ELECOM Co., Ltd. (6750): The Batch's Cheapest Name, a ¥58bn Floor, and the Founder Who Won't Open It — A Fortress You Cannot Enter

Verdicts

Lens	Verdict	Buy below	Most load-bearing items
Buffett	watch	¥1,350	B72; B84; B23/B25
Munger	watch	—	M83; M52; M90
Pabrai	watch	¥1,350	P53/P66; P40; P13
Li Lu	too-hard	—	L1; L39; L24
Claude	watch	implied ¥1,500	C39/C47; C74/C73; C33/C35

Four of five lenses reached watch; one — Li Lu — reached too-hard: a watch consensus (4-of-5). ELECOM is pick #3 of the "quality at a fair price" batch and its best shot at a buy — the batch's lowest multiple (P/B 1.36× D1), its biggest, most reachable-looking net-cash floor (+¥58.0bn D32, ~39% of the ¥147bn cap D35, ~¥720/sh D34), and an ex-cash operating business earning ~28.6% ROIC D30 at only ~7× ex-cash normalized earnings D40 — genuinely cheap and good. And it still drew only watch, every buy-below (¥1,350–1,500) below the ¥1,830 stamp, clustered near BOOK. Two findings decided it, and they converged across all five lenses. First, the ¥58bn floor is founder-trapped, and management just proved it: the register is a founder bloc (Hada 20.78% E79 + Sands 15.64% E79 + foundation 2.23% E79, 38%, plus 12.67% treasury E80), the founder personally sets director pay E106 and his son is an executive E98, no activist — and the tell: the ¥7bn buyback was FY2025, not FY2026 E126; in FY2026, with standing board authority to act E90 and the stock near its low, the treasury purchase was **¥1M** E126 and management reissued 4,154,667 treasury shares to fund the Nippon Antenna acquisition E127, so net float rose. Second, the FY2026 record is one-off-flattered: a ¥7,648M non-cash negative-goodwill gain E134 from buying Nippon Antenna at ~50% of book D48 collapses the reported 21.2% ROE / 7.1× P/E to a normalized ~13% ROE D26 / ~11–12× P/E D22 — dead on management's own 13% target E31 — while operating cash flow fell to ¥9,877M from ¥17,354M F63F62. The batch's arc completes: EBARA JITSUGYO's moat was rented, Daitron's was owned, ELECOM's floor is trapped — three clean quality names at fair prices, all watch, none a buy.

The business

エレコム (ELECOM) is the small stuff in every desk drawer. It began in 1986 as an OA-furniture maker selling PC desks to consumer-electronics mass retailers, added accessories in 1987, and launched its first mouse in 1988 E4. Forty years on it develops and sells mice, keyboards, cables, chargers, docking stations, power strips, mobile batteries, storage and NAS, plus a business-to-business solutions arm and — bolted on by acquisition — hair dryers and broadcast antennas E12E13. A shopkeeper can follow how the cash comes in: design the product, have a contract factory build it, move it fast through electronics retailers and Amazon Japan. It is a single reportable segment E12 and reports by product category rather than by division E12.

The defining structural fact is that ELECOM makes almost nothing itself. It is fabless — apart from part of its subsidiaries the Group holds no manufacturing plant, and production is outsourced to contract makers in China and Taiwan, with the finished goods imported mostly in US dollars E35E42E43E44. Management names its own edge plainly: a strong supply chain — fast product development, procurement scale, and agile sales and logistics that move goods to the shelf on trend E36. That is genuine operating competence, not a toll bridge; the same risk pages concede daily fierce price competition and an inability to always pass rising costs into price E46.

Growth comes from buying brands, not from organic compounding. ELECOM is a serial acquirer — Logitec (2004), Hagiwara Solutions and JDS, DX Antenna (2017), Tescom Denki (2023), and Nippon Antenna (2025) E5E6E8E9 — and it runs a multi-brand portfolio (ELECOM, Logitec, HAGIWARA Solutions/JDS, DX Antenna, Tescom, and now Nippon Antenna) under one roof E13. Two of those brands push it beyond PC peripherals into broadcast/telecom infrastructure: DX Antenna already exceeds 10% of consolidated sales E16, and Nippon Antenna makes broadcast/telecom equipment and does electrical-telecom construction E19. There is also a healthcare arm — Elecom Healthcare, run by the founder's physician son E17E98 — and a BtoB solutions business (rugged tablets, NAS, GIGA-school keyboards) that grew 29.6% to ¥42,909M in FY2026, partly on the six-month Nippon Antenna consolidation E62. The largest single customer is Amazon Japan, at 11.4% of sales E68.

The whole thing sits in a maturing end-market, and management says so. Its own environment note describes market maturation across PC, smartphone/tablet, and TV/AV, with intensifying competition from emerging global makers E25E41, offset by resilient demand in peripherals and networking from e-commerce, corporate and public digitalization, security and education. Short product life cycles force monthly inventory disposal and quarterly write-downs E45; an industry practice called "inventory compensation" reimburses retailers when ELECOM cuts prices on stock they already hold E49. This is a good, durable, cash-generative accessories business — but a price-taker in a commoditizing category, not a franchise with pricing power.

The numbers

At the ¥1,830 stamp — a P/B of just 1.36× D1, the batch's lowest — ELECOM is a good-enough operating business buried under a fortress of idle cash, and the reported "record" is a mirage a hurried reader must dismantle.

The ROE record is moderate, and FY2026's headline is a one-off. Return on equity ran 13.2 / 10.0 / 11.9 / 11.0 / 21.2% across FY2022–26 [F49–F53]; the first four years are a low-teens band around the company's own ≥13% target E31, never a compounding machine. The FY2026 21.2% F53 is flattered by a ¥7,648M 負ののれん発生益 (negative-goodwill gain) F218 booked because ELECOM bought Nippon Antenna for about half its book value — ~¥15.3bn of fair-value net assets D46 for a ¥7,669M acquisition cost E131, a ~50.1%-of-book deal D48. The gain was also non-taxable, cutting the effective rate to 18.4% versus the 30.6% statutory D18 — so the record flatters twice, in the gain and in the tax line. Strip it: normalized net income is ~¥11,863M cleaner / ~¥12,543M simple D18D16, normalized ROE ~12.5–13.2% D26D25 — right at the target — and the reported 7.07× P/E D3 becomes a normalized ~11.4–12.0× D21D22. And the cash tells the same story: this best-ever earnings year generated less operating cash than the year before — ¥9,877M in FY2026 versus ¥17,354M in FY2025 F63F62, because the FY2026 cash-flow statement correctly deducts the ¥7,648M non-cash gain E126.

