Abalance Corporation (3856): A Regulatory Window That Closed

Verdicts

Lens	Verdict	Buy below	Most load-bearing items
Buffett	pass	—	B59 cockroach rule; B1/B2 outside circle; B23/B36 no franchise
Munger	too-hard	—	M12 opaque book; M54 character screen; M86 negative lollapalooza
Pabrai	pass	—	P1 permanent-loss floor fails; P49 promoter integrity; P20 leverage kills
Li Lu	too-hard	—	L46 trust in accounting/people; L30 controller alignment; L33 listed entity's claim on cash
Claude	pass	—	C54 NCI leakage worsening; C98 overseas book unknowable; C33 bear EPV brackets stamp

Implied buy-below for the Claude lens: buy-below ¥300 (mid-stress EPV; derived independently of the stamp — see The five lenses below).

The business

Abalance Corporation began as an internet knowledge-exchange site in 2000 E4 and reached its current form by acquisition: WWB in 2011, VSUN in 2020, TOYO in 2024 E5. The Tokyo-listed parent today employs 36 people F60 and is, by its own disclosure, almost purely a holding vehicle — affiliate investments and loans account for 89.8% of its standalone assets E39.

The manufacturing engine is VSUN Solar Energy Company in Vietnam, which Abalance holds at 44.3% indirect voting rights F176, and TOYO Solar Manufacturing, held at 45.4% F177. The business makes solar panels sold primarily to US customers. Revenue grew from ¥26,901M in FY2021/6 F1 to ¥215,284M in FY2023/6 F3 — an 8-fold expansion in two years — on the back of a US tariff exemption for Vietnam-origin panels. At the peak in FY2024/6, North America accounted for 68.8% of revenue D9, or ¥143,783M F157.

Customers kept paying because the tariff window made Vietnamese panels cheap relative to Chinese alternatives. When that window closed — the exemption expired June 2024 E11 and final AD/CVD rates were set E13 — the revenue engine stalled. Then, decided 2026-01-11, US Customs denied VSUN's panels entry under the Uyghur Forced Labor Prevention Act (UFLPA), stopping US sales entirely, stranding inventory, and imposing demurrage costs E66. Estimated loss on the UFLPA disruption: ¥30–50 billion [F248, F249]. The parent did not learn of the denial from its subsidiary until 2026-03-18 — two months after the fact E66.

The company's strategic response is a three-region supply chain: Vietnam manufacturing continues for non-US markets, an Ethiopia cell plant (4 GW capacity, $106.8M of investment across two phases [F252, F253]) addresses the US tariff arbitrage by origin, and a Texas panel plant ($30M Phase 1 F254) targets domestic US assembly. Ethiopian cells showed genuine demand in H1 FY2026/3 E73. Whether the pivot works as a like-for-like revenue substitute for the closed US channel is unresolvable from available filings.

The numbers

Revenue peaked at ¥208,972M in FY2024/6 F4 and dropped to ¥72,417M in the nine-month transitional period FY2025/3 F5 as the US tariff collapse took hold. Operating income fell from ¥23,349M F75 to ¥3,602M F74 in that same period, an operating margin of 4.97% D10. ROE had been 53.8% and 52.4% in the peak years [F44, F45]; it fell to 4.0% F46 in FY2025/3. H1 FY2026/3 revenue recovered to ¥58,662M F200 with operating income of ¥6,274M F201, but ¥5,885M of special gains were embedded in that period — including a ¥4,306M gain on fixed-asset sales F230 and a ¥1,410M reversal of the export-tariff provision F133 — so the underlying operating picture is weaker than headline suggests.

The structural feature that matters most: in FY2024/6, of ¥20,203M total consolidated net income F87, ¥10,673M — 52.8% D6 — went to non-controlling interests F86. In H1 FY2026/3 that fraction worsened to 66.2% D7. Parent-attributable net income was ¥9,530M at the peak F14 and fell to ¥951M in the nine-month period F15. The listed entity holds a minority economic slice of the manufacturing operations.

The balance sheet carries ¥56,318M of total borrowings F125, of which ¥32,310M is short-term F104. The rolling structure turned over ¥41,009M of new short-term borrowings against ¥35,850M of repayments in the nine-month period alone E23. Assets pledged as security total ¥38,634M against ¥50,352M of secured debt [F134, F135] — collateral is already underwater against the secured claims. An export-tariff provision sat at ¥16,470M at FY2025/3-end F130, equivalent to 38.0% of total net assets D8; it declined to ¥11,638M at September 2025 F132 as partial cash settlements were made. Vietnamese tax authorities may disallow the deduction for this provision, creating an additional unestimable liability [E37, E80].

