The Silent Lien

The market prices Korean semiconductors as blended AI-beta: an HBM cash engine whose surplus is freely allocable between distributions and capacity, with the won an exogenous overlay applied after the allocation decision. The actual variable is a senior, unfiled claim on that surplus — denominated not in node class but in domestic footprint, and serviced through the treasury's dollar-conversion decision rather than the dividend line. This report calls it the Silent Lien: the capital commitment that ranks ahead of minority distributions in practice while appearing in no covenant, no filed liability, and no balance-sheet line.

The lien is constructed in three moves and settled in a fourth. It is originated when the capital committee converts an unwritten obligation to build, convert and staff at domestic scale into a capex schedule that no longer flexes with the return calculation (§II). It is serviced when the treasury elects how, and whether, to convert the dollar receipts that fund it — an instrument choice filed disclosure now adjudicates directly (§III). It echoes when that conversion decision enters a currency whose flow map is dominated by structural, non-corporate participants (§IV). It is settled, or deferred, where a statutory clock meets a distributable residual that arithmetic has already fixed (§V).

The lien reframes the axis. Consensus, bullish and bearish alike, argues about which node class earns its cost of capital. The Q1 2026 disclosure record attributes the dominant capital line not to legacy sustenance but to advanced-process expansion, conversion and infrastructure — the binding commitment is locational and institutional, not node-specific. What is pledged is the domestic footprint. What is subordinated is the residual.

Key Takeaways & Risk Boxes

The lien is locational, not node-class. Q1 2026 group capex of 11.23 trillion won included 10.19 trillion won in device solutions — roughly 90% of the total — with the filed narrative attributing it to advanced-process expansion, conversion and infrastructure. Confirmed The commitment defended is domestic build scale, not a node's return.

Filed evidence contradicts the forward-hedging mechanism. One duopoly issuer discloses current forward-contract assets of 194 million won against liabilities of 171 million won, while pre-tax profit sensitivity to a 10% currency move is 2,953,682 million won. Hedge accounting is not applied. Confirmed Ring-fencing runs through retention, not forward-selling.

Netting does not neutralise the exposure. The natural hedge from dollar equipment and materials offsets an estimated 20–50% of receipts Scenario, leaving net exposure at 50–80% of gross dollar receipts Scenario. Most of the flow survives netting — the retention channel is larger than a "fraction of gross" framing implies.

Sign correction, published. Won weakness expands disclosed pre-tax profit. Any chain asserting margin compression via currency must route through imported energy and industrial tariffs, not revenue hedging. Backtest window 2010–2026 supports the energy-and-tariff route. Scenario

The distribution ceiling is arithmetic — and not yet binding. Q1 2026 operating cash flow of 26,330,119 million won against investing outflows of (17,634,899) million won Confirmed leaves a derived residual of 8.70 trillion won, more than fourfold the upper bound of the 0.5–2.0 trillion won programme-consistent gap band Scenario. The ceiling binds in the engine-normalisation regime, not in the observed quarter.

The statutory clock pre-empts the paralysis it was assumed to await. The Commercial Code Third Amendment took effect 2026-03-06, imposing rolling one-year cancellation clocks on newly acquired treasury shares and a fixed 2027-09-06 deadline for pre-existing holdings. Confirmed HOUSE-ANCHOR Forced action arrives on statutory time — but cancellation is non-cash at execution, which is why it coexists with the lien.

The load-bearing wall is a legislative variable. A direct-subsidy regime at 5–15 trillion won annually Scenario externalises the mandate to the fiscal ledger. The internal cross-subsidy persists only while that valve stays metered.

I. The Consensus Ledger: Two Camps, One Shared Premise

Consensus here is not monolithic. The first camp prices the complex as an AI-beta vehicle: HBM volume and pricing carry earnings, foundry losses are openly flagged rather than ignored, domestic mature capacity is optionality on industrial recovery. The second — the trailing-node camp — argues the reverse: wafer crowd-out into advanced products created genuine trailing-node shortage, re-rating legacy from drag to profit engine. That view was itself crowded inside the current window.

Attacking either on its own terms produces a strawman. What both hold, and neither prices, is the premise beneath the disagreement: that capital allocation is a free corporate optimisation, and the currency an exogenous overlay applied to the result. Both assume cash is fungible at management's discretion. Both assume the won happens to the income statement rather than being partly produced by it. The Silent Lien attacks that shared premise — allocation constrained by a state-adjacent unfiled obligation, and the currency as, at the margin, an output of treasury behaviour.

The bridge is narrow. If allocation is free and the currency exogenous, the risk metric is memory pricing. If allocation is encumbered and the currency partly endogenous, the risk metric is the residual after the lien is serviced — governed by disclosure lines, statutory clocks and a fiscal valve, none of which appear in a memory pricing model.