The balance sheet is a genuine fortress, and the floor is large. Cash and deposits of ¥58,497M F78 against interest-bearing debt of just ¥536M — a ¥500M short-term bank line, ¥36M of leases, no bonds, no long-term borrowings E142D33 — leave net cash of +¥57,961M D32, ~¥58bn, about 39% of the ¥147.4bn net market cap D35 and roughly ¥720 a share D34 inside a ¥1,830 stock. Equity ratio is 74.4% F48. The business is capital-light: capex ¥4,773M E76 roughly matches D&A, so owner earnings sit close to normalized net income.

The operating business, stripped of its cash, is genuinely good and genuinely cheap. Set the ¥720/sh of net cash aside and the market is paying about 7× normalized operating earnings — the ex-cash operating multiple is 6.9–7.3× D40D39 — for an enterprise earning **28.6% ROIC on the capital actually deployed in it D30**, versus a blended 9.9% D31 once the idle cash is included, and versus a P/B of just 1.36× D1, the batch's lowest. Gross margin is a stable ~39.6% D8 on the fabless model; operating margin ~11.7% D10 sits mid-range of a tight recurring band. Organic growth over the five-year window is only **5.3% p.a. D49**, recurring profit ~3.6% D50, and book value per share ~9.8% D51 — the growth and the "gains" are M&A artifacts, not an organic engine.

Capital return is real on paper but paused in practice — the tell. The mid-term-plan policy is a progressive dividend plus a payout ratio of 30%+ of consolidated net income E82, and a ¥7.0bn buyback authorization F507. But the ¥7,005M treasury purchase visible in the cash flow is the prior-year FY2025 execution E126; FY2026's own-share acquisition was ~¥1M E126. Worse, FY2026's treasury movement went the other way — 4,154,667 shares were reissued from treasury to fund the Nippon Antenna share exchange E127, so net float rose. FY2026 dividends of ¥57.00 DPS included a ¥5 40th-anniversary commemorative and produced a payout of only 22.0% reported / 35.4% ex-negative-goodwill, down from FY2025's 40.3% E83. So in the year the stock was cheapest, with a live authorization and standing board power to act E90, management bought essentially nothing, reissued equity, and cut the payout.

The five lenses

Buffett — watch

Let me start where I always do — with the business, not the ticker. This company began in 1986 selling computer desks and the humble mouse E4, and today it develops and sells the small stuff that clutters every desk drawer: mice, keyboards, cables, chargers, docking stations, storage, plus a BtoB arm and, bolted on by purchase, hair dryers and broadcast antennas E12E13. It designs and outsources the manufacturing — fabless E35 — and moves goods fast through electronics retailers and Amazon, which alone is 11.4% of sales E68. I can explain that to my sister in one breath, so we are inside the circle. Good.

Now the number that will fool a hurried man. Reported earnings hit a record ¥20,191M F15, and against the ¥1,830 price that looks like a P/E of 7 D3 — a steal. It is not. Fully ¥7,648M of that "profit" is a non-cash, non-taxable negative-goodwill gain F218E134 booked because they bought Nippon Antenna for about half its book value D48 — a real bargain for them, but a one-time accounting windfall, not the till ringing. Take it out and normalized earnings are roughly ¥12bn D16, the honest P/E is nearer 11 to 12 D22, and that squares with the four prior years' return on equity — 13.2, 10.0, 11.9, 11.0 [F49–F52] — and with the company's own 13% target E31. The "record" is the same low-teens-return business wearing a party hat.

Here is what genuinely attracts me. This is a fortress: cash and deposits of ¥58.5bn against essentially no debt E142, net cash of about ¥58bn D32, ~39% of the whole net market value D35, roughly ¥720 a share D34; equity 74% of the balance sheet F48. Set that cash aside and you're paying about 7 times normalized earnings D40 for a business earning close to 28.6% on the tangible capital actually working in it D30. On paper, that is the Buffett dream sentence.

But it isn't mine to run, and that is the crux. The founder personally owns 20.78% E79; add Sands at 15.64%, the foundation at 2.23%, and 12.67% in treasury E80, and the outcome is decided by one man and his affiliates. He sets the individual pay of every director himself E106; his son just joined the board as an executive officer E98; there is no activist and no need for a poison pill E91 — the register is the poison pill. And the sin is what the cash is doing. In the very year earnings looked like a record, the effective payout was cut to 22% from 40% the year before E83, and the "¥7bn buyback" people cite was last year's outlay — this year they repurchased ¥1M, essentially nothing E126. Apply the one-dollar test — has each retained dollar created a dollar of market value? — and a price at 1.36 times book D1 is a polite no. The cash is real; my access to it is not.

And I cannot lean on a moat to make me patient, because there isn't one. Management's own risk pages describe daily fierce price competition E46, an inability to pass rising costs into price E44, a maturing end-market E25, and inventory that goes obsolete fast E45. Their real strength is speed and low cost E36 — respectable operating, but a low-cost fast-follower is a price-taker, not a toll bridge. Growth comes mostly from buying other companies E5E9 while the organic top line has compounded only about 5% D49.

That is a watch, not a buy — the best watch in the batch. A discount that merely persists inside a founder's balance sheet is not a margin of safety; it is the market correctly pricing trapped capital. I'd own it happily at a price near book (BPS ¥1,341.68 F35) — around ¥1,350, where I'm paying little more than the cash for the whole operating business — or if someone lit a fire under that ¥58bn net-cash pile D32.

What a student should take from this: when a "record year" and a P/E of 7 appear together, find the one-time item before you get excited — a non-cash, non-taxable gain F218 turned a low-teens-return business into a fake bargain here. And cash you cannot reach is not the same as cash you own: a ¥58bn fortress behind a controlling founder who just cut the payout and stopped buying back E83E106 is a reason to wait for a catalyst or a lower price, not a reason to pay up.