The stamp price of ¥515 implies a market cap of ¥9,760M D1. The equity-holder BPS at September 2025 is ¥1,956 D4, giving a P/B of 0.263× D5 on an attributable basis. Against the last audited BPS of ¥1,358.44 F31 the stamp P/B is 0.379× D2. The equity-holder equity itself — ¥37,070M at September 2025 F215 — rests on an uninvestigated overseas balance sheet, and the auditor who reviewed it issued a disclaimer of conclusion E76.

The five lenses

Buffett: Pass

The starting question: what does this company sell, who pays, and why do they keep paying? Abalance sells commodity solar panels from a Vietnamese factory it only partly owns. Customers paid because a US tariff exemption made the product cheap. That exemption ended in June 2024 E11. The UFLPA denial arrived in January 2026 E66. The business in one sentence: a commodity manufacturer with one dominant channel, one regulatory permission slip, and a structure where the majority of profits flow to someone else.

The economics looked spectacular while the window was open. Revenue grew eight-fold in two years [F1, F3]. ROE reached 53.8% and 52.4% [F44, F45]. But of FY2024/6's ¥20,203M of consolidated net income F87, ¥10,673M — 52.8% D6 — went to non-controlling interests. By the nine-month FY2025/3, the leakage rose to 68% of the much smaller pie. You own a holding company collecting a minority share of a minority-owned manufacturer.

Then overlay the balance sheet: ¥56,318M total borrowings F125, ¥32,310M short-term F104, ¥50,352M of secured debt against ¥38,634M of pledged assets [F134, F135]. An export-tariff provision of ¥16,470M F130 with unresolvable Vietnamese tax deductibility E80. An ongoing UFLPA-related loss of ¥30–50 billion [F248, F249] booked against withdrawn guidance E69. And a cockroach kitchen.

B59 closes the case. A third-party committee found "intentional and organized fraudulent accounting" — same-day circular cash flows between WWB and construction contractors with no economic substance [E43, E44]. The company disclosed this to the public as mere "error" — the committee said this may have violated FIEA Article 158 E46. The company's own rebuttal committee called the fraud conclusion "extremely crude" E55, but even that friendlier body found no evidence that intent was absent, and separately documented that 龍潤生 borrowed from WWB to fund his own share subscription during the investigation, raising fictitious-payment concerns E56. The TSE designated the stock Security on Alert for "insufficient awareness of group governance" E60. The SESC conducted an on-site inspection in May 2025 E52. The auditor issued a disclaimer of conclusion on H1 FY2026/3, citing undiscovered misstatements possibly "material and pervasive" E76.

The price looks cheap against book. But a melting-bargain check (B65) requires NCAV that is stable and growing while you wait. Operating cash flow was negative ¥10,361M for nine months F51. Current assets of ¥89,038M F96 are set against total liabilities of ¥102,463M F109 at FY2025/3 — no positive NCAV exists even before the UFLPA losses. The audit disclaimer means we cannot verify the numbers anyway.

The ten-year ownership test answers itself. I would not buy a farm whose crop was legally blocked from sale, whose title was under investigation for falsification, and whose harvest numbers an independent accountant could not certify. Verdict: pass — not because the price is necessarily wrong, but because I cannot verify what I am buying.

What a student should take from this: A low P/B is only meaningful when the book has been honestly computed and is stably retained. Here the auditor could not form a conclusion, the structure leaks value to minority holders, and the assets face unresolvable tax and legal contingencies. Cheapness measured from an unreliable balance sheet is not cheapness — it is an invitation to lose money with the comfort of arithmetic.

Munger: Too-Hard

I start by inverting. How does this business die? I count the kill paths before reading a single revenue line.

First path: the US market closes permanently. North America was ¥143,783M of FY2024/6 revenue F157 — 68.8% of total D9. The UFLPA denial stopped US sales entirely E66 and the parent learned of it two months late E66. An import ban that could have been appealed in January was instead met with raised guidance in the same two-day window E73.

