Futaba Corporation (6986): A Hard Floor, a Melting Business, and No Hand on the Wheel

Verdicts

Lens	Verdict	Buy below	Most load-bearing items
Buffett	too-hard	null	B2 — no honest ten-year earnings estimate; B42 — owner earnings negative (~−¥2.2bn); B65/B103 — a melting net-net with no conversion agent
Munger	pass	null	M34 — moat absent and narrowing; M37 — not great-at-a-fair-price, a mediocre business at a cheap price; M86 — negative lollapalooza
Pabrai	watch	¥380	P1/P20 — asset floor + no leverage; P7 — melting ice cube (monetization still winning); P53 — ~0.6× liquid value, not ≤0.5×
Li Lu	too-hard	null	L1 — ten-year knowledge bar fails; L35 — melting, not compounding; L21/L22 — profit manufactured from self-liquidation
Claude	watch	¥420 (implied)	C35 — whole-owner yield negative at ¥566; C39/C47 — ~2× floor, ~1.0× cash-covered; C57/C58/C62 — no forcing agent with a date

The business

Futaba Corporation is a small Japanese manufacturer — the Company plus 23 consolidated subsidiaries E1F402 — that sells two quite different kinds of things through two reportable segments.

Electronic Devices makes composite modules, industrial and hobby radio-control equipment, robotics products, and organic-EL displays E2. It is the smaller segment by revenue — ¥15,386M of external sales in FY2026/3 F304 — and it is the more visibly transitional one. Management is winding down the fluorescent-display-tube (VFD) business and the touch-sensor business, and moving organic-EL displays out of in-house production mid-stream — its own words are a "business-scheme change" E28E29E96. The growth hopes here are early and small: industrial servos tied to North-American UAV work were described as strong, and industrial drones aimed at the inspection, disaster-prevention, and (next) defense markets E28E38.

Production Equipment is the larger segment — ¥27,595M of external sales F306 — selling plate products, mold materials and parts, and molding/production-rationalization machinery into the plastics supply chains that feed autos and appliances E3. It leans heavily on one foreign market: sales to South Korea were ¥11,848M F337, better than a quarter of the whole company, and this is where cheap Chinese imports and a weak Korean auto/appliance market are doing the most damage E30.

The global footprint is real but contracting. Manufacturing and selling entities sit in China, Korea, Thailand, Vietnam, Taiwan, and Alabama E4E5E6; the US Alabama subsidiary and Kishin Seiki in Korea each exceed 10% of consolidated sales E12. But the direction of travel is retrenchment — the Singapore subsidiary completed liquidation in December 2025 E8; Korea, multiple China entities, the Philippines, and (post-balance-sheet) a Shanghai trading subsidiary have all been resolved to dissolve, suspend, or liquidate E9E96; group headcount has fallen from 4,006 to 2,384 in five years F73F77. The shape is a shrinking legacy core with a handful of unproven emerging niches — a company being taken apart and rebuilt at the same time, under a "FY2024–2026 Mid-term Management Plan" of structural reform whose own final-year target was revised down to a revenue of ¥45.0bn and an operating loss of ¥1.3bn E14E15E19.

The numbers

The two halves of Futaba's numbers point in opposite directions, and the whole thesis lives in the gap between them.

The operating business is shrinking and loss-making. Revenue fell from ¥60,326M in FY2023/3 to ¥42,982M in FY2026/3 F2F5 — down roughly 29% from the peak, and down 10.7% in the latest year alone E26. The consolidated operating result was a loss in every year the ledger records it, and the loss widened: −¥1,141M (FY2024/3), −¥1,292M (FY2025/3), −¥2,280M (FY2026/3) F11F222F223 — a −5.3% operating margin D2. Both segments lost money: Electronic Devices −¥1,322M, Production Equipment −¥959M F308F310E88. Gross profit of ¥6,923M F219 no longer covers SG&A of ¥9,204M F221. Management concedes in its own governance report that it "has not yet achieved profitability exceeding its cost of capital" E82.