Munger — watch

Invert first. How does buying ELECOM at ¥1,830 produce a permanent loss? The primary kill path is not business failure — this company has made PC peripherals for thirty-nine years and never reported a loss in the five-year record [F11–F15]. The kill path is the cash. You own a business generating roughly ¥11–12bn of normalized net income D16 atop ¥58bn of idle cash D32 — 39% of market cap D35 — controlled by a founder who has placed his son on the board E98, decides director compensation personally E106, and faces no activist and no mechanism short of extraordinary resolution to force redeployment E79.

The business itself is genuinely good. Strip the cash and the operating economics snap into focus: ex-cash ROIC of roughly 28.6% D30 at an ex-cash earnings multiple of 6.9–7.3× D39D40, gross margin 39.6% D8 on a fabless model with no fixed-asset burden. But the profit growth is sluggish — ordinary profit compounded only 3.6% over five years D50 against revenue at 5.3% D49, and normalized ROE runs 12.5–13.2% D25D26, essentially at the company's own minimum target E31. The operating business merits a "fair" grade; the normalized P/E of 11.4–12.0× D21D22 is not heroic.

The moat mechanism is narrow — two tendencies, not five. ELECOM is the largest multi-category PC-peripheral brand in Japan, fabless, with a Japan-China two-pole fast-development model E75 giving genuine speed. Customer habit and ubiquity matter in sub-¥3,000 accessories: you grab the brand you recognize from the shelf or from Amazon Japan (11.4% of sales E68) without reconsidering. But the brand commands no lasting pricing power — the filing itself names daily fierce price competition and an inability to always pass through cost increases as a named risk E46. That is active management plugging holes, not pricing power absorbing them.

Incentives: the founder is the story and the risk. Hada controls roughly 38.6% of the effective float (20.78% + Sands 15.64% + foundation 2.23% E79), the treasury holds a further 12.67% E80, and he sets director compensation E106 under a cap he effectively approved. Outside directors are four ex-bankers E99 — competent, unlikely to tell a founder who has compounded at 13% for decades that he is wrong. The "who can tell the CEO no" question [M83] has a weak answer. And the serial-acquisition program is the crux: the Nippon Antenna deal at 50% of book D48 is good capital allocation when it happens, but it requires a steady supply of distressed sellers you cannot plan around — and the same portfolio shows a Tescom impairment where recoverable value-in-use was judged zero E124, one year after acquisition. The single-segment reporting means the operating quality of each acquired subsidiary is unverifiable from the consolidated figure table E12 — a complexity flag that pushes toward watch.

The permanent-loss map is benign: 74.4% equity ratio F48, no bonds, net cash ¥58bn E142. A fifty-percent-price-drop test is straightforwardly passed. The only permanent-loss path is a decade of value-destroying acquisitions at full price, funded by the hoard — a risk, not a certainty.

Verdict: watch. At 7× ex-cash earnings D40 for a genuinely high-ROIC operating business, the statistical case is real. But the forcing agent for cash deployment is absent E79, the moat is narrower than the multiple implies, and the clearest observable signal is damning: in FY2026, with the ex-cash multiple among the cheapest in the company's history, buyback execution was essentially zero E126 while the ¥7bn authorization F507 sat on paper. That is not Singleton discipline. I will not buy a cheap cage hoping the cash escapes by accident. I issue no buy-below number: the block is governance, not price, and it will not clear at a mark I can name today — it clears only when the capital-allocation behavior clarifies.

What a student should take from this: the operating business buried under a large cash pile is a classic value-trap disguise — the cheapness is real but so is the structural obstacle to realizing it, a controlling founder with no exogenous pressure to change course. When quality and price are both attractive but the capital-allocation agent is unaccountable, the right response is to watch and wait, not to hope. And a low-teens ROIC including idle cash looks mediocre; stripping the cash to reveal 28.6% ex-cash ROIC D30 shows what the business actually does with the capital it deploys — always separate the business from its balance-sheet excess before judging the economics.

Pabrai — watch

Let me start where I always start — the downside — because with this one the balance sheet is the whole argument, and the argument does not quite close.

The business is simple enough to pass my paragraph test. ELECOM is a fabless E35 Osaka brand that designs PC peripherals and digital accessories, outsources manufacturing to China and Taiwan E43, and sells through mass retailers and Amazon Japan E68. One segment E12, at it since 1986 E4, grows by buying brands — Logitec, DX Antenna, Tescom, now Nippon Antenna [E5–E9]. A ten-year-old could follow it.

Now the fortress. Cash and deposits of ¥58,497M against ¥536M of interest-bearing debt — no bonds, no long-term borrowings E142 — leave net cash of +¥57,961M D32, about 39% of the ¥147.4bn market cap D35, or ¥720 a share D34 inside a ¥1,830 stock. Equity ratio 74.4% F48. This is exactly the Japanese hunting ground I love: a company that could survive a shutdown for years on cash alone, dilution-proof, with survival never in question. Tails, I do not lose the company. And the operating business underneath is genuinely good — strip the cash and the market pays about 7× normalized operating earnings D40 for a business earning ~28.6% return on deployed capital D30. That is a fifty-cent dollar's skeleton.

So why only watch, and why below the stamp? First, the floor is hard but it is not mine. The ¥58bn is controlled by a founder who shows no intention of handing it to outside owners on my timetable. Chairman Hada owns 20.78% personally E79, plus Sands 15.64% and his foundation 2.23% E79 — ~38% in the founder bloc — and the company holds another 12.67% in treasury E80. He personally sets each director's pay E106, and his son just joined the board as an executive officer E98. There is no activist E91, but also no lever to pull. The dividend yields 3.11% D7 on a 22% payout E83, and the ¥7.0bn buyback everyone points to is an authorization, not this year's cash — FY2026's actual treasury purchase was ¥1M E126, because the treasury shares went to Nippon Antenna's holders, not to me E127. The cash is not returning fast; it is being stockpiled to buy more companies E30.