Second path: the accounting cannot be trusted. A 2024-03 restatement reduced FY2023/6 revenue by ¥2,130M and net income by ¥479M [F250, F251]. The third-party committee found "intentional and organized fraudulent accounting" [E43, E44]. The original auditor resigned citing inability to assemble audit quality E36. The replacement auditor issued a disclaimer of conclusion on H1 FY2026/3, not a qualified opinion — a disclaimer — finding undiscovered misstatements possibly "material and pervasive" E76. A fourth investigation is running on the Abit case, where a transaction was booked with a counterparty that did not legally exist E62. FY2026/3 results are delayed 62 days E70. At the stamp date I cannot read audited financials for the current year. A book I cannot reconstruct routes to too-hard — full stop [M12].

Third path: the structure leaks all owner-attributable economics. Even in the peak year, the parent captured only 47.2% of consolidated net income; by H1 FY2026/3 that fraction had fallen to 33.8% D7. VSUN and TOYO — both held at 44–45% indirect [F176, F177] — generate the revenue, but Abalance shareholders are permanent minority participants in their economics.

Fourth path: the founder's governance is terminal. The third-party committee recommended 龍潤生 promptly resign and that his shareholder influence be eliminated E50. The friendly verification committee agreed E59. What happened? He borrowed from WWB to fund his own share subscription — the verification committee called this "outrageous" E56. Post-AGM June 2026, he stepped from representative director to plain director but remains on the board E72. He chairs every major operating subsidiary E27 and is the only person who holds information about the overseas business generating over 90% of revenue E57.

The negative forces are not merely additive — they compound. Six independently-sufficient problems: unauditable accounts, UFLPA ban, governance and character failure, NCI leakage, commodity economics with no moat, and an SESC investigation. Cheap on unverifiable figures is not cheap. At no price does "I can't read the accounts" become investable.

Verdict: too-hard. Falsifier: clean audited FY2026/3 accounts, UFLPA ban lifted or Texas/Ethiopia capacity operational at similar margins, 龍潤生 no longer a director or significant shareholder, no further restatements — only then does the price question become meaningful.

What a student should take from this: When the accounting is in dispute, the multiple is irrelevant. A low P/B on an unverifiable book is not a margin of safety — it is an invitation to guess. Governance failure and accounting failure travel together. Regulatory-window economics look like moats in the revenue table but vanish the moment the window closes.

Pabrai: Pass

Start where Dhandho starts: tails. What happens if everything goes wrong?

Cash and deposits at September 2025: ¥36,213M F240. Gross short-term borrowings: ¥34,775M F241. Add long-term borrowings ¥10,381M F108, installment payables ¥10,402M combined [F120, F121], and lease obligations ¥1,715M [F122, F123], and total financial debt approaches ¥57bn. Net cash is deeply negative. Pledged assets ¥38,634M secure ¥50,352M of debt [F134, F135] — secured claims exceed pledged collateral. Frontline could sell one ship and walk away. Abalance cannot sell the Vietnam factory piecemeal; VSUN is 44.3%-owned indirectly F176 and secured creditors would see proceeds first.

Then: export-tariff provision ¥16,470M at FY2025/3 F130, declining to ¥11,638M at September 2025 F132, with Vietnamese tax authorities potentially disallowing the deduction [E37, E80]. UFLPA loss estimated ¥30–50 billion [F248, F249]. Guidance withdrawn entirely E69. No audited current financials; an 80-page disclaimer from the auditor instead of a review conclusion E76.

The financing structure is the Horsehead problem and the Delta Financial problem in one package. In the nine-month FY2025/3 period, ¥41,009M of short-term debt was drawn and ¥35,850M was repaid — a continuous rolling structure E23. If any one bank gets nervous during the accounting investigation, the music stops. With roughly ¥57bn of gross debt against ¥36bn of cash, the company cannot survive a credit-market closure of even a few quarters.

P1 arithmetic: market cap ¥9,760M D1. Book equity attributable to shareholders ¥37,070M at September 2025 F215 — superficially a 4× asset cushion. But strip away what I cannot recover: goodwill ¥3,899M F99 haircut to zero; finished goods ¥16,618M F93 being sold outside the US at lower prices E67; export-tariff provision ¥11,638M F132 plus open tax exposure; UFLPA mid-loss ¥4,000M; and ¥50,352M of secured creditors standing ahead of equity on ¥38,634M of pledged assets [F134, F135]. The stressed liquidation value available to equity is near zero or negative. "Tails, I don't lose much" does not apply.

