price falsifier by construction, and Pabrai's pass carries none. So there is nothing of that kind to pre-register from the lenses, and we say so plainly rather than manufacture one.

What the record does carry, as the priced thresholds a future review scores against:

• The red team's falsifiable buy-below: ¥340 (≈0.35× the post-impairment book of ¥968.15 F50; ≈¥13bn market cap) — the adversary's genuine watch price, demanding "the market pay you for the refinancing risk you are underwriting," escalating to a buy once the MUFG maturity is termed out and a second consecutive profitable quarter lands. A future review can score whether the stock reached it and what followed.
• Claude's implied buy-below: ¥15 (the probability-weighted going-concern equity, each scenario floored at zero — its honest content is "far below the stamp; effectively no floor," not a price to the yen [C44]) — published for falsifiability, not issued as a verdict.

And the observable that would move the consensus off too-hard: the blocking unknown is the MUFG maturity-runway coverage (Claude C98/C46) — resolvable only by a lender decision or the passage of time, not by any observation of the company at a threshold. Concretely, the consensus would upgrade toward watch on a committed, disclosed, non-dilutive refinancing of the MUFG term loan E38 (a signed facility with terms, not merely "discussions have commenced" E13). Conversely, a rescue equity raise, a third-party allotment, or a convertible re-strike below ¥676 E41 would confirm the dilution path (H5) and hold the decline. Claude's pre-registered probabilities (in profiles/claude.md) put the MUFG loan remaining a going concern past 2026-11-30 at 0.75 but a materially dilutive recapitalization at 0.45 and the too-hard posture being correct at 0.60 through two reporting cycles — i.e. the consensus expects survival-with-possible-dilution rather than a clean term-out, and has said so on the record.

What this taught the checklists

Queued for the next study (F2 revision proposals; see docs/process/evolution.md), attributed per lens:

• Pabrai — a dedicated P20 sub-trigger for a covenant-already-breached-and-waived state, distinct from "thin headroom." A live waiver-dependent breach with a 重要事象/material-events disclosure (short of a formal going-concern note) is a specific, recognizable failure signature — the Delta Financial live-wire case — currently caught only under the general maturity-coverage language. Rationale: the waiver is revocable and the refinancing is unclosed, so a breached-and-waived structure is categorically more dangerous than mere low headroom and deserves its own trigger E13E14F294.
• Li Lu — two rationale-only notes. (1) L20 guidance for distressed-refinancing companies: when a filing discloses a lender waiver plus a near-term refinancing dependency E13E14, treat L20 as a presumptive fail unless trough-scenario coverage is shown without assuming the refinancing closes — this case is a clean fixture for that rule. (2) Compute the margin of safety against tangible (ex-goodwill) equity as the primary figure — L18 and L47 both fired on "book is half goodwill" F119, but nothing forces netting goodwill out before computing the P/B margin; a student could pass on a headline 0.41× and miss that tangible book is far thinner.
• Claude — three, each with rationale. (1) Add a "levered call-option" branch to the downside/buy-below method (C33/C44): when net debt > ~2× market cap and haircut NAV is negative, the standard floor-based buy-below degrades to a misleadingly precise low-yen figure (here ¥15) that is a floor, not the convex option value; instruct pricing the equity as a call on enterprise value struck at net debt, or declaring too-hard. (2) Promote C46 to an explicit verdict-gating item for distressed names, with a required "committed vs indicative financing" test — the whole verdict turned on whether the refinancing was committed or merely in discussion E13. (3) Sharpen C79 to flag "a metric that pays out in a loss year" — FY2025 performance pay was earned on an FCF-margin KPI in a −¥27bn loss year because the metric rewarded a working-capital release E57.
• Buffett — three Buffett-only observations. (1) B61 needs a Japan-specific sub-item on covenant-reset cadence: the covenant text tests at each fiscal-year-end AND each Q2 E15 — semi-annual testing doubles the frequency of potential acceleration events versus the standard annual framing. (2) B52 should note that IFRS goodwill impairment under IAS 36 is not reversible — unlike inventory write-downs, once the CGU is impaired F255 that goodwill cannot be restored even if the business recovers; check whether the same CGU has been impaired before. (3) B37 overrides B85 — when the only viable thesis requires trusting new management who have not yet produced the record, "management in place" (B85, passes because officers are seated) cannot compensate for "no turnaround dependency" (B37, fails on the not-yet-demonstrated execution).

---

Corrections