Second, is this the good kind of uncertainty? My whole method is to buy where the market prices uncertainty as risk. But ELECOM is not cheap because of a nameable, temporary fear — there is no cyclical trough, no lawsuit, no forced selling. It trades at 1.36× book D1 and ~11–12× normalized earnings D22 — a fair price for a decent business. The "cheapness" is entirely the ex-cash optical trick; strip it and the operating business is a low-teens-ROE D26, maturing-end-market E25, price-competed E46, fabless reseller — a fine business at about the right price. That fails my fifty-cent test at ¥1,830. Third, the serial-acquirer engine cuts both ways. The Nippon Antenna deal at ~50% book D48 booked a ¥7,648M gain E134 that flattered ROE to a headline 21.2% F53 that is really ~12.5% D26. Buying below book is a Mittal-style virtue if repeatable — but it means my floor could be redeployed at any moment into the next deal, at a price I do not control.

Nothing here is fatal. Survival is certain, the sole audit matter is a routine refund-liability estimate, not the acquisition E146. This is the best fifty-cent-dollar candidate in the batch precisely because the floor is so hard. But at ¥1,830 the discount is not there — I'd be paying full price for the business and getting founder-trapped cash as a maybe.

Verdict: watch, with a buy-below at roughly book — ¥1,350 — where the ex-cash operating business would cost ~5× and the margin of safety would finally be wide enough that I do not need the founder's permission to win.

What a student should take from this: net cash worth 39% of the market cap is a magnificent floor — but a floor only counts as yours if owners can actually reach it, and a founder who owns ~38%, sets his own board's pay E106, and stockpiles cash to buy more companies controls that floor, not you. And beware the "ex-cash P/E of 7" optic: strip out cash you can't touch and what's left is a fair-priced, low-teens-ROE reseller in a maturing market. Cheapness is a relationship between price and reachable value, not between price and a balance-sheet cash line.

Li Lu — too-hard

I begin, as I always must, with the question that governs everything: can I honestly say I understand this business's next ten years better than almost anyone who owns it? If I cannot, then no price rescues it — cheapness is not a substitute for knowledge, and intellectual honesty about the boundary of my competence is the whole discipline.

ELECOM is a fabless brand: it develops PC-peripheral, digital-accessory and home-electronics products, outsources all manufacture to Chinese and Taiwanese contract makers, imports the finished goods mostly in US dollars, and sells through mass retailers and Amazon Japan E12E35E42E44. It is a single reportable segment E12, and it has grown — revenue compounded from ¥107,358M to ¥132,132M, ~5.3% p.a. F1F5D49. The record looks strong at first glance: FY2026 net income ¥20,191M at a P/B of only 1.36× F15D1. But it is one-off-flattered. Of that ¥20,191M, ¥7,648M is a non-taxable negative-goodwill gain from buying Nippon Antenna below book F218E134. Strip it and normalized earnings are ~¥11,863–12,543M D18D16, normalized ROE ~12.5–13.2% D26D25 — squarely in the 10–13% band of the four prior years [F49–F52], right at the company's own 13% target E31. The "21.2%" is not a step-change in profitability.

Now to the ten-year knowledge bar. The variables that decide ELECOM's earnings power a decade out are: whether a fabless accessory brand's shelf position is durable pricing power or a substitutable commodity; the trajectory of a PC/digital end-market the company itself calls maturing E25E41; and the durability of its serial-acquisition engine. On every one, the filings push me toward "I cannot know." The yūhō concedes "daily fierce price competition" and that the company "may be unable to pass" cost increases into price E46, short life cycles and inventory obsolescence E45, intensifying competition from emerging global makers E25. This is not the language of a moat; it is the language of a price-taker in a hardware race. I see no switching cost, no license barrier, no winner-take-all line being crossed. A fabless brand's advantage can erode quietly, and the filings give me no evidence it will still earn its return in ten years against emerging makers and Amazon's own private label.

Where ELECOM is genuinely strong is the balance sheet and the structure, and I want to give this its full due, because it is what tempts me. Net cash is +¥57,961M, ~39% of the ¥147.4bn cap, no bonds and no long-term debt D32D35 — a genuinely reachable floor of ~¥720 per share D34. The claim on that cash is clean: no listed parent E154, subsidiary dividends reach the parent, no VIE, related-party dealing trivial. EY ShinNihon has served 22 years with a clean opinion and one benign Key Audit Matter E103E146. Structurally, this passes.

But the founder control is total, and the cash is trapped behind it. Hada holds 20.78% personally, plus Sands 15.64%, plus his foundation 2.23%, plus treasury 12.67% E79E80. He is chairman, sets each director's pay E106, and his son is now a director and executive officer E98. There is no activist and no rights plan E91. Capital allocation, which for me is the defining measure of management, turns ambiguous: the Nippon Antenna deal at ~50% of book D48 was a genuinely good act, but the ex-cash operating ROIC is ~28.6% while the blended figure is dragged to ~9.9% by the cash pile D30D31. A rational owner earning 28.6% on deployed capital does not let 39% of the company sit at deposit rates. The buyback is real but lumpy — ¥7,005M in FY2025, ¥1M in FY2026 E126 — and payout only reached the 30% floor E82. This is a hoard explained more by founder comfort than by the business — precisely the pattern that keeps cheap Asian names cheap.

Put it together. If the exchange closed for ten years, I would be relying on a ~12–13% ROE business whose moat I cannot verify and whose surplus cash the founder may never release. The permanent-loss mechanisms — obsolescence E45, price war E46, FX I cannot hedge forever E44 — are not bounded by the filings; they are the business. The knowledge bar gates everything, and I fail it here. This is not a business I can predict better than its owners: too-hard.

What a student should take from this: a fortress balance sheet, a clean structure, and a headline P/B of 1.36× D1 are not enough — the knowledge bar comes first, and a fabless accessory brand in a maturing, price-competitive, FX-exposed market is a business whose ten-year earnings power I cannot honestly forecast. Learn also to normalize before you admire: the "record" year here is a low-teens-ROE year wearing a ¥7.6bn one-off gain. And ~39% of the price sitting as founder-trapped cash is not automatically your margin of safety — value only counts if it can eventually reach the owner.