II. The ROIC Ledger: Node Class Is the Wrong Axis

Neither duopoly issuer discloses return on invested capital by node class. No segment line isolates trailing-node from advanced memory; no capex line separates sustenance from conversion. Any claim that a node class earns below its cost of capital is an inference on a proxy stack, labelled accordingly rather than promoted to fact. The stack has four rungs: segment capex and depreciation disclosures; utilisation inference from disclosed output and pricing commentary; allocation of shared cleanroom capital by wafer-start share; and a post-credit tax adjustment at the national strategic technology facility-investment credit rate of 10–25% Scenario. The third rung breaks: shared cleanroom capital cannot be allocated by node without an assumption the issuer does not publish. SPECULATIVE — direct entity disclosure for node-class spreads does not exist in the public domain.

The falsifying disclosure line is easy to name and does not exist: a node-class capex or segment-return breakout in the periodic filings. The comparison must therefore be run band against band, not point against point. Post-credit blended legacy-node return sits at 4–8% Scenario against a won-denominated hurdle of 4.5–6.0% Scenario SPECULATIVE. Those bands overlap across their entire mutual range. The honest verdict is that "legacy is structurally sub-WACC" holds only in the lower half of the return band and fails in the upper half — an indeterminacy, not a finding, and the report is written to survive both halves.

Two rate observations push the hurdle band upward rather than downward. The Federal Reserve's projection medians stand at 3.8% for end-2026, 3.6% for end-2027 and 3.4% for end-2028, against a current target range of 3.50–3.75%. Confirmed HOUSE-ANCHOR Dot dispersion around those medians is wide — candidate ranges of 3.50–3.90%, 3.35–3.85% and 3.25–3.75% respectively Range — so any WACC or currency scenario built on a single path understates regime variance and must be rebuilt band-first. Domestically, the policy rate stands at 2.75% Confirmed within a 2.50–3.00% band Range, against consumer prices of 3.2% year-on-year in June 2026 Confirmed HOUSE-ANCHOR inside a 2.7–3.3% band Range. The implied real rate spans −0.8% to +0.3% Range — no longer unambiguously negative. A raised policy floor raises the hurdle, which tightens rather than relieves the lien.

That distinction is the structural pivot, and the audit record forces it. Where the upstream anchor treated mature-node capacity as the object of a quasi-sovereign security mandate, the disclosed attribution points to advanced capacity, conversion and infrastructure. The claim must therefore be redefined from node-class to footprint: what the state-adjacent obligation fixes is where and at what scale capital is deployed, not which node it feeds. Roughly nine-tenths of group capex sat in one segment whose own operating result could fund it several times over. A free-allocation model reads that as a profitable segment reinvesting. The lien model asks whether the location and scale of that reinvestment would survive a pure return test, or whether it is fixed by commitments that predate the test.

Time-stamping is mandatory. The trailing-node pricing regime moved inside the current window: wafer reallocation into advanced products withdrew trailing supply and repriced legacy output upward. A return computation drawn from a 2023-vintage income statement describes a regime that has since inverted. Every return statement here is a through-cycle statement, dated against a post-flood regime rather than the present windfall, and carries a Scenario label. The report does not claim legacy assets are losing money today. It claims today's commitment decisions are made under a constraint today's pricing conceals.

Forward content rests entirely on that commitment-versus-P&L distinction. Foreign trailing-node additions are estimated at 20–40% of global legacy capacity addition Scenario SPECULATIVE, landing with a ramp lag of 4–8 quarters Scenario, while the domestic commitment is contracted now, at shortage-window equipment and construction prices, into cleanroom capital with no alternative use. Backtest window 2015–2026 establishes the pincer: current windfall plus incoming flood traps surplus in legacy maintenance. Scenario Capital locks at peak replacement cost, the offsetting credit window is finite, and the ASP regime justifying the lock is the one most likely to have expired by the time the asset depreciates. The option value of not committing is destroyed in a window where the P&L makes destruction look like prudence.

The advanced-memory return premium is directionally credible; "exponential" is rhetoric and is stripped. On a full-cost basis — conversion capital, through-silicon-via and stack yield burn, multiple wafer consumption per delivered bit, qualification cost carried before revenue, concentration discount — the surviving spread against blended legacy is 5–15 percentage points Scenario SPECULATIVE. Bounded, not unbounded. A margin premium is not a return premium, and the gap between them is where the subsidy capacity of the surplus is determined: a 5-point spread finances a materially smaller lien than a 15-point spread, and the report's own strength scales with where in that band the truth sits.

Single-engine risk is the honest vulnerability and is stated rather than buried. The surplus servicing the lien concentrates in a product with top-customer share of 40–70% Scenario SPECULATIVE, supply constrained by qualification cadence and packaging throughput rather than domestic wafer capacity. If the engine normalises — capacity catch-up, qualification loss, ASP reset — the constraint story does not become false; it becomes acute. The lien does not amortise when the surplus falls: the commitment schedule is contracted, the cash generating it is not. This is the regime in which the distribution ceiling of §V converts from slack to binding, and it is priced nowhere in a memory-beta framework.