The one profitable year is a costume. FY2026/3 reported positive net income attributable to owners of ¥2,522M F16F256 — but the ordinary result was a loss of ¥683M F239, and the black ink comes entirely from ¥4,226M of special gains F244: ¥3,673M gain on the sale of buildings, land, and machinery F241E90 plus ¥553M from liquidating an affiliate F242. The company says so plainly: the profit exists "owing to booking a gain on sale of fixed assets" E32. Operating cash flow of ¥1,770M F57 was likewise flattered — by a ¥2,617M inventory drawdown and asset-disposal mechanics — leaving a pre-working-capital cash subtotal of just ¥1,035M F281E34.

The balance sheet is the opposite story — deep, unlevered, net-cash. Net assets were ¥95,200M against total assets of ¥108,884M F211F165, a consolidated equity ratio of 77.0% F36 (88.3% at the parent F122). Cash and deposits were ¥36,189M F133 against nil short-term borrowings F171 and only ¥340M of non-current lease obligations F181 — net cash of ¥36,189M D1, with a ¥5.0bn Chiba Bank committed line undrawn E36 and a clean going-concern note E86. Behind the cash sit ¥17,333M of investment securities F157 (of which ¥14,724M is listed cross-holdings E74) and a ¥10,002M net defined-benefit asset F159. Book value per share is ¥1,975.55 F46; at the ¥566 stamp the stock trades at 0.29× book, and the ~¥24.0bn market cap is roughly equal to net cash alone.

And the guidance says the melt continues. For FY2027/3 the company guides to revenue of ¥45,000M F380, another operating loss of −¥1,300M F381, and a net loss of −¥3,900M F383 (EPS −¥91.95 F384) — with no asset-sale rescue built in. The FY2026/3 year-end dividend is ¥18 (¥763M total, AGM-pending) F120F395; the next-period dividend is explicitly undecided (未定) F389E47; the buyback was 492 odd-lot shares E45; and cross-holdings are only planned to be trimmed to under 10% of net assets by 2030 E21E73.

The five lenses

Buffett — too-hard

I start where I always do, with the business, not the stock. Futaba is explicable at the top level — Production Equipment sells mold parts and plate products, heavily into Korea (¥27,595M of sales, ¥11,848M to Korea) F306F337; Electronic Devices sells radio-control gear, robotics servos, composite modules and organic-EL (¥15,386M) F304E2E3. A shopkeeper could follow it. That clears the first gate, and only the first.

The second gate stops me. I ask of every business whether I can make a rough, defensible guess at what it earns ten years out. Here I cannot, because the company itself plainly cannot: it is winding down VFD and touch-sensor lines and moving organic-EL out of in-house production mid-stream E28E29, has dissolved or suspended something like seven subsidiaries in barely a year E8E9E96, took its own three-year plan and revised the final-year target down to an operating loss of ¥1.3bn E19, and guides FY2027 to a −¥3,900M net loss F383E93. Severe change and a durable estimate don't mix. That is a too-hard, decided at that item (B2).

A Graham-trained man stops caring what it earns and counts what it owns, and the count is arresting: equity to the common of ¥83,793M, ~¥1,976 a share F350F46, bought at 0.29×; net current assets after all liabilities and after the ¥11,406M minority F151F189F209 work out near ¥872 a share, two-thirds of that ~¥581, and the ¥566 price sits just under it — a formal net-net (B63/B89). Cash plus securities attributable to the common runs ~¥675 a share F133F157. On paper you buy the operating business for less than nothing.

Here is why I still won't reach. A bargain in a melting ice cube is a race, not a margin of safety. Operating losses widen every year I can see F11F222F223; the +¥2,522M net figure is a costume over ¥4,226M of special gains F244F241E32; owner earnings come to roughly negative ¥2.2bn once I add back ~¥1.0bn depreciation and take out ~¥0.9bn of maintenance capex F278F287 (B42). And for a business in secular decline, Graham and I demand a catalyst — a controlling owner, an activist, an announced liquidation — someone with the motive and the votes to convert idle assets to cash on a visible clock. Dempster only worked because Harry Bottle was sent in. Futaba has no such hand: no parent, no controlling shareholder E44E84; and management is diluting the hoard I'd be buying — cross-holdings held for "relationships," trimmed only to under a tenth of net assets by 2030 E71E73, a ¥4.2bn six-year systems rebuild ahead E95, and a buyback of 492 odd-lot shares E45. Too-hard. Cheapness is not a thesis when the ice is melting and no one has been sent to carry the blocks to shore.