Claude — watch

My figures-blind priors made four load-bearing bets, and the ledger lets me score them. H1 (moat durability): I predicted a maturing, price-taking accessory business whose returns quietly revert, not a durable franchise. The ledger confirms this, with one correction. The reported ROE series 13.2/10.0/11.9/11.0/21.2% [F49–F53] normalizes to ~12.5–13.2% D25D26 once the ¥7,648M negative-goodwill gain F218E134 is stripped — exactly at the company's own ≥13% target E31, never above. Organic growth is ~5.3% p.a. D49 in a market management itself calls maturing E25 with daily price competition it "cannot always pass through" E46. But the correction matters: the ex-cash operating business is better than good-enough — ROIC ex-cash ~28.6% D30 versus ~9.9% blended D31. So H1 resolves as: not a moat, but a good operating business dragged to mediocrity by idle cash — the reachability variant, not the quality variant, of a compounder. (Self-distance: I hold this verdict, built the reconciled figure table all five lenses consumed, and wrote the synthesis below — read all three with that concentration of authorship in mind.)

H2 (floor reachability) — the load-bearing prior. I predicted the ¥58bn cash pile is founder-trapped, worth much as downside insurance and little as an accessible unlock. The ledger confirms this hard, and adds the tell I hoped for. Net cash is +¥57,961M D32 — ~39% of the ¥147bn net cap D35, ~¥720/sh D34. The register: founder Hada 20.78% + Sands 15.64% + foundation 2.23% E79 + treasury 12.67% E80 — a decisive controlling bloc, no visible activist, no pill needed E91. The founder sets director pay E106, his son joined the board as executive officer E98. And the tell arrives precisely: the ¥7bn buyback was FY2025, not FY2026 — the ¥7,005M treasury purchase in the cash flow is the prior-year execution; FY2026's own-share acquisition was ¥1M E126. FY2026's treasury movement went the other way — 4,154,667 shares reissued from treasury to fund the Nippon Antenna exchange E127, so net float rose. In a year the stock was available near its low, management bought essentially nothing and reissued equity. Revealed preference: the cash funds M&A E64, not the minority.

The deciding work is the owner arithmetic. Normalized owner earnings ¥11,863M D18 (capex ~D&A E76, so owner earnings ≈ normalized NI) on the ¥147.4bn net cap D5 is an 8.0% whole-company yield; strip the cash and the operating business trades at EV ¥89.4bn D36 / normalized NI ~7.5× D37, a ~13% ex-cash yield — plus ¥58bn earning ~0%. Bear case, valued first (C33): take the observed FY2023 trough net income ~¥8.1bn F12, value the operating core at 8× no-growth = ~¥795/sh, add the ¥720/sh cash floor assumed to hold = **¥1,515/sh**, a ~17% draw from ¥1,830 — cushioned, not eliminated, by real cash. The floor caps the downside but does not close it. My jury of selves pushed every divergent item the same way — toward requiring more margin and a lower entry — and confirmed the number.

Verdict: watch, implied buy-below ¥1,500. At ¥1,830 this is the batch's cheapest name (P/B 1.36× D1) with the biggest floor and a ~7× ex-cash operating business at 28.6% ROIC D30 — the three ingredients a buy needs. But the bear-case margin of safety is only ~7–17%, and a founder-trapped floor with no forcing agent and a demonstrably switched-off buyback demands more margin, not less, because the ¥720/sh of cash is insurance I cannot compel and the operating business, while good, is maturing. It is not a pass: the downside is cash-cushioned, the dividend is progressive E82, and the ex-cash operating business is genuinely good. It is not a buy here: the price is fair-to-slightly-cheap on what I can compute, not cheap enough to underwrite the trap. That gap between ¥1,830 and ¥1,500 is the founder-control discount I am declining to pay. This is a watch with a live path to buy: an executed return of capital that shrinks the float closes it.

What a student should take from this: a record ROE of 21.2% that normalizes to 12.5% once a non-taxable one-off is stripped D26E134 is not a profitability step-change — always find the one-off before you find the multiple. And a large, real cash floor (~¥720/sh, 39% of the cap D34D35) is only worth its face value to a minority if someone can reach it; when the founder controls the register, sets his own directors' pay E106, reissues equity rather than buying it back at a low E126, and faces no activist, the floor is downside insurance, not upside — so it earns a bigger required discount, not a smaller one.

Synthesis

Where the lenses agree

Four lenses reached watch and the fifth (Li Lu) reached too-hard — a watch consensus (4-of-5) — and beneath the split-of-one the panel is unanimous on the three facts that are the whole study.

First, the operating business is genuinely good, and cheap on its own capital. Strip the idle cash and it earns ~28.6% ROIC D30 at ~7× ex-cash normalized earnings D40, on a fortress balance sheet — equity ratio 74.4% F48, net cash +¥58.0bn D32. Every lens credits this. Buffett: "the Buffett dream sentence." Munger: "the operating business merits a fair grade … 28.6% ex-cash ROIC shows what the business actually does with the capital it deploys." Pabrai: "a fifty-cent dollar's skeleton." Even Li Lu, who declined, wrote "I want to give this its full due." This is not a melting-ice cube; it is a real, above-average operating business.

Second, the FY2026 record is one-off-flattered, and management said so in writing. The ¥7,648M negative-goodwill gain F218E134 from buying Nippon Antenna at ~50% of book D48, non-taxable and so double-flattering D18, collapses the reported 21.2% ROE / 7.07× P/E to a normalized ~12.5–13.2% ROE D26D25 / ~11.4–12.0× P/E D22D21 — dead on the ≥13% target E31, the floor of the aspiration, not a beat. And the best-ever earnings year generated less operating cash than the year before — ¥9,877M versus ¥17,354M F63F62. Every lens normalizes the record down; this is the panel's operating-line discipline catching a flattered headline, the batch's third in a row.

Third, the ¥58bn floor is founder-trapped, and the buyback tell confirms it. The register is a founder bloc — Hada 20.78% + Sands 15.64% + foundation 2.23% E79, plus 12.67% treasury E80 — with the founder setting director pay E106, his son on the board E98, and no activist E91. And the revealed preference is damning: a ¥7bn authorization F507, ~¥1M executed in FY2026 E126, with 4.15M treasury shares reissued so the float rose E127. Every lens reads the same fact. Buffett: "the register is the poison pill." Munger: "not Singleton discipline." Pabrai: "the floor is hard but it is not mine." Claude: "a cash-box whose owner likes the cash."