On the governance and compensation gate: all-fixed pay E34 — ¥47M total for four executive directors F192 — with no performance linkage, no remuneration committee E30, and individual pay delegated to the president's own discretion E35. The gate is moot here: P1 fails first, and P1 alone routes to pass regardless.

Verdict: pass — unconditional. No buy-below.

What a student should take from this: Optical cheapness (P/B below 0.4) is not a margin of safety. Safety requires a hard asset floor net of all claims, which demands honest accounting of who stands in front of the equity holder. When the auditor cannot issue any conclusion at all, you cannot even measure the floor.

Li Lu: Too-Hard

I have sat with this company for a long time. Let me be honest with you.

The knowledge bar is the decisive gate. To predict this company's earnings power in ten years requires knowing: whether the UFLPA ban is permanent or reversible on what timeline; whether the Ethiopia-Texas supply chain produces the tariff advantage claimed; what the Vietnamese tax authority will do with the export-tariff deductions; what corrections to prior-period financials the auditor will ultimately require; what the SESC proceeding produces; whether the 55–56% minorities in the operating subsidiaries will be managed fairly toward the listed entity going forward; and whether the organizational culture shaped by a decade of "Ryu's shop" can be rebuilt under new management E48. None of these can be answered from the ledger. Several are unknowable even to insiders.

L46 closes the case: trust in the accounting and the people. An interim review disclaimer E76, a prior-year restatement from organized circular invoicing E44, a live SESC proceeding E52, an auditor resignation E36, a possible FIEA Article 158 violation E46, and an Abit fictitious-customer transaction E62. These are not degrees of qualification — a disclaimer means the financial statements cannot be evaluated at all.

L30 compounds it: controller alignment and the ten-year minority record. 龍潤生 chairs every major operating subsidiary E27, is documented as the sole holder of overseas subsidiary information E57, borrowed from WWB to fund his own share subscription during the fraud investigation E56, and was recommended for removal by both committees [E50, E59].

L33 completes the picture: the listed entity's claim on the cash. The parent is a holding shell, 89.8% of standalone assets in affiliate investments and loans E39. VSUN and TOYO are 44–45% held [F176, F177]. Non-controlling interests absorbed 52.8% of FY2024/6 net income D6 and 66.2% of H1 FY2026/3 net income D7. H1 NCI dividends paid out ¥3,296M F236 — flowing to the 55%+ of the operating subsidiaries Abalance does not own.

A cheap price does not compensate for ignorance. The boundary of competence is not a valuation floor; it is a prior condition that must be met before valuation is meaningful. I would rather sit still.

Verdict: too-hard.

What a student should take from this: When a company is cheap because of a crisis, the first question is not "is the price low enough?" but "can I trust what I am reading?" Fraudulent accounting does not become safe to analyze at a lower price — it becomes a deeper unknown. A live SESC investigation, a disclaimer rather than a review conclusion, and an auditor running retrospective fraud procedures are the clearest possible signals that the ledger is not yet reliable enough to reason from.

Claude: Pass — Implied buy-below ¥300

The outside view was registered figures-blind before the ledger was opened. Reference class: below-mid-cap Japanese holding company, overseas manufacturing in a commoditizing sector, export-channel disrupted, live accounting investigation, founder-dominant ownership. Bottom-quartile survival probability pre-registered at 0.40; delisting probability at 0.25. The ledger updates both.

What the ledger confirmed. The equity cushion is larger than the priors implied: equity-holder BPS ¥1,956 D4 gives a P/B of 0.263× D5 at the stamp — far cheaper than the inside view suggested. The known quantitative impact of the domestic accounting crisis is not balance-sheet-threatening by itself: the restatement removed ¥2,130M from FY2023/6 revenue F250 and ¥479M from net income F251; the Abit fictitious-customer transaction had a net P&L impact of ¥0.2M E64.

What moved in the wrong direction. NCI leakage is worse than the priors registered. In FY2024/6, 52.8% of consolidated net income accrued to non-controlling interests D6; by H1 FY2026/3 that fraction rose to 66.2% D7. Parent-attributable net income peaked at ¥9,530M F14 and fell to ¥951M in the nine-month period F15. The US-channel disruption is total: North America was ¥143,783M of FY2024/6 revenue F157 and is now effectively zero. H1 FY2026/3 operating cash flow of ¥32,723M F225 looks exceptional but includes ¥16,729M of contract-liability increase [F243, F242] — advance customer payments not yet earned — so underlying H1 cash generation is roughly ¥16,000M. The forensic read on overseas integrity is impossible: the third-party committee's mandate explicitly excluded overseas subsidiaries covering ~95% of revenue E51, and the auditor issued a disclaimer on the consolidated H1 statements E76.