What a student should take: A statistical net-net is not automatically a Graham buy — the balance-sheet discount only pays you if the assets sit still while you wait or someone with the votes converts them on a visible clock. When the operating business loses money every year, the reported profit lives entirely on asset-sale gains, and there is no controlling holder or activist to force the conversion, "cheap" collapses back into the circle-of-competence test — which a serially self-reinventing company fails.

Munger — pass

I begin with inversion, the only honest way to start: not "why own Futaba" but what must be true for this to be a catastrophic mistake — and the answer assembles itself uncomfortably fast. Both segments lost money in FY2026/3 E88; the consolidated operating loss was ¥2,280M on ¥42,982M of revenue F223F215, a −5.3% margin D2; revenue is down nearly 29% over three years F2F5; four net losses in five years F12F13F14F15F16; FY2027 guided to a −¥3,900M net loss F383. The ¥2,522M that makes the stock look profitable came entirely from ¥3,673M of fixed-asset-sale gains and ¥553M of subsidiary liquidations F241F242E32. Selling buildings is not a business model; it is a countdown.

Now the moat. A decade of data shows no mechanism that protects returns — VFD wind-down, touch-sensor wind-down, organic-EL restructuring in Electronic Devices E28E29; cheap Chinese imports, weak Korean autos, and price competition in Production Equipment E30E31. What specific mechanism — scale, switching cost, brand, license, structural cost position — protects margins against those forces? Nothing in the record names one. The textile-loom lesson applies directly: when everyone installs the same efficiencies, the benefit flows to customers, not owners.

The numbers that seem attractive I'll meet head-on. Net cash of ¥36,189M F133D1 against a ~¥24bn market cap means you nominally buy the cash at a discount and get the operating businesses for negative value — the classic cheap-for-a-reason situation, and the "free" businesses burn roughly ¥2bn a year with guidance for more F381E93. On permanent loss, the balance sheet is genuinely strong: 77% equity ratio F36, no financial debt F171, ¥36,189M cash F133, clean going-concern E86 — it cannot blow itself up. That is the one strong card. But the permanent-loss risk here is not financial, it is operational erosion — the boiling frog: revenue down 10.7% E26, headcount down 40% over four years F73F77, the mid-term plan already revised down with the growth transition "challenged" E19E20. This is a mediocre-to-poor business at what looks like a bargain, and a mediocre business at a bargain price is inferior to a great business at a fair price. The opportunity-cost hurdle is dispositive. Pass.

What a student should take: Apparent cheapness below book can persist and deepen when the business destroys operating value faster than the asset base can be liquidated. A large cash cushion does not offset an operationally loss-making business — it merely delays the reckoning. The Munger test is not whether the stock is cheap against book; it is whether there is a durable mechanism generating returns above the cost of capital. When you cannot name the mechanism and the five-year trend is relentlessly downward, move on regardless of the statistical discount.

Pabrai — watch, buy below ¥380

Let me tell you what I actually see, because on the surface it looks like a gift and my job is to work out whether it's a gift or a trap. Futaba passes my simplicity gate — it sells components and it sells the tooling other manufacturers use to stamp their own parts E2E3. What it does not pass is "is this business any good": revenue down from ¥60.3bn to ¥43.0bn F2F5, an operating loss every year I can see and widening F11F222F223, four net losses in five years F12F13F14F15, and management's own FY2027 guide to a −¥3,900M net loss on a −¥1,300M operating loss F383F381. Last year's profit was ¥3,673M of gains from selling buildings and land F241E90 plus ¥553M from winding up a subsidiary F242. That is a company selling the furniture.