III. Treasury Mechanics and the FX Exposure Ledger

The claim that capital committees aggressively forward-hedge dollar receipts to ring-fence them is the most mechanically consequential assertion upstream, and the filing record settles it without inference. The observables are specific and public: currency-risk notes in periodic business reports, resident foreign-currency deposit series, bank net forward position statistics, and cross-currency swap basis behaviour around settlement seasons. This report goes to the first directly.

Two corrections follow, both cutting against the upstream framing. First, a sign correction. A ten per cent appreciation of the dollar against the won is disclosed as increasing pre-tax profit by 2,953,682 million won. Confirmed Won weakness is accretive to reported earnings at the revenue channel, not dilutive. Any chain asserting that hedging necessity inflates a won cost basis and crushes operating margins is directionally wrong and is withdrawn. It is re-routed: the legitimate won-cost channel runs through imported energy, industrial tariff normalisation and domestic construction inflation, and is developed in the Dark Matter Map. Backtest window 2010–2026 supports that routing. Scenario

Second, the netting table the anchor never built — and the audited magnitudes invert the rhetorical use it was destined for. Dollar-denominated equipment, materials and construction content constitute a standing natural hedge estimated at 20–50% of revenue Scenario. That offset is real, and presenting gross receipts as the hedgeable quantity remains a fabrication-adjacent error. But the offset does not neutralise: net exposure after natural hedge stands at 50–80% of gross dollar receipts Scenario SPECULATIVE. The methodological discipline survives; the deflationary conclusion does not. Most of the flow clears netting and remains a live conversion decision — which makes the retention channel materially larger than a "fraction of gross" framing implies, and correspondingly raises what §IV must then test rather than assume.

The adjudication inverts the anchor's sign at the sovereign boundary. Forward-selling pushes banks to hedge, supplying spot dollars and supporting the won. Retention — holding receipts in foreign-currency deposits and deferring repatriation — withholds spot supply and is won-negative. Opposite mechanisms, opposite consequences; the filing record at one duopoly issuer is consistent with the second. The transmission chain is written with the retention sign, the aggregate carries Scenario pending the deposit and net-forward series, and one issuer's notes cannot establish sector behaviour. What they establish is that the forward-based version is unsupported where it was assumed strongest.

One implication survives either instrument. Retention is not a hedge; it is a decision to hold the lien's servicing balance in the currency it was earned in — rational against a commitment schedule denominated partly in imported equipment. It converts a corporate financing decision into a flow decision inside the domestic currency market, which is where the corporate and sovereign ledgers stop being separable.

IV. The Sovereign Echo: Ranking the Flow Map Before Claiming It

The flow map must be ranked before any currency claim is made. The won's structural drivers include pension overseas allocation, retail overseas equity demand, bond-index-inclusion inflows, the goods trade balance, and corporate exporter treasury behaviour. Semiconductor treasury flow is one line, not the table. The audited materiality ratio — semis net dollar flow against daily interbank turnover — sits at 0.05–0.20 Scenario SPECULATIVE. That band is verdict-splitting, and §III's netting result forces the split to be taken seriously rather than dismissed: with 50–80% of gross receipts surviving the natural hedge, the flow entering this ratio is the larger of the two candidate quantities. At the lower bound the corporate limb is a marginal amplifier and nothing more. At the upper bound, one-fifth of daily turnover concentrated into settlement clusters is capable of episodic price-setting without being a structural driver at any other time. The house verb therefore remains amplifies at the margin, with an explicit upper-band episodic case — never dictates, at any point in the band.

Lag structure is the second discipline consensus skips. Conversion and retention decisions roll continuously with settlement cycles; central-bank responses are episodic. A flow accumulating over weeks cannot be matched to a response measured in meeting dates without specifying where the intermediate signal appears. Three series are specified: cross-currency swap basis around quarter-end settlement clusters, resident foreign-currency deposit balances at monthly frequency, and the timing of official smoothing commentary relative to those clusters. If retention is doing sovereign work, the footprint appears in the basis first, deposits second, official communication third — with a lag of weeks, not days.

The confirming observable must be named or the claim is unfalsifiable. The named test: an identified retention episode — a step change in resident foreign-currency deposit balances concentrated in the exporter cohort — coinciding, within the specified lag, with documented smoothing activity or explicit deferral in official rate-path communication. Absent that coincidence, the sovereign limb is boxed as hypothesis, not published as mechanism. What cannot be used: there is no published explicit USD/KRW level threshold for intervention. Smoothing is undisclosed and volatility-based, not level-based. Confirmed HOUSE-ANCHOR Any report citing a specific intervention level is citing a number that does not exist.