Bear-case EPV (C33). Revenue ¥60bn (non-US annualized with Ethiopia), operating margin 5%, tax 25%, 65% NCI leakage → parent NOPAT approximately ¥788M; EPV at 10% hurdle = ¥7,875M / 18.95M shares = ¥416/share [D1, D5, D7, F176, F177]. Stress scenario (3% margin, 12% hurdle): approximately ¥208/share. The stamp of ¥515 sits at or above the optimistic end of the bear EPV range.

Implied buy-below: ¥300/share D5. Derived from mid-stress EPV: ¥60bn revenue × 4% operating margin × 75% after-tax × 35% parent take after 65% NCI leakage / 11% hurdle = ¥5,727M / 18.95M shares = ¥302/share, rounded to ¥300. This derivation used no reference to the stamp price. The implied buy-below is 42% below stamp and makes no allowance for potential overseas restatement, the export-tariff-deductibility tax risk E80, or further UFLPA litigation costs. Even at ¥300, the bear EPV requires the overseas balance sheet to hold, the non-US pivot to generate 4%+ operating margins on ¥60bn+ revenue, and NCI leakage not to worsen further — three conditions none of which are verifiable from available documents.

Jury result. The jury of three deliberate selves on C33 (bear EPV), C35 (private-owner yield), C84 (short case), and C98 (epistemic limits) found divergence across all four eligible items. The bull-framed self reached above-stamp EPV using H1 annualized revenue; the bear-framed self reached well-below-stamp using non-US-only annualized figures. The divergence confirmed pass: the stamp sits within the bear EPV range (¥208–416), offering no safety margin at any plausible revenue or margin assumption. The C98 jury found the single structurally unknowable unknown — overseas balance-sheet integrity — load-bearing enough that no watch verdict is warranted before the FY2026/3 retrospective audit produces a clean opinion. The C89 anchoring check passed: the implied buy-below derivation was conducted without reference to the stamp.

Verdict: pass. Implied buy-below ¥300.

What a student should take from this: A low P/B and a low P/E are necessary but not sufficient for cheapness. When NCI leakage consumes the majority of consolidated earnings — 66.2% in H1 FY2026/3 D7 — the optical multiple severely overstates what the controlling-share buyer actually purchases. The correct denominator is parent-attributable earnings on the fully-diluted count, not the consolidated headline. The investigation mandate matters as much as the investigation finding: a committee that covers only the domestic entity — roughly 5% of revenue — and issues a clean bill produces no information about the overseas operations it did not examine E51.

Synthesis

Where the five lenses agree. All five lenses declined. The three grounds of agreement — reached independently — are: (1) the accounting cannot be relied upon (auditor disclaimer on H1 E76; FY2026/3 annual results delayed E70; overseas book uninvestigated E51); (2) the NCI leakage structure means Abalance shareholders own an economically thin slice of the manufacturing operations — not the full consolidated book [D6, D7, F176, F177]; and (3) the UFLPA import denial has stopped the dominant revenue channel with consequences that are not bounded or timestamped [E66, F157, D9].

Where they differ. The three pass lenses (Buffett, Pabrai, Claude) and the two too-hard lenses (Munger, Li Lu) diverge on emphasis, not on the fact of decline. Buffett reaches pass through the cockroach rule (B59) and circle of competence (B1): the fraud finding and audit disclaimer end the analysis before valuation. Pabrai reaches pass through P1: the stressed liquidation floor is near zero or negative, so "tails, I don't lose much" fails regardless of the stated book discount. Claude reaches pass after completing a full owner arithmetic: the bear-case EPV range brackets or falls below the stamp price [D1, D5], and even the bull-framed jury self could only reach above-stamp EPV using optimistic revenue and margin assumptions.