So why am I not walking away? The balance sheet, and this is the whole game. ¥36,189M of cash and deposits F133 against nil interest-bearing borrowings F171 — net cash of ¥36.2bn D1 — plus ¥14,724M of listed cross-holdings E74 and a ¥10,002M pension-surplus asset F159. Crayon math: ¥566 × ~42.4m shares F95F352 is a market cap of about ¥24.0bn, and net cash alone is one and a half times that. Book is ¥1,975 a share F46; I'm paying twenty-nine cents on the book dollar. This is the Dhandho setup in its purest Japanese form — "heads I win, tails I don't lose much" — because the equity cushion here is actual money and actual listed stock, not a soft accounting number (P1). And leverage, the thing that has killed more of my ideas than anything else, is simply absent: equity ratio 77% F36, undrawn ¥5bn line E36, going-concern not applicable E86 (P20).

What stops me pounding the table is the item this verdict turns on: the melting ice cube (P7). A pile of cash in front of a cash-burning business is not automatically safe — Sears taught me that. So is the melt eating the floor faster than management can act? Right now, no — asset monetization is winning: proceeds from selling PP&E were ¥2,423M F289 and cash actually rose ¥4,670M F299, outrunning the operating burn. That's the Frontline pattern, not the Sears pattern — but it's closer to the line than I'd like and the operating-loss trend is the wrong way, with a ¥4.2bn ERP spend ahead E95 and the next dividend undecided E47. And do I get paid to wait? Partly — ¥763M of dividends E48, a cross-holding sell-down by 2029/2030 E21E73, and Brandes has held 8.8% E43 — but buybacks are literally odd-lot E45. Conservative liquid value: net cash ¥36.2bn D1 plus haircut securities less two-to-three years of melt gets me to roughly ¥900 a share of hard, liquid value before I credit the operating business at a yen. Half of that is my line. At ¥566 I'm paying right up against net cash while the business is guided to lose more, so the margin of safety on my terms isn't there yet (P53). Watch — and I want it well below where it trades, buy-below ¥380.

What a student should take: A pile of cash in front of a shrinking business is only a margin of safety if the cash is falling slower than you can collect it — the balance sheet answers "can I lose everything?" (no, here); the cash-flow statement answers "is the floor melting?", and you must read both. Net-cash-above-market-cap earns a serious look; a widening operating loss and a guided loss-making year ahead mean you demand the discount to the liquid value, not to book, and you wait for the price to come to you.

Li Lu — too-hard

I begin, as always, with the only question that matters first: can I honestly claim to understand this business's next ten years better than almost anyone who owns it? For Futaba I cannot, and the failure is instructive. It is not complicated the way a bank is complicated; it is that the business, as the filings present it, is a collection of small, declining product lines being reorganized in real time — I am asked to predict a decade of an entity actively becoming something different.

Watch the top line: ¥53,450M, ¥60,326M, ¥56,360M, ¥48,116M, ¥42,982M F1F2F3F4F5 — down ~29% from the FY2023 peak, and every single year since. A good business, if it is truly good, tends to get better; here the arrow points the other way through no single crisis but through steady erosion. On every year the ledger records a consolidated operating figure it is a loss F11F222F223, ordinary loss in four of five years F6F7F8F9F10, and five years of net income to owners sum to a cumulative loss of roughly ¥5.8bn F12F13F14F15F16 (L21). The one black-ink year is where a careless reader is trapped and 0.29× P/B starts to feel like a gift — but the company tells you the profit exists "owing to booking a gain on sale of fixed assets" E32; strip the ¥4,226M of special gains F244 and the year was an ordinary loss F239. You cannot sell your factory twice.

The melting-ice-cube test is decisive for me (L35). Is intrinsic value growing or eroding? Management supplies the answer no bull wants — it has "not yet achieved profitability exceeding its cost of capital" E82, it is winding down VFD and touch sensors and moving organic-EL out of in-house production E28E29, it has dissolved or suspended subsidiaries across Korea, China, Singapore, the Philippines, Shanghai E8E9E96, and employees have fallen 4,006 → 2,384 F73F77. The value the bull points to is entirely on the balance sheet — net cash ¥36,189M D1, net assets ¥95,200M F26, investment securities ¥17,333M of which ¥14,724M listed F157E74, a ¥10,002M pension asset F159 — the Korean hidden-asset pattern I have taught (L27). But I taught that pattern about businesses that also earned money. Here the operating business behind the cash is a loss-maker consuming that pile slowly, the "hidden assets" are substantially policy cross-holdings the company has only just resolved to unwind by 2030 E21, and a pension surplus that is not distributable owner cash. The dollar-at-fifty-cents logic requires the fifty cents to be stable — an asset base attached to a melting operating business is not a static dollar (L18). Could I state, from filings alone, this company in ten years? Honestly, no. That is the definition of too-hard, and no price rescues a business whose ten-year state I cannot predict. Pass — today it is a melting business at an asset discount.