Where the lenses diverge

The verdict is a 4-of-5 watch, but the buy-belows spread from ¥1,350 to ¥1,500 — narrower than earlier in the batch — and every explicit number sits below the ¥1,830 stamp, clustered near BOOK (BPS ¥1,341.68 F35). The spread is not noise; it is three different definitions of the discount required, plus one lens whose bar the price could not reach at all.

Pabrai and Buffett both anchor to the net-cash floor and demand to pay barely more than it — a buy-below of ¥1,350, at book (BPS ¥1,341.68 F35). Pabrai's crux is the fifty-cent-dollar test (P53): "at the stamp the discount is not there … it clears at roughly book, where the ex-cash operating business would cost ~5× and the margin of safety would finally be wide enough that I do not need the founder's permission to win." Buffett arrives at the same buy-below from his own route — B72 (the one-dollar test fails at 1.36× book D1 against a mounting hoard) and B84 (the blocker map: founder + affiliates + treasury control the outcome with no catalyst to free it). Both pay up only where the balance sheet does almost all the work.

Claude prices the thin margin one notch higher, an implied buy-below near book, by normalization. C39/C47 (the floor is real but governance-unreachable) and C74/C73 (the buyback tell) drive an implied ¥1,500 — the price at which the observed-trough bear (operating core on FY2023 trough net income F12 plus the ~¥720/sh cash floor D34, landing ~¥1,515/sh) produces ~zero loss and the ex-cash operating yield clears a ~9–10% hurdle even after founder leakage. Claude will pay slightly more than Pabrai because he credits the good operating business explicitly, but still demands the founder-control discount the gap to the stamp represents.

Munger prices the governance block itself and declines to name a number. His M83 (who can tell the CEO no?), M52 (the switched-off buyback), and M90 (the decisive variable is cash deployment) lead him to withhold a buy-below entirely: "the block is governance, not price." His intrinsic-value range only holds if you believe the cash is extractable; under "the cash never comes out," value converges on the operating business alone — essentially the stamp. He waits for behavior, not a mark.

Li Lu prices the knowledge bar and steps aside (too-hard). His L1 gate is neither the floor nor the normalized yield but predictability: he can map the decisive variables (the maturing end-market, the fabless-brand moat, the M&A engine) to disclosures E25E46, but cannot answer them "with the confidence my standard demands." L39 (growth is acquisition-driven, ~5.3% organic D49) and L24 (the founder-trapped hoard at the 30% floor E82) sharpen it. Where the other four found a price that resolves the name, Li Lu found a knowledge gap that no price resolves — the same discipline that separated him from the others earlier in the batch, applied here to a business he judged genuinely unpredictable a decade out. The gap between his too-hard and the others' watch is not a disagreement on the facts; it is a disagreement on whether a low-teens-ROE fabless brand in a maturing market can be underwritten at all.

The red team, engaged

Because four lenses converged on watch (a consensus non-decline), a fresh adversary (ledger only) argued the stronger PASS case — that ¥1,830 (1.36× book D1) is not a "wait for cheaper" but a name to drop, because the quality on display is borrowed of its own cash, peaked, and flattered, so no future price reliably converts this into a sound minority holding. The synthesis must meet its strongest points by name. Each sharpens the crux; the bear is largely right on the facts, and the split is watch-vs-pass on weight.

• "The floor is founder-trapped, and management proved it this year — in the exact window that mattered." The bear's central point, and it is true as stated. With a live ¥7.0bn authorization F507 and standing board power to act E90, FY2026 buyback was ~726 shares / ~¥1.2M E81 while the stock was cheap, and management reissued 4,154,667 treasury shares E127 so the float rose. This is precisely the crux the whole panel already priced — it is why every buy-below sits below the stamp. The rebuttal to pass is that a trapped floor is a reason to demand a discount, not to declare the business un-ownable at any price: unlike a value trap with no operating business under the cash, ELECOM's stub earns ~28.6% ROIC D30 and is not melting, so at book (¥1,350–1,500) the floor plus a good, cheap operating business give a minority a margin it can monetize without the founder — you own the operating business at ~5× and hold the cash as free insurance. The bear proves the floor won't be opened; it does not prove the stub isn't worth owning cheap.

• "The record is a purchase-accounting artifact that normalizes to the aspiration, not a beat of it." Fully conceded, and the panel conceded it first. The ¥7,648M non-cash gain F218E134, non-taxable D18, collapses 21.2% ROE / 7.07× P/E to ~12.5–13.2% D26 / ~11.4–12.0× D22 — dead on the 13% target E31 — and this "record" year's operating cash fell to ¥9,877M from ¥17,354M F63F62. This is exactly why every lens normalizes down and none pays 7× the headline. But a normalized ~13% ROE at ~12× on a fortress balance sheet is a fair price for a fair business — a reason to value on the normalized number, not to conclude the through-cycle business is worthless. The bear's normalization is the panel's normalization; it caps the price, it does not zero the franchise.

• "A maturing, price-taking core with bought growth and roll-up wobble." Conceded on the facts — management-disclosed maturation E25 and daily price competition it can't always pass through E46, ~5.3% organic revenue D49, against a Tescom impairment written to zero value-in-use E124. This is the strongest strand and the one Li Lu weighted to a too-hard. The panel's four watchers answer it not by denying it but by pricing it: a moderate, price-taking business is precisely a watch at a discount to book, not a buy at a full price — which is exactly where the buy-belows land. The roll-up wobble (one below-book steal, one impairment to zero) argues for a bigger discount and against paying up for the M&A engine — which the panel does.

• "You buy the whole entity at 1.36× book D1, not the stub — the ex-cash multiple is a decomposition, not a purchasable security." This is the bear's sharpest logical point, and it is correct. A minority cannot separate the good operating business from the trapped cash and the founder; you own all three, priced together. The panel does not dispute it — it answers it with the discount. At the stamp, the bear is right that the decomposition is a temptation, not a thesis. But at the buy-belows near book (BPS ¥1,341.68 F35) the whole entity is cheap enough that even owning the cash at ~0% and the founder as a fixed cost, the operating business at ~5× ex-cash D40 plus a net-cash cushion of ~39% of the price D35 gives a minority-monetizable margin. The decomposition is not the thesis; the whole-entity price near book is.