Munger and Li Lu reach too-hard because the accounting is not merely questionable — it is literally unauditable. M12 (opaque book) and L46 (trust in accounting and people) are categorical failures that mean the question of price is not yet meaningful, irrespective of where that price sits relative to book. The difference between pass and too-hard here is not a valuation disagreement; it is a sequencing judgment. Munger and Li Lu say: resolve the accounting first, then ask about price. Buffett, Pabrai, and Claude say: the accounting failure is itself the reason to pass, and passing is immediate.

Both sequences produce the same output: no buy.

The exchange between the masters. Buffett's observation on the cockroach rule prompts Munger to sharpen it: one cockroach in the kitchen and we exit; here we have six compounding ones. Pabrai adds the Horsehead parallel — the Delta Financial failure mode arrives not from fundamental insolvency but from the rolling short-term funding machinery stopping E23. Li Lu frames the knowledge bar: unlike a cyclical trough where the fundamentals are knowable and patience is rewarded, here the unknown is structural — only 龍潤生 holds the overseas information E57 — and no amount of patience resolves what has not been investigated. Claude's contribution is arithmetic: even granting every favorable assumption in the bull case (non-US revenue recovering to ¥60bn, margins restoring to 5%, NCI stabilizing), the bear EPV barely reaches ¥416/share [D5, D1, D7] against a stamp of ¥515. There is no comfortable entry price at which the five simultaneously-unresolved uncertainties wash out.

Self-distance note. All five lenses in this study ran on a single model family (Claude Sonnet) in one session, completed in-session after the headless batch crashed on a session limit mid-stage-4. The consensus is therefore single-family — a hypothesis the red team stress-tests below. The Claude lens's §1 outside view was genuinely figures-blind: a separate context that never received figures.md registered priors to profiles/claude-priors.md before any ledger exposure, satisfying the isolation protocol for that stage. However, the four master-voice lenses (Buffett, Munger, Pabrai, Li Lu) and the synthesis were all written in the same session against the same ledger. Treat the agreement between them as within-session corroboration, not independent multi-model confirmation.

Governance and compensation gate. Compensation and governance data are present in the ledger: officer cash comp ¥47M for four directors F192 and ¥8M for five outside officers F193; all-fixed structure with no performance link E34; individual pay delegated to the president without a remuneration committee E35; governance structure including absence of nomination/remuneration committees [E30, E31] and no formal independence criteria for outside directors E31. The gate is satisfied (data present). Quality assessment: catastrophic — but the gate requires presence, not quality, and all five verdicts are at pass or too-hard, so the gate caps nothing.

Verdict accounting (fixed ex-ante)

• 3 pass / 2 too-hard — no buy-below issued by any lens. (The Claude implied buy-below ¥300 is an analytical output, not a scored verdict — no lens crossed the watch threshold.)
• 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.

Red team

The red-team memo steelmans the bull case. Its strongest points, engaged in their own terms:

"Cash covers market cap almost four times over" [F240, D1]. The red team correctly observes that cash and deposits at September 2025 of ¥36,213M F240 exceed the market cap of ¥9,760M D1 by a wide margin. The engagement: the cash sits mostly inside VSUN and TOYO, entities 44–45% owned [F176, F177]. The parent cannot freely extract it: secured debt of ¥50,352M has prior claims on ¥38,634M of pledged assets [F134, F135]; ¥11,638M of the export-tariff provision F132 represents real cash obligations; and the short-term debt of ¥34,775M F241 at September 2025 is a rolling daily obligation. Net deployable cash after netting short-term debt against gross cash is approximately ¥1,400M — not a cash box, but a working-capital float.

"The manufacturing asset base is not in the market cap" [F174, F175, F97]. The Ethiopia plant absorbed $106.8M of capex [F252, F253] and delivers 4 GW of capacity E24. At ¥9,760M market cap D1 the factories appear free. The engagement: at 44.3% indirect ownership F176, Abalance cannot monetize VSUN's ¥35,646M of net assets F174 without minority consent and without secured creditors' consent on pledged assets. More fundamentally, VSUN's ¥35,646M net assets figure is single-sourced from the yūhō subsidiary table and unverified: the investigation excluded overseas subsidiaries E51 and the auditor disclaimed the consolidated H1 E76. The "manufacturing asset base" may be exactly what the audit would confirm — or may contain undetected misstatements the auditor explicitly flagged as possible E76.