What a student should take: A deep asset discount is only a margin of safety when the assets sit still; behind a business whose operations lose money every year and whose only profitable year came from selling its own factory, "cheap on book" is a melting ice cube with a price tag, not a dollar bought at fifty cents. And the knowledge bar gates everything — a company mid-transformation, dissolving subsidiaries and winding down product lines, is one whose ten-year state you cannot honestly predict, and no discount buys you out of that.

Claude — watch, implied buy below ¥420

I registered Futaba figures-blind as a cash-rich, below-book Japanese small-cap industrial-hardware maker with a shrinking legacy core, small growth niches, sub-cost-of-capital returns, and no controlling shareholder to force the cash out — expecting it to land below its class median on total return with a real-cash downside cushion. The ledger confirmed the floor at full strength, confirmed the melting business, and — more sharply than I priced — showed the melt widening and the forcing agent weak. That combination is what makes this a watch, not a buy and not a pass.

The floor is real, deep, and mostly reachable. At ¥566 the market cap is ~¥24.0bn; cash and deposits are ¥36,189M F133 against nil debt F171 and only ¥340M of lease F181; parent-attributable equity, excluding ¥11,406M of NCI F209, is ¥83,793M F392, a P/B of 0.29× (¥566 / ¥1,975.55 F46). Behind the cash sit ¥17,333M of investment securities F157 and a ¥10,002M pension asset F159. My haircut liquidation — receivables at 80%, inventory at 50% F147, PP&E at 70% F153, securities after-tax, pension excluded, all liabilities and the full NCI removed — still yields ~¥1,264 a share, 2.2× the stamp; deployable cash alone, after NCI, an operating buffer, and repatriation friction, is ~¥24bn — roughly 1.0× the entire market cap. The market is paying net-cash-only and assigning zero-to-negative value to the operating business, the securities, the pension asset, and ¥16,837M of PP&E F153.

The operating business is melting, and the melt is accelerating. Revenue fell 60,326 → 56,360 → 48,116 → 42,982 F2F3F4F5; the operating loss widened −1,141 → −1,292 → −2,280 F11F222F223, a −5.3% margin D2; both segments lose money F308F310. The +¥2,522M net income F16 is a trap — ordinary loss was −683 F239, and the black ink is entirely ¥4,226M of special gains F244F241F242. FY2027 is guided candidly to a −¥3,900M net loss F383. But the melt is slow relative to the pile: at a ¥2–3.9bn annual drain, cash alone funds nine to eighteen years before the securities and pension asset are even touched, and financial income of ¥1,119M F282 partly self-funds the wait. So the central case is time-destruction, not capital-destruction.

The deciding unknown — does the cash reach minorities, is there a forcing agent — resolved against the bull, though not fatally. There is no controlling shareholder and "no major shareholder" E84E44; the register is led by a Master Trust nominee (10.18%), the Futaba Electronics Memorial Foundation (7.67%), founder-linked individuals, and Chiba Bank (4.38%, also the lender and main cross-holding counterparty) E42E75. Brandes filed 8.80% as of Nov-2024 E43, but the current holding is unconfirmed. Against that, revealed behaviour is weak: the cross-holding cut is a June-2026 plan with no disposal record — securities actually rose 13,343 → 17,333 F156F157 — the buyback was 492 odd-lot shares E45, the FY2027 dividend is undecided F389, and DOE is 1.0% F391. Behaviour is accumulation, not return. A real ~2× floor that is ~1.0× cash-covered takes deep permanent loss off the table, but a floor only becomes return through a mechanism with a date, and there is none. I withhold buy-below because at ¥566 the whole-company owner yield is negative — the operating loss drains the financial income (C35) — and I will not pay for an unlock with no agent and no date. The implied buy-below is ¥420 (0.21× book), derived with the stamp nowhere in the chain. Watch.