• "No reachable discount even at the watch price — a cheaper entry doesn't convert trapped cash or a one-off record into a margin of safety." This is the bear's real challenge to watch itself, and it is the honest tension. The rebuttal, and the reason the panel holds watch rather than pass: the net cash is a genuine downside cushion whether or not it is activated (~24% of the price even after a bear draw), the operating business is good and not melting (unlike the deep-value trapped-floor names earlier in the batch, which had no good business under the cash), and the price is the batch's cheapest with buy-belows at book. So it is a reachable watch at ~book — a good business you would own cheaply while the cash sits as insurance — not a value trap to discard. The bear is right that a lower price does not fix a founder-trapped floor; the panel never claimed it would. The buy-belows assume the bear case (normalized earnings, no unlock) and simply demand to buy the good-through-cycle stub at book.

The honest resolution: the bear is largely right on every fact — the floor is trapped, the record is flattered, the core is maturing and price-taking, the growth is bought — and the split is watch-vs-pass on weight, not on facts. The panel holds watch (and Li Lu holds too-hard, one notch more cautious still) because, unlike the deep-value trapped-floor names the record has declined, ELECOM's operating business is genuinely good (~28.6% ROIC D30, not melting) and the price is the batch's cheapest with buy-belows near book — so it is a reachable watch at ~book, not a value trap. The red team did not move a verdict off watch, but it earned its keep: it is why the synthesis states the crux as sharply as it does — is the floor reachable, and is the record real? — and the answer is that both are exactly what the ~25–30% discount below the stamp is demanding.

Self-distance note. The Claude lens holds one of the five verdicts compared above (watch) and wrote this synthesis; it also built the dual-blind reconciled figure table and evidence ledger all five lenses consumed, and the red-team ran on the same model family. That is an unusual concentration of authorship — the answerer, the ledger-builder, one of the five voters, and the adversary are the same system. Read the synthesis with that in mind.

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.

Verdict accounting (fixed ex-ante)

• A buy-below-¥X verdict is price-falsifiable against the unadjusted stamp. Two lenses issued explicit buy-belows — Buffett ¥1,350 and Pabrai ¥1,350; Claude publishes an implied buy-below of ¥1,500; Munger issues no number (the block is governance, not price); Li Lu's verdict is too-hard, with no buy-below.
• pass / watch / too-hard are recorded but unscored in any future review. Four verdicts here are watch and one is too-hard; the buy-belows sit below the ¥1,830 stamp and are the price at which each watching lens would revisit toward buy.
• 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. (There is no stock split in this company/period; the ¥1,830 stamp and all per-share figures are on the ordinary, unadjusted basis.)

Red team

A consensus red-team (four of five lenses converged on watch, a non-decline) was dispatched to argue PASS — that ¥1,830 / 1.36× book is not a "wait for cheaper" but a name to drop, because the FY2026 quality is trapped behind its own cash (a founder-controlled ¥58bn floor the family just refused to open), peaked (a purchase-accounting record that normalizes to the 13% target, not a beat), and flattered (a ¥7,648M non-cash gain plus a commemorative dividend), so no future price reliably makes it a sound minority holding. Its strongest points, verbatim-faithful, and the synthesis's engagement with each, are in "The red team, engaged" above. In brief, the adversary's three ranked points were: (1) the FY2026 non-buyback-plus-reissue tell — a live ¥7.0bn authorization F507 and standing board power E90, yet ¥1.2M bought E81 while the stock was cheap and 4,154,667 treasury shares reissued E127 so the float rose; (2) the record is a purchase-accounting artifact that normalizes to the aspiration — ¥7,648M non-cash negative goodwill F218E134, collapsing 21.2% ROE to ~12.5–13.2% D26, with operating cash falling to ¥9,877M from ¥17,354M F63F62; (3) a maturing, price-taking core with bought growth and roll-up wobble — market maturation E25, daily price competition E46, ~5.3% organic D49, a Tescom impairment to zero value-in-use E124, and the 28.6% ex-cash ROIC D30 inseparable from the trapped cash and the founder. The synthesis concedes every factual point — the floor is trapped, the record is flattered, the core is maturing and price-taking, the growth is bought — and meets the load-bearing claim (that these make the business un-ownable at any price) by the reachability-of-the-discount test: unlike the deep-value trapped-floor names the record has declined, ELECOM's operating business is genuinely good (28.6% ROIC D30, not melting) and the price is the batch's cheapest with buy-belows near book, so a minority can monetize the discount at ~book without the founder's permission. The bear's own falsifier — a large executed buyback at cheap prices, an activist filing, a special dividend, or two-plus years of clean ~13%+ ROE with rising operating cash D26F63 — is the same observable the panel is waiting on. A consensus that faced its strongest opponent and, on the facts, largely agreed with it — the disagreement is watch-vs-pass on weight, resolved by the reachable discount near book.

What would change our minds

Pre-registered falsifiers, per lens issuing a buy-below or holding a verdict. The verdicts converge on a single observable: does a forcing agent appear — a large EXECUTED buyback, an activist, or a special dividend — to make the ¥58bn floor reachable — or does the price fall to ~book (¥1,350–1,500) where the floor plus the cheap operating business give a margin a minority can monetize without the founder's permission.