"Terminal volatility" is not a measurable object and is banned until defined. The accepted definition is a regime object: realised volatility sustained above the 8–15% annualised band coincident with historical smoothing episodes Scenario, cross-checked against implied volatility across a stated observation window. The breach condition invalidating the currency chain is symmetric: realised volatility sustained inside that band through a settlement season in which retention balances rose materially, demonstrating the flow moves without moving the regime. That is checkable from public series.

The reflexivity bound is the argument this thesis must make against itself. A flow large enough to alter sovereign liquidity is large enough to move forward points and basis against the party executing it. As the cross-currency basis deteriorates through the +50–150bp throttle band Scenario, the marginal cost of the next unit of hedging rises and the optimal ratio falls. The mechanism self-attenuates: a reflexive equilibrium, not a ratchet, and any version requiring unbounded corporate flow is internally inconsistent. This is precisely why the retention channel is the more durable mechanism — non-conversion carries no forward-point cost, only opportunity cost and translation exposure, so it throttles far more slowly than a derivative overlay. The reflexivity bound constrains the instrument the filings rule out more tightly than the one they support.

The macro frame: with medians at 3.8%, 3.6% and 3.4% Confirmed HOUSE-ANCHOR inside candidate ranges of 3.50–3.90%, 3.35–3.85% and 3.25–3.75% Range, and a domestic policy rate of 2.75% Confirmed against 3.2% inflation Confirmed HOUSE-ANCHOR, the differential does not resolve on a schedule that relieves the currency. If the materiality ratio settles at the bottom of its band, the corporate treasury is demoted to marginal amplifier and this chapter compresses to a scenario box — which the house will do without complaint.

V. The Governance Vise: Clocks, Claimants and the Residual

Read structurally rather than as judgment on any filer, that arithmetic is the governance argument in one line. Distribution capacity is what remains after investment absorbs operating generation. Where the investment schedule is fixed by commitment rather than chosen by return, the residual is not a policy variable; it is a remainder. Every stewardship conversation about payout expansion is mechanically a conversation about that remainder's numerator — and the numerator is set by a claimant who is not at the table.

The inequality must be written explicitly, and written honestly against the audited magnitudes. The programme-consistent payout requirement measured against distributable free cash flow post-capex resolves to a gap band of 0.5–2.0 trillion won Scenario. Against the disclosed quarter's residual of 8.70 trillion won, that gap is covered more than fourfold. The ceiling does not bind at the observation point, and this report will not assert that it does. The claim that survives is conditional and through-cycle, consistent with the dating discipline of §II: the residual is a remainder whose numerator is the surplus, and the gap band becomes binding only when the surplus compresses toward it — the engine-normalisation regime of §VIII, and the same regime in which §II's single-engine risk turns acute. The lien's danger is not that it consumes today's residual. It is that the commitment schedule is contracted while the residual is not, so the coverage ratio observed this quarter is the most favourable reading the framework will ever produce. Readers should note the inventory does not state the gap band's period; it is not annualised against the quarterly residual here, and the comparison is presented as a coverage observation rather than a solved inequality. The programme itself is measurable in participation terms: 174 companies had disclosed plans as of end-2025. Confirmed HOUSE-ANCHOR

The stewardship actor was mis-specified upstream, and the correction strengthens the case. Pension actuarial adequacy is governed by contribution rates, replacement rates and demographics — parameters set by enacted reform, not by two issuers' dividend decisions. Domestic equity weight is modest and declining under published mid-term policy. The "mathematically impossible threshold" framing overstates the dependence: if the actuarial link is immaterial, actuarial arithmetic constrains nothing. What remains, and is analytically sturdier, is the fund's role as a governance actor exercising stewardship and proxy authority. The dilemma is not that the fund needs the dividend. It is that its public mandate to press for payout expansion collides with a national industrial commitment the same state apparatus originated.

The villain must be recast, because the conventional casting is incentive-inverted. The assumption that controlling shareholders suppress distributions to fund the mandate runs against two documented incentives in the opposite direction. First, abolition of the largest-shareholder valuation premium for inheritance tax took effect 2025-01-01 while the in-kind payment option remained rejected on mid-2026 reconfirmation Confirmed HOUSE-ANCHOR — improving headline valuation treatment while leaving a hard cash-liquidity requirement that structurally favours distribution. Second, where an intermediate holding company sits in the chain, its debt service depends on upstreamed dividends, which again favours distribution. Neither is consistent with controller-driven suppression. The suppressing agent is more plausibly the commitment schedule itself — state-adjacent in origin, corporate in execution. This is a governance-critical judgment and is rendered at the structural-category level, without naming groups, families or individuals.