What a student should take: A deep, reachable asset floor caps your loss but does not make you money — return requires a mechanism with a date, and "a floor with no forcing agent against a melting business" is the anatomy of a value trap, not a bargain. Price the floor as insurance (which sets a genuine buy-below well below the current price) while refusing to pay for an unlock that is plan-only, un-track-recorded, and un-compellable — and let the rate of the operating melt, not the size of the cash, decide how much time the thesis can afford.

Synthesis

Where the five lenses agree

For once, the five lenses do not disagree about the facts. They agree on three, and the divergence is entirely about what the facts mean for a minority buyer at the stamp price.

One — the discount is genuine, and the balance sheet behind it is net-cash and unlevered, with no going-concern risk. Every lens credits the same pile: net cash of ¥36,189M D1 against nil short-term borrowings F171, a consolidated equity ratio of 77.0% F36, book value per share of ¥1,975.55 F46 against a ¥566 price (0.29× book), plus ¥17,333M of investment securities F157 and a ¥10,002M pension asset F159. The going-concern note is "not applicable" E86 and a ¥5.0bn credit line is undrawn E36. Buffett computes the formal net-net (~¥581 two-thirds-NCAV, ~¥675 net financial assets per share) F151F133F157; Pabrai and Claude both find deployable cash roughly equal to the entire market cap; Munger and Li Lu concede the floor is real. Nobody thinks the company can go bankrupt.

Two — the operating business is melting. Revenue is down ~29% from the FY2023 peak F2F5; the consolidated operating result was a loss in every year on record and widened to −¥2,280M F11F222F223; both segments lose money F308F310; four of the last five years show net losses F12F13F14F15F16; and FY2027 is guided to a −¥3,900M net loss F383. All five lenses note the single load-bearing accounting fact: the only profitable year, FY2026/3's +¥2,522M F16, was manufactured entirely by ¥4,226M of special gains F244 — mostly the ¥3,673M sale of buildings and land F241E90 — over an ordinary loss of −¥683M F239. Management itself concedes it has "not yet achieved profitability exceeding its cost of capital" E82.

Three — there is no forcing agent. There is no parent and no controlling shareholder E44E84; the register is a foundation at 7.67%, a trust-bank nominee, founder-linked individuals, and the main bank E42; the activist of record, Brandes at 8.80%, is a Nov-2024 filing whose current stake could not be confirmed at the stamp E43. The revealed capital-allocation behaviour is accumulation, not return: the buyback was 492 odd-lot shares E45, the FY2027 dividend is undecided (未定) F389E47, the cross-holding sell-down is a plan dated to 2029/2030 with no disposals yet E21E73, and DOE is 1.0% F391. Nobody found a hand on the wheel.

Where the lenses diverge

The split is not about the facts; it is about whether asset value below book is reachable by a minority without a catalyst, and how fast the operating melt threatens it. Put the masters in a room.

Munger inverts the "cheap" claim into a value-trap and will not move off it. "Apparent cheapness below book can persist and deepen when the business destroys operating value faster than the asset base liquidates (M37). Name me the mechanism that earns above the cost of capital — you can't, because there isn't one; there is a negative lollapalooza instead: shrinking end-markets, cheap-Chinese displacement, product wind-downs, and an ERP cost burden, all reinforcing (M86). A large cash cushion does not offset that. It delays the reckoning. Pass."

Pabrai presses the other way — the floor changes the arithmetic. "Agreed the business is bad; I said so in my first paragraph. But leverage is what kills you, and there is none — net cash exceeds the market cap and the going-concern note is clean (P1, P20). That makes this 'heads I win, tails I don't lose much' at the right price. Today asset monetization is still outrunning the operating burn — PP&E sales of ¥2,423M F289, cash up ¥4,670M F299 — so it's the Frontline pattern, not Sears (P7). My problem is not the thesis; it's the entry. At ¥566 you pay right against net cash; my line is half of a ~¥900 liquid value, so I watch with a buy-below of ¥380."