"Non-US revenue is real, growing, and survives the ban" [F200, F206, F242]. H1 FY2026/3 revenue of ¥58,662M F200 was 28.6% above the prior-interim ¥45,613M F206, and contract liabilities tripled to ¥25,361M F242 — evidence of advance customer orders. The engagement: contract liabilities are cash received for goods not yet delivered. They inflate H1 OCF F225 by ¥16,729M of advance payments [F243, F242] that will either convert to revenue on delivery or be returned. Underlying H1 operating cash generation was approximately ¥16,000M — still positive but not the ¥32,723M headline. The Ethiopia cells generating H1 demand E73 were cited in the same disclosure that raised guidance while management did not yet know (or had not acted on) the UFLPA denial [E66, E73].

"The export-tariff provision is a declining liability, not a fixed bomb" [F131, F130, F132]. The provision fell from ¥18,447M F131 to ¥11,638M F132 over 15 months — a 37% reduction. The engagement: the remaining ¥11,638M F132 is a real cash obligation, and Vietnamese tax authorities have not yet resolved whether the deduction is allowed [E37, E80]. If disallowed, the provision generates an additional tax liability that is explicitly described as unestimable. The rate of decline is real; the remaining risk is not zero.

"The fraud finding is genuinely contested" [E43, E44, E53, E54, E55]. The verification committee rejected the "intentional and organized fraud" conclusion as "extremely crude" E55. The engagement: even the friendlier committee found governance awareness "markedly lacking, unchanged to this day" E54, called the WWB loan to fund 龍's subscription "outrageous" E56, and agreed that management must be urgently renewed. The two committees do not disagree on whether problems exist — they disagree only on the label (fraud vs. negligence). Either label routes to the same investment conclusion: the accounts are unreliable and the controller's conduct is disqualifying.

"A new management team is now in place" E72. Yanase installed as president, 龍 reduced to non-representative director, three new outside AC directors E72. The engagement: this is the fourth presidential change since 2023 E28. 龍 remains a board director and 24.7% shareholder F189. He chairs every major operating subsidiary E27 and remains the overseas information black box E57. A change in the Tokyo title does not restructure the overseas authority that drives 90%+ of revenue.

The one bull point neither the synthesis nor any lens directly rebuts. The red team's weakest acknowledgment — but strongest asymmetry argument — is that the stock is so neglected (80.2% individual holders F191, TSE Security on Alert E60, zero institutional anchor) that even a partial resolution of the multiple crisis layers simultaneously could produce a large rerating. The lenses do not dispute this. The issue is whether any one resolution becomes likely enough to stake capital on. The answer from four of five lenses: not without a clean audited book first.

What would change our minds

Since no lens issued buy-below or watch, these are the observable conditions that would move Abalance from avoid/too-hard to investable — pre-registered now against the stamp conditions.

1. Clean audited FY2026/3 accounts. The FY2026/3 annual audit (targeted by 2026-08-31 per the 62-day extension E70) issues a clean or merely-qualified (not disclaimed) opinion with no material restatement of VSUN/TOYO net assets below the ¥35,646M reported at FY2025/3 F174. This is the single necessary condition for any re-evaluation; it resolves the structurally unknowable item from C98. Falsifier for the bear case: if audited VSUN net assets are impaired by more than 20%, permanent impairment is confirmed and the verdict deepens.

2. UFLPA ban lifted or Ethiopia/Texas capacity operational at equivalent margins. Either CBP reverses the VSUN denial E68 or the Ethiopia-Texas three-region supply chain E74 demonstrates that Ethiopian-origin cells have no UFLPA exposure and the Texas plant reaches commercial production at margins competitive with the pre-ban North America channel. Revenue recovery toward ¥150bn annually with operating margin above 7% on that base, audited, would move the bear EPV above the stamp.

3. 龍潤生 exits the board and divests materially. The third-party committee E50 and the verification committee E59 both recommended removal and reduction of his shareholder influence. If he exits the board entirely AND divests below 10% of outstanding shares, the governance overhang compresses materially. The overseas information black box E57 must also be addressed: verifiable independent oversight of VSUN/TOYO is needed, not merely a Tokyo title change.

4. Verified overseas balance sheet. An independent, unrestricted audit of the overseas subsidiaries — VSUN, TOYO SOLAR, TOYO — specifically covering the 95% of revenue that the third-party committee explicitly excluded E51. Until this exists, the equity-holder equity figure of ¥37,070M F215 is an unverified claim, and all valuation arithmetic that uses it carries a structural disclaimer.