• Buffett (watch, buy-below ¥1,350). Toward buy: a price near ¥1,350, where he pays barely more than net cash for the whole ex-cash operating business (~4–5× normalized operating earnings D40D39); or the ~¥720/sh net cash D34 put to work for owners on a dated clock — an activist stake, a payout that stays ≥30% of normalized earnings E82 and a buyback that actually shrinks the fully-diluted count, or a control event. Kill the watch if organic (ex-acquisition) revenue turns down two years running D49 or the effective payout falls again in a normalized-up year E83.
• Munger (watch, no buy-below). The falsifier flips toward buy if the founder deploys the cash shareholder-friendly at scale — a real buyback at cheap prices rather than authorization theater E126, or a clearly-accretive below-book acquisition. It flips toward pass if the founder deploys the ¥58bn hoard into value-destroying acquisitions at full price, OR the ex-cash operating business deteriorates to normalized ROIC ex-cash below 15% for two consecutive years (operating margin sustained below 9%), OR the share count grows >3%/yr via equity-funded M&A without a commensurate normalized-EPS offset.
• Pabrai (watch, buy-below ¥1,350). Buy-below holds only while the ~¥720/sh net cash D34 stays a genuine floor AND the ~7× ex-cash operating business D40 keeps earning ~28.6% ROIC D30. It breaks — sell/pass regardless of price — if the founder bloc E79 uses the cash for a below-return empire-building deal or a related-party transfer rather than owners; if core ex-cash ROIC falls below ~15% or normalized ROE breaks below ~10% D26 as the maturing market E25 and daily price competition E46 compress the model; or if the ¥1,321M refund-liability estimate E146 proves materially incomplete.
• Li Lu (too-hard, no buy-below). The knowledge bar could clear — moving this off too-hard toward a watch — only if the fabless-brand moat becomes namable and durable (a demonstrated, sustained switching cost or shelf-control mechanism rather than speed-and-cost E36), the maturing end-market E25 stabilizes on organic (not acquired) growth D49, and the founder-trapped cash E80 finds a reachable path to owners E82. Absent all three, no price rescues a business he cannot predict a decade out.
• Claude (watch, implied buy-below ¥1,500). A dated, executed capital-return step that shrinks the net float and closes the cash-to-market gap — an actually-executed buyback of ≥¥7bn cancelling shares, or a policy shift to DOE/≥50% payout, in the FY2027/3 disclosures — flips watch → buy at ¥1,830. Conversely, normalized operating income falling below ~¥9.5bn in FY2027/3, or the founder bloc using cash for another below-transparency related deal, confirms the trap and pushes toward pass.

The single observable most lenses converge on is whether a forcing agent makes the ¥58bn floor D32 reachable — a large executed (cash-out, not authorized) buyback E126, an activist E91, or a special dividend that draws down the pile E82 — versus the price falling to the ¥1,350–1,500 range near book F35 where the floor plus the cheap, good operating business D30D40 satisfy each watching lens's discount.

What this taught the checklists

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

• Buffett — add a "controlled-cash / trapped-general" item to the Verdict/asset-play section. The profile handles melting businesses (B39, B103) and hidden separable assets (B92), but ELECOM is a distinct, common Japanese case: a sound, cash-rich, founder-controlled general where the discount is real but structurally inaccessible to a minority. A dedicated item — "controlled-cash discount: is the separable cash reachable by a minority owner, and if not, is the price low enough that you pay only for the operating business?" — would force the access question that B72/B84 today carry only by implication E79E106. Also: a one-line pointer in B42/B52 to "non-taxable negative-goodwill gains from below-book acquisitions" would make the normalization reflex explicit for the Japanese serial-acquirer case F218E134.
• Munger — M53 data-insufficient handler: where consolidated single-segment reporting obscures individual acquisition returns E12, score data-insufficient and add one complexity flag under M14. A new item between M38 and M52 — cash-as-optionality vs cash-as-drag: when net cash exceeds 30% of market cap D35 and the operating business earns >20% ROIC ex-cash D30, ask whether the cash is current optionality (a serial acquirer with demonstrated below-book deal flow) or a permanent drag (no forced-distribution mechanism, no demonstrated buyback discipline E126) — specifically relevant to Japanese net-cash value stocks where the governance discount is structural.
• Pabrai — add a "reachable-floor" test to P1/P17. An asset floor can be marked-down-liquid and still not accrue to a minority because a founder controls its redeployment E79E106. Distinguish an owner-reachable floor (cash on a path to dividends/buybacks/liquidation the minority can force or expect) from a controller-trapped floor (cash a dominant insider stockpiles for M&A E30), and route the "trapped" case to watch even when survival (P20) passes. And sharpen P39/P17 on authorization vs execution: require executed below-IV repurchase over the trailing window and flag announced-but-unexecuted authorizations as non-credit — the ¥7.0bn "buyback" E33 was net-undone by the FY2026 reissue that raised float E127.
• Li Lu — L1/L39 interaction, a "moat-or-maturing" sub-prompt for fabless/brand roll-ups. For a develop-outsource-sell brand, name the switching cost or shelf-control mechanism and cite it, or mark the moat unknown — so the lens does not drift toward "cheap on book" when the durability question is open E36E46. And an L2/L3 archive-depth note: when only a 5-year filing window exists [F1–F5], a one-line reminder that a sub-decade archive caps the worst-case items at data-insufficient prevents over-crediting a shallow record as resilience.
• Claude — two checklist sharpenings, plus a library seed. (1) Sharpen C47 (deployable vs trapped cash) to score two axes — "operationally deployable? / minority-reachable?" — since a name can pass the first and fail the second, as here (fully deployable E64, governance-trapped E106); the two traps have different falsifiers (a repatriation tax vs an activist filing). (2) Sharpen C74 (buyback execution) to check for treasury reissuance in the same window — a buyback executed in year N−1 can be silently reversed by a share-exchange reissue in year N E126E127, so net-float change over a 2-year window is the honest measure. Library (Class-level) — a new corner, the reachability variant of the quality-at-a-fair-price corner: "a genuinely good operating business (high ex-cash ROIC) behind a large, real, but founder-trapped cash floor, priced fair-to-slightly-cheap." This is not the pure-floor pattern (no good business under the cash) and not the EBARA/Daitron quality-compounder corner (reachable returns) — it is the hinge between them, where the floor and the operating business are both real and the verdict turns solely on reachability. First member ELECOM (6750), N=1: normalized ROE ~12.5% D26, net cash 39% of cap D35, buy-belows ~18–26% under the stamp near book. The batch's three names now map cleanly: EBARA JITSUGYO's moat rented, Daitron's owned, ELECOM's floor trapped — three clean quality names at fair prices, all watch (with one too-hard here), none a buy. Where the deep-value corners ask whether a discount closes, this corner asks whether a fair price is worth paying when the floor is unreachable — and it needs its own base rate: for good-operating-business, founder-trapped-floor names at a fair price, how often does the market later offer the ~30% discount at book at which the floor plus the cheap stub become a margin a minority can monetize?

---

Corrections