Sequencing was posed upstream as an open question — which clock binds first — and the audit closes it: the statutory deadlines force distribution-adjacent action and directly pre-empt the paralysis equilibrium. The Third Amendment passed 2026-02-25, was promulgated and took effect 2026-03-06, and establishes rolling one-year cancellation clocks from the acquisition date of newly acquired treasury shares — first cohort maturing from March 2027 — plus a fixed 2027-09-06 deadline for pre-existing holdings, comprising a six-month grace period and a one-year cancellation window. Confirmed HOUSE-ANCHOR Layered on are expanded director fiduciary duty in force since 2025-07-22 and separate election of audit committee members under the three per cent rule with the one-third independent-director requirement, effective late July 2026. Confirmed HOUSE-ANCHOR

Pre-emption does not, however, dissolve the lien — and the reason is the mechanical difference between the two forms of shareholder return. Cancellation of treasury shares already acquired is non-cash at execution: it retires share count without drawing on the residual. Dividend expansion is cash and draws on it directly. A statute that compels the first while leaving the second discretionary compels exactly the action a footprint-constrained allocator can afford. The binding question therefore migrates: not whether the board acts, which the clock settles, but which action the clock forces and whether it touches the residual at all. Paralysis was the wrong object. Substitution is the right one — statutory compliance delivered through the channel that leaves the commitment schedule untouched.

Finally, the valve. The internal cross-subsidy is load-bearing only while the state declines to fund the mandate directly. Credits, expanded policy finance and direct subsidy each convert an internal drain into a fiscal line — same mandate, different ledger, no constraint on the residual. That such valves open is not speculative: the comparable regime abroad raised its facility investment credit rate to 35% while leaving the construction-start deadline of 2026-12-31 unchanged. Confirmed HOUSE-ANCHOR The domestic equivalent is politically metered. The dissolving threshold is a direct-subsidy scale of 5–15 trillion won annually Scenario SPECULATIVE — the single most important legislative variable in the sector, not because it lifts earnings but because it removes the claimant.

VI. Vulnerability Profile: Where This Report Is Weakest

Consolidated rather than distributed, because an objection buried beside its rebuttal is an objection concealed. Five vulnerabilities carry real kill-radius.

One contamination check. The "mature-node" discourse is largely imported from a foreign policy debate about mature logic capacity, where the state's mandate object is genuinely legacy sustenance. The domestic instrument set — credits, cluster megaprojects, policy finance — and the disclosed capex attribution both point to advanced capacity and its ecosystem. The foreign framing is not re-imported, and that correction is precisely why the lien is defined by footprint rather than node.

VII. Thresholds and the Falsification Ladder

A thesis with a single binary falsifier is unfalsifiable in practice: the binary event is rare enough that the claim survives by default. The upstream single trigger is replaced with a graded ladder — four soft rungs, each killing a chain, and one hard rung terminating the architecture. Readers should treat the soft rungs as live monitoring objects.

Rung 3 is asymmetric to the others. Rungs 1, 2 and 4 falsify by revealing the mechanism was mis-specified. Rung 3 falsifies by removing a correctly specified mechanism — the constraint was real and was then legislated away. Those are different epistemic events and should be scored differently by any reader tracking whether this framework earned its keep.

VIII. Scenario × Asset × Impact Matrix

Directional phrasing only. Impact cells are conditional-state mappings, not sourceable data, and carry the Scenario label. No cell constitutes a recommendation.

IX. Risk Parameter Translation & Monitoring Blotter

Reader-side parameters, not a recommendation sheet. No execution windows are framed as house advice. Thresholds carry the Scenario label; unresolved limits are marked for Master override.

X. Hidden Structure / Dark Matter Map

Scope is the domestic semiconductor sector. Foreign policy rates, foreign trailing capacity and foreign customer concentration enter strictly as exogenous downstream variables.

Causal Chain — The Corrected Won-Cost Channel

Elevated foreign policy path → sustained rate differential → won under pressure → dollar energy import bill rises → state utility absorbs the gap under constrained industrial tariffs → utility balance sheet erodes → transmission capex deferred → cluster energisation slips → domestic build schedule extends → the lien's principal grows → residual thins.

This is the correct mechanisation of the upstream won-cost claim. The revenue-hedging route is sign-inverted against the filed sensitivity disclosure in §III and is discarded. The energy-and-tariff route reaches the same destination with the correct sign, through a sovereign macro variable — the imported energy bill entering the current account, governed by the same policy path anchored at 3.8%, 3.6% and 3.4% Confirmed HOUSE-ANCHOR within its dispersion band. Backtest window 2010–2026. Scenario

Infrastructure Module

INF-1 — The Utility Transducer Consensus Blindspot: YES — Underpriced. Incentive map: a state utility politically constrained from cost-reflective industrial tariffs; an industrial ministry mandated to deliver cluster power; conglomerates externalising energy-cost risk; ratepayer politics capping normalisation speed. Every participant is rational and the aggregate is a deferred bill. Gridlock: suppressed tariffs erode the utility balance sheet, deferring transmission capex and putting cluster energisation at risk. The utility's own cost of capital is unearned by design. The normalisation gap sits at 10–30% against cost-reflective level Scenario. Backtest window 2010–2026 establishes grid delay and tariff suppression as structurally hidden costs inflating the fab capex baseline. Sovereign channel: dollar energy imports → current account → won; won weakness → import bill → tariff pressure → fab opex. The loop closes on itself. Why it is unpriced: sell-side models fab capex, not grid dependency or tariff-normalisation passthrough into operating cost.