Buffett and Li Lu refuse to price it at all, and for the same reason. Buffett: "I can compute the net-net to the yen, and it's real — but a company revising its own plan down to an operating loss, guiding to −¥3.9bn F383, winding down product lines and dissolving subsidiaries E9E19, is being reinvented faster than I can forecast it (B2). And the discount only pays if the assets sit still or someone converts them — nobody here has the votes (B65, B103)." Li Lu: "Precisely. The ten-year knowledge bar gates everything (L1). And the assets are not a static dollar — the only profitable year was manufactured by selling the very factory a value investor separately credits on the balance sheet, while operations bled (L21, L35). Cheap on a melting book is not fifty cents on the dollar. Too-hard."

Claude names what none of them can fully know, and lands between Pabrai and the too-hard pair. "Both sides are right about something. The floor is real and reachable — a haircut NAV of ~¥1,264 is 2.2× the stamp, and deployable cash is ~1.0× the market cap, which takes deep permanent loss off the table (C39, C47). That is why it is not a pass. But a floor only becomes return through a mechanism with a date, and the mechanism here is plan-only, un-track-recorded, and un-compellable — no controller, an unconfirmed activist, a nil buyback, an undecided dividend (C57, C58, C62). At ¥566 the whole-company owner yield is negative, because the operating loss consumes the ¥1,119M of financial income F282 — so I will not pay for an unlock with no agent and no date (C35). It is time-destruction, not capital-destruction: a decade at a discount that may never close. Watch, implied buy-below ¥420."

The crux, then: is a net-cash, below-book, melting business a bargain (Pabrai and Claude, at a lower price), a trap (Munger), or unknowable (Buffett and Li Lu)? The three watch/pass/too-hard families are not arguing about the number on the balance sheet — they agree on it. They are arguing about whether that number is collectible by an outside minority before the ice melts, and none of them can point to the agent who would collect it.

Self-distance note

The Claude lens holds one of the five verdicts compared above (watch, implied buy-below ¥420), and this synthesis is likewise Claude-authored. In this autonomous run the reconciled figure table all five lenses consumed was built by a (Claude-driven) dual-blind extraction — two independent passes reconciled per-(metric, period) against the page-delimited source. Read the synthesis with that in mind: the same author's fingerprints are on the ledger, on one of the five memos, and on the reconciliation you are reading.

Governance and the activist question

Governance and compensation were fully extracted, as the gate requires for any verdict, and they come back clean. Futaba is a company with an audit & supervisory committee (監査等委員会設置会社) E49; the board is 8 directors, half (4) of them independent outside directors E50E59, with a voluntary Nomination & Remuneration Committee chaired by an outside director and holding an outside majority E51. Director compensation is fixed 70% plus performance-linked 30%, the variable portion swinging ±30% across five equally-weighted KPIs (consolidated sales YoY, value-added YoY, free cash flow, operating margin, ROE) E63E64E65; no individual is paid ¥100M or more E70; related-party transactions in FY2026 were none E87; and the auditor (Deloitte / トーマツ) has been continuous since 1982 E62. Because no lens issues a verdict stronger than watch, the governance/compensation cap binds nothing here — but had a buy-below wanted to rest on a stronger claim, the record would have supported it.

The load-bearing governance fact is not a flaw; it is an absence. There is no forcing agent. No controlling shareholder and no parent E44E84; a register led by the Futaba Electronics Memorial Foundation (7.67%), a Master Trust nominee (10.18%), founder-linked individuals (Yoshida 4.71%, Kawasaki 4.39%), and Chiba Bank (4.38%) E42; and an activist — Brandes at 8.80% — whose stake is a Nov-2024 large-holding report that the company could not confirm was still held at the 2026-03-31 register date, and so excluded from the major-shareholder table E43. This cuts both ways, as Claude's C59 notes: the absence of a controller removes squeeze-out risk and removes the one party who could compel a return of capital. It is exactly the fact that keeps this a cheap, below-book, net-cash company with no one positioned to close the discount.