None of (1)–(4) is visible at the stamp date. Conditions (1) and (4) are necessary before any buy-below can be issued. Conditions (2) and (3) are necessary for the buy-below to exceed the ¥300 implied level.

What this taught the checklists

Buffett checklist (0.4.0):

1. NCI leakage screen (gap item). No dedicated item catches majority-revenue, minority-owned manufacturing subsidiaries where NCI absorbs 50–66% of reported net income. The holding-company-minority trap would be partially caught by B104 (sum-of-parts) for conglomerates but the checklist has no explicit pass/fail for the case where the primary operating subsidiary is a specified sub under majority-other-shareholders and the leakage is structural. Propose adding a B-series item between B48 and B49 requiring: for businesses where consolidated net income exceeds parent-attributable income by more than 30%, compute and table NCI leakage across five periods and state the parent's actual economic claim. Abalance illustrates that peak ROEs of 53.8% and 52.4% [F44, F45] coexist with ¥951M of parent-attributable income F15 on ¥72,417M of revenue F5.

2. Regulatory-access moat flag (B25 gap). The apparent cost advantage that sustained above-commodity returns for years was tariff-exemption access — neither a low-cost production advantage nor a brand. A note should be added to B25 explicitly asking whether the cost advantage rests on regulatory access (quota, exemption, certificate) that can be administratively revoked. If yes, treat as no moat regardless of historical margins.

3. Audit disclaimer as automatic verdict trigger (B59 enhancement). B59 currently asks about restatements, regulator accounting actions, and documented integrity lapses. An auditor disclaimer of conclusion is substantively more serious than a restatement — it means the auditor cannot determine whether the accounts require adjustment at all. The item should explicitly name a disclaimer as an automatic hard-fail routing to pass, not merely to a burden-shifting watch.

4. NCI-inclusive vs equity-holder BPS (B89 clarification). The net-net check uses current assets minus all liabilities. When consolidated net assets include large non-controlling interests [F116, F239], the apparent book discount on the headline P/B ratio overstates the parent's claim. The item should explicitly require deducting NCI before computing the per-share bargain.

Munger checklist (0.4.0):

1. M2 (Permanent-loss map) should explicitly flag the NCI-strip as a permanent-loss vector. The current item covers debt and dilution but not the structural redirection of earnings to non-controlling interests that the parent shareholder can never recover. In cases where NCI captures more than 50% of consolidated net income D6, this belongs in the map.

Pabrai checklist (0.6.0):

1. P12 (Revenue concentration) — geographic/regulatory sub-question. The current item tests customer-level concentration. Abalance had no disclosed customer above 10% but 68.8% of revenue D9 depended on a single regulatory regime — North America ¥143,783M F157 — that closed. Propose adding: "What share of revenue depends on a single regulatory permission or tariff regime, and what happens if that permission is revoked?"

2. P23 (Funding-model dependence) worked cleanly. The ¥41bn short-term rolling structure E23 was identified directly. No revision needed.

Li Lu checklist (0.4.0):

1. L46 should flag the disclaimer outcome explicitly. The "Good" criteria should add: a disclaimer of conclusion is a categorical fail, not a degree of qualification — it ends analysis regardless of price.

2. Add an explicit item for the NCI structural claim. L33 asks whether the listed entity carries an enforceable claim on operating cash flows. A sub-question — "What fraction of consolidated operating income is retained by the parent after NCI distributions, and what is the mechanism by which that fraction reaches shareholders of the listed entity?" — would surface the 44.3% indirect ownership of VSUN F176 and 45.4% of TOYO F177 before any accounting assessment.

Claude checklist (0.4.0):

1. C17 (Manufacturing unit economics) — add throughput-unit sub-item. Future studies of manufacturing businesses should explicitly request GW of nameplate capacity vs. actual production vs. revenue/GW.

2. C46 (Year-by-year debt maturity) — flag data-insufficient explicitly. If the annual maturity schedule is absent from the borrowings table, flag this and request it. Without the ladder, rollover-cliff analysis is impossible.

3. C55 (Listed affiliates) — TOYO Nasdaq price check as standard step. TOYO Company Limited is Nasdaq-listed; a price check would quantify the after-tax mark on the 45.4% stake F177. Stage this as a standard step at the source gate before the study begins for companies with listed subsidiaries.

---

Corrections

Append-only below this line. Use templates/correction.md. Never edit above it.