INF-2 — Cluster Siting Gauntlet Consensus Blindspot: YES — Underpriced. Incentive map: local authorities extract siting rents; transmission corridors face precedent-setting opposition; the national timeline is politically non-negotiable. Reconciling the three lands on the corporate balance sheet as unmodelled schedule slippage. Gridlock: a consent-based siting regime without compulsory fast-track makes transmission lag, not fab construction, the critical path. Energisation lag against fab-ready date: 6–24 months Scenario. Sovereign channel: cluster delay is national-security-posture delay, deepening state entanglement in private capital sequencing. Schedule delay expands the lien's principal directly and thins the residual further. The mandate grows endogenously out of its own failure to deliver. Why it is unpriced: timeline risk is modelled as execution noise rather than as a structural feature of the consent regime.

INF-3 — Construction Capacity Constraint Consensus Blindspot: YES (marginal). Incentive map: domestic engineering and construction capacity split between cluster build-out and competing demand; a strengthened safety-liability regime raising contractor risk pricing. Gridlock: labour and contractor bottlenecks inflate won-denominated capital cost — a second "won cost" channel entirely independent of the exchange rate, widening the lien's principal with no currency move at all. Industry-wide quantification: [N/A - Requires Master Manual Override | industry-wide construction and labour CAPEX inflation, % of baseline] N/A. Sovereign channel: capital cost inflation widens the servicing requirement, tightening the treasury and governance chains simultaneously.

Product Module

PRD-1 — Wafer Fungibility Coupling: the ledger separates what the cleanroom cannot Consensus Blindspot: YES — Underpriced. Incentive map: product-mix committees arbitrate customer-pressured advanced allocation against spot-priced commodity output on shared physical capacity. Financial ring-fencing coexists with complete physical fungibility — the "schism" is an accounting construct atop a single coupled asset. Gridlock: every wafer moved into advanced products withdraws trailing supply. Legacy pricing is endogenous to the growth of the engine that supposedly subsidises it. Any model treating the chains as separable double-counts the surplus — which is also why the 5–15pp full-cost spread cannot be read as a stable subsidy capacity. Sovereign channel: domestic mix decisions set global commodity memory pricing, feeding trade-balance composition and the currency. Why it is unpriced: this endogeneity is the single most under-modelled structure in the upstream anchor, not merely in consensus.

PRD-2 — The Trailing-Node Pincer Consensus Blindspot: sequencing sub-node YES. Incentive map: the current windfall rewards staying in legacy; foreign capacity landing at 20–40% of global addition Scenario over 4–8 quarters Scenario punishes having stayed. Credit-window expiry makes commitment cheapest exactly when it is most dangerous. Gridlock: the option value of exit is destroyed by mandate plus windfall myopia. Backtest window 2015–2026. Both regimes are separately well covered; the sequencing is not. Sovereign channel: foreign legacy supply enters only as an ASP regime input, never as a primary node.

PRD-3 — Single-Engine Qualification Concentration Consensus Blindspot: coupling YES — Underpriced. Incentive map: advanced-memory revenue concentrates in few qualified customer slots at 40–70% top-customer share Scenario. Treasury planning is built on a flow with binary re-qualification risk. Gridlock: the true supply constraint is qualification cadence and advanced packaging throughput, not domestic fab capacity. Confirmed HOUSE-ANCHOR Engine throughput is governed externally, by counterparties and packaging providers. Sovereign channel: if the engine stalls, treasury, currency and governance chains stall simultaneously — and the distribution coverage ratio of §V inverts in the same move. Concentration is known; its coupling to the commitment architecture is not.

Regulatory Module

REG-1 — The Fiscal Bypass Valve Consensus Blindspot: YES — Underpriced. Incentive map: an industrial ministry wanting capacity, a finance ministry resisting outlays, conglomerates preferring subsidy to internal drain, opposition politics oscillating. The valve exists and is metered. Gridlock: the internal cross-subsidy persists only while the valve stays shut. The load-bearing wall is a legislative variable, modelled by consensus neither as risk nor catalyst. Dissolving scale: 5–15 trillion won annually Scenario. Sovereign channel: opening the valve converts a corporate treasury constraint into a fiscal deficit line. Same mandate, different sovereign ledger, different holder of the residual claim.

REG-2 — Commercial Code Clockwork Consensus Blindspot: NO as a clock, YES as a substitution channel. Incentive map: boards under expanded fiduciary duty and cancellation machinery, controllers whose distribution preferences are inverted relative to the conventional story, stewardship actors facing an enforce-versus-preserve dilemma the statute has partly resolved for them. Gridlock: rolling one-year clocks from acquisition date and the fixed 2027-09-06 deadline for pre-existing holdings Confirmed HOUSE-ANCHOR run on statutory time and pre-empt paralysis outright. The residual survives not because action is deferred but because the forced action is non-cash at execution — compliance and commitment are compatible, which is the structure consensus does not model. Sovereign channel: the governance regime is itself a state instrument shaping distribution capacity — and, in this configuration, shaping which form of distribution capacity is compelled.