Prediction-vs-actual

VOID. This is an autonomous cycle; the human blind prediction is voided (void: no-human-prediction) and no prediction-vs-actual scoring applies. The Claude lens's pre-registered probabilities (in profiles/claude.md) and the two watch-lens falsifiers below stand as the study's forward, resolvable record.

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.

What would change our minds

The two watch lenses each pre-registered a falsifier; these are what a future review scores against, not hindsight.

Pabrai (watch, buy-below ¥380) — two-part falsifier. (1) Asset-floor break: net cash (cash & deposits F133 less any new interest-bearing debt) plus listed cross-holdings E74 falls below ~1.2× market cap — i.e., the hard floor stops covering the price with room to spare. (2) Melting-clock break: two consecutive fiscal years in which the melt (operating loss F223/F381 plus real cash restructuring outflows) exceeds monetization proceeds AND period-end cash & deposits F133 falls year-on-year — the ice cube is then shrinking the floor faster than sales replace it, and this stops being "tails I don't lose much."

Claude (watch, implied buy-below ¥420) — forcing-agent / deterioration falsifier. Upgrade to buy-below only if a forcing agent is demonstrated within ~2 reporting cycles: (a) a ≥¥3bn board-resolved buyback actually executed with shares cancelled (not the 492-share odd-lot of E45), OR (b) traced cross-holding disposal proceeds ≥¥5bn returned to shareholders ahead of the 2029-09 first-tranche date in E21, OR (c) an activist 5% amendment (Brandes E43 or new) with a stated balance-sheet purpose and a shareholder proposal on the AGM agenda. Kill the watch (→ pass) if two of: FY2027 revenue prints below ¥42bn AND operating loss below −¥2.3bn in the FY2027/3 tanshin; the FY2027 dividend F389 is set at ¥0; or cash & deposits F133 fall below ¥30bn without offsetting securities disposals returned to holders.

What this taught the checklists

Three revision notes surfaced from the profiles, queued for the next study per the revision contract (see docs/process/evolution.md).

• Pabrai — P40/P41 skin-in-the-game items misfire on no-controlling-shareholder Japanese cases. Here there is genuinely no owner-operator and no controlling holder E44E84 — which is good for minority protection (no one to trap value) but scores as a fail on "skin in the game." The lens should distinguish a misaligned entrenched controller (bad) from diffuse ownership + professional management + an activist present (neutral-to-good). Brandes's presence E43 and the absence of a controller arguably raise the odds value reaches a minority, yet the current items give no credit for that. (Pabrai also flagged that P7 needs a stated, numerical monetization-vs-melt ratio, and P1's asset floor should explicitly instruct crediting listed cross-holdings at a haircut.)

• Buffett & Li Lu — the "profit-from-self-liquidation double-count trap." Reported net income can be manufactured by selling the very assets a value investor separately credits on the balance sheet — Futaba's only profitable year was entirely fixed-asset-sale gains F241E32 while operations lost money F223. Buffett routes this to a "minority-interest wedge" note on the net-net items (NCAV per common must divide by the parent's share, not group current assets ÷ total shares, given ¥11,406M of NCI F209) and a cross-reference making the melting-bargain → too-hard path mandatory. Li Lu proposes an explicit item flagging "book-value discount + asset-sale-driven earnings" as a specific double-count trap on Japanese asset-heavy decliners, and separating distributable from merely booked assets (the ¥10,002M pension surplus F159 and 2030-dated cross-holdings E21 inflate book but are not near-term owner cash).

• Claude — forcing-agent / owner-yield handling. Three item-weaknesses captured: score a mid-term plan's own mid-flight downward revision E19 as the calibration datapoint when only one year of filings is in-archive (C7/C8); treat an explicitly-undecided forward dividend F389 as an unpaid wait rather than crediting a trailing yield the company has declined to commit (C42); and score asset-floor realization per asset class — demonstrated (real-estate sales E90) vs planned (securities E21) vs theoretical — rather than as a single monolithic path (C39/C45).

---

Corrections