REG-3 — Payout Tax Lever Consensus Blindspot: YES as an interaction term. Incentive map: separate taxation of dividend income flips the controller payout preference at the margin, perturbing the distribution equilibrium of §V. Legislative status is an unresolved fact slot and is not asserted. Why it is unpriced: covered as a tax story, uncovered as a cross-subsidy variable — the lever changes who wants the residual, and therefore whether the cash-drawing channel gets chosen over the non-cash one.

REG-4 — Labour and Liability Regime Consensus Blindspot: YES. Incentive map: research velocity against labour-standards politics; contractor risk pricing feeding cluster capital cost. Status of the working-hour exemption question is an unresolved fact slot and is not asserted. Sovereign channel: this is where the employment limb of the mandate becomes documentable rather than assumed. No formal instrument mandates legacy-node employment; there is a documentable localisation drive following the 2019 export-control episode and a documentable labour-normalisation record. The documentable subset is used; the rest is treated as informal political constraint and flagged as such. Why it is unpriced: treated as governance-and-sustainability noise rather than capital cost structure.

REG-5 — FX Microstructure Reform and Index Inclusion Consensus Blindspot: NO as flows, YES as counterweight — Underpriced. Incentive map: authorities engineering deeper offshore access to the domestic currency. Non-resident structural flows then enter the exact channel this thesis assigns to corporate treasuries. Inversion: index inclusion drives a non-resident flow surge that can overtake corporate treasury retention in the flow ranking, demoting the currency chapter's protagonist from principal to bystander. It appears here as a counterweight to the report's own argument, deliberately. Related exogenous variable: the annual-licensing regime that replaced validated end-user status on expiry at 2025-12-31 Confirmed HOUSE-ANCHOR introduces recurring binary renewal risk to cross-border capacity operations, entering strictly as a downstream input.

XI. Consensus & Usage

(a) Consensus Under Interrogation. The prevailing view treats these issuers as blended AI beta: advanced-memory cash generation lifts aggregate margins, domestic mature capacity is a benign cyclical option, and the currency is an overlay applied after the fact. A second camp inverts the legacy limb and treats trailing nodes as a profit engine. This report contests neither camp's pricing. It contests the premise both share — that the surplus is freely allocable and the currency exogenous — and repositions the sector as a footprint-constrained allocator whose treasury behaviour is one input into the currency it is measured in.

Bridge for the non-Korea allocator. Holding this sector as an AI expression from outside the market means holding a claim on a currency whose corporate flow channel you have not modelled — with 50–80% of gross dollar receipts surviving the issuer's natural hedge and no derivative overlay on the dollar leg — and a distribution profile governed by a statutory clock and a subsidy vote rather than a payout policy. The exposure is not the one on the fact sheet.

Single regime-invalidating line. The trailing-node pricing regime may have cyclically inverted the incinerator premise, and the return and hurdle bands overlap across their mutual range. Every return claim here is a through-cycle claim dated against a post-flood regime, not the present windfall. Full treatment in §II and §VII.

XII. Observation Markers

Falsifiable, dated, and checkable by any reader against public sources. No single-name language, no price targets, no portfolio actions.

Marker 1 — The instrument and netting test. Observable: currency-risk notes, derivative disclosures and import/equipment cost composition in the FY2026 half-year business reports of both duopoly issuers, filed to the regulator's electronic disclosure system. Deadline: 2026-08-31. Confirms: the derivative overlay on the dollar leg remains immaterial relative to disclosed sensitivity, and disclosed dollar cost composition implies net exposure inside the 50–80% band of gross receipts — retention, not forward-selling, is the operative channel, and netting does not neutralise it. Invalidates: either a materially expanded forward book on dollar receivables with hedge accounting adopted, or disclosed cost composition implying net exposure below 50%. The first inverts the currency chain's sign; the second restores the deflationary netting reading and demotes §IV independently of the materiality gate. Either outcome forces §III to be rewritten.

Marker 2 — The retention test. Observable: resident foreign-currency deposit statistics across the third-quarter settlement season, read jointly with bank net forward position data. Deadline: 2026-10-31. Confirms: corporate foreign-currency deposit balances rise through the settlement cluster while net forward positions are broadly unchanged. Threshold magnitude: [N/A - Requires Master Manual Override | resident FX deposit balance change, USD bn] N/A — direction is checkable now; magnitude awaits Master population. Invalidates: balances flat or falling while net forward positions expand — Ladder rung 1, terminating the sovereign limb.

Marker 3 — The substitution test. Observable: board resolutions and disclosure filings on treasury-share cancellation for the first rolling cohort — shares acquired after the amendment took effect on 2026-03-06 and reaching their one-year clock from March 2027 — read against cash distribution decisions disclosed for the same period. Deadline: 2027-03-31, with the fixed statutory backstop for pre-existing holdings at 2027-09-06. Confirms: statutory compliance delivered through non-cash cancellation while cash distribution remains at or below programme norms, leaving the 0.5–2.0 trillion won gap unclosed. The clock is satisfied through the channel that does not touch the residual — the substitution mechanism operating as described in §V. Invalidates: cash distribution expanded beyond programme norms alongside cancellation, absent a subsidy enactment. The residual was never the binding constraint and the governance chapter fails on its own terms. Note that mere occurrence of cancellation no longer confirms anything — the statute compels it — which is why this marker tests the channel, not the event.

XIII. Alpha & Acre Macro Methodology v1.0 — Surgeon's Layered Anatomy

Public-interest statement. This report's analysis of corporate governance and capital allocation structures is conducted in the public interest — to improve information transparency for investors and market participants regarding matters of public corporate disclosure. All factual statements are sourced to identified public disclosures, regulatory filings, statutory text, or other verifiable primary sources cited herein; analytical interpretation is clearly distinguished from sourced fact throughout.

The Four Layers. Layer 1 — Surface Narrative: what the market is saying and what it concedes, mapped to §I, the split-consensus ledger and the isolation of the shared premise, attacked in place of the caricature. Layer 2 — Organ / Macro Vectors: the load-bearing economic organs and the macro vectors acting on them, mapped to §II and §III — node-class disaggregation, the band-against-band return test, the dollar exposure ledger and instrument adjudication. Layer 3 — Nerves / Transmission: how signal travels from corporate decision to sovereign variable and back, mapped to §IV and §V — flow-map ranking, lag structure, reflexivity bound, statutory clocks and the distribution residual. Layer 4 — Scars / Regime Shifts: structures left by prior regimes that constrain the next one, mapped to §VII, §VIII and §X — the falsification ladder, the regime matrix, and the Dark Matter Map's infrastructure, product and regulatory modules.

Data Source Hierarchy. Tier 1 (Sovereign / Regulator): statutory text and promulgation records; regulatory electronic disclosure filings; central bank statistical series and official communications; national statistical office releases; foreign central bank projection materials. Tier 1 sources may be stated as fact with the Confirmed label. Tier 2 (Listed Issuer IR / Consensus): company periodic reports and notes, investor guidance, press releases; sell-side aggregate positioning where used to characterise consensus rather than establish fact. Tier 2 disclosure lines carry Confirmed; consensus characterisations carry Market Estimate. Tier 3 (Macro Proxies): derived ratios, allocation inferences, proxy stacks, and any quantity not disclosed at the granularity the argument requires. Tier 3 quantities may never carry Confirmed and are rendered as Scenario, as Range where a band is externally sourced, or as unresolved open quantities.

Open Quantity Register. This report carries three distinct open quantities, appearing in four pipe-delimited override slots. The marker string [N/A - Requires Master Manual Override] appearing without a pipe-delimited unit descriptor — as in this sentence — is definitional prose and is never a data slot.

Anchor credibility inheritance. Predicates flagged suspect or unverifiable upstream are never promoted to fact; they appear under conditional phrasing, a Scenario label, or an open-quantity slot. "Terminal volatility" appears only as a defined volatility-band regime in §IV.

Band-first discipline. Where dot dispersion or estimation error is material, the report states both bounds of a band and the verdict at each bound. A point estimate is used only where a primary source discloses a point.

Directional call suppression. No single-name buy, sell or hold recommendation, no price target, no trade list. Where an adversarial question demanded a trade answer, the answer is the mechanism and the threshold, not the call.

Korea jurisdiction override on named entities. Named listed issuers appear only in neutral factual contexts: disclosed financial lines, filing dates, regulatory receipt numbers. All governance-critical and evaluative analysis uses anonymised structural categories, which is why the §V classification table is built on archetypes. Owner families, executives and other natural persons are not named in any context; institutional attribution is used throughout. Statutory bodies and regulators are named where relevant.

Operating-entity precision. Metrics are attributed to the operating unit that disclosed them. Segment figures are not attributed to a parent or group, and affiliate results are not consolidated by implication.

Derived figures and unit integrity. Arithmetic differences computed from Confirmed disclosure lines are marked derived and stated in the filing's own unit. The residual in §V is stated as 8,695,220 million won against disclosed lines of 26,330,119 and (17,634,899) million won; where a downstream input conflicted with this unit convention, the primary filing governed and the conflict was returned to the Master rather than absorbed silently.

Rebuttal vitality. Where audited quantities inverted the directional premise of an adversarial rebuttal, the rebuttal was rewritten to defend the corrected quantitative reality rather than deleted or softened. Five rebuttals were re-tuned on this basis; each retains its original anchor.