Desk Summary & Risk Insulation

• The halo is event-level, not firm-level. Section 48D at 35% HOUSE-ANCHOR — the operative rate at the top of a statutory 30–35% band Range — is node-agnostic with elective-pay monetization. Any screen built on “subsidized firm vs unsubsidized firm” is mis-specified at the source.
• The subsidy skew is a gradient, not a category. Dollar- and project-weighted composition runs 60–80% advanced against 20–40% mature and foundational Market Estimate. A categorical “advanced-only” premise mis-maps both the target set and the intensity of the tax.
• The unpriced object is the externality term, not the transfer. Consensus is split on leading-edge versus legacy; both camps share the premise of spillover-neutrality. That shared premise is the target.
• Incumbency Tax = factor-cost inflation + relative funding penalty + subsidy-financed asset depreciation. None of the three appears on an incumbent income statement as a policy line item.
• The first discriminator has partially fired — against this framework. National producer prices for new industrial building construction rose +45.4% cumulatively 2020-01 to 2026-06 Confirmed, inside the hub band rather than beneath it. On construction cost there is no geographic gradient. The equipment and specialist-labour legs remain untested, and the channel now stands or falls on them.
• The rate floor descends, but not far enough or fast enough. The SEP glide holds the terminal band above the pre-2022 regime across the entire window in which mature-node maturities and maintenance capex both fall due. Slope, not flatness, is what the framework requires — and the slope is audited.
• First-crossers are an intersection, not a sector. Commodity-exposed ∧ hub-co-located ∧ harvest-mode ∧ outside the subsidy perimeter ∧ unmoated. Remove any one condition and the impairment sequence reorders.
• The unsubsidised cohort is populated but regime-split. Wholly unsubsidised funding covers 30–60% of domestic legacy capacity Market Estimate, distributed across three to four distinct funding regimes rather than a single binary.
• Qualification moats outrank node class. AEC-Q automotive and ITAR / Trusted Foundry accreditation preserve repricing power. The operative split is intra-legacy, not advanced-versus-legacy.
• The 2026-12-31 start-of-construction deadline HOUSE-ANCHOR is time-shaped distortion — and is itself contested. The credit rate was raised; the deadline was not moved. Extension advocacy is active and unresolved, which makes the cliff a dated but not a certain event.
• Credit stress travels by private channel, not commercial paper. CP clears at 1–5% and is IG-gated; private-credit, sale-leaseback, equipment-ABS and convertible structures clear at 6–12% Confirmed. The channel is priced; the sector-specific instances are not yet observed.
• The pricing cap is regional, not global. Offshore mature-node scaling presses global legacy ASPs by −10% to −30% Market Estimate, but an effective 50–75% Section 301 band Confirmed partially re-valves domestic sales. A single global cap over-prices the damage.
• Distress does not require a backstop to resolve. Quiet consolidation by IG incumbents absorbs capacity without a public event. The boring path is the base path.

Risk Box 01 — Attribution (regime-invalidating). Mature-node ROIC compression must decompose into foreign supply glut, inventory cycle, and domestic subsidy externality. If the third term is not separable from the first two, this report has no incremental claim and reduces to a restatement of consensus. Separability is the single load-bearing test, and it has now returned its first negative reading on the construction leg.

Risk Box 02 — Census. The audited unsubsidised share of 30–60% establishes that the cohort exists; it does not identify its members. The census table — owner, rating, leverage, grant and 48D status by facility — is not built. Until it is, the framework’s target set is bounded in aggregate and unresolved in the particular.

Risk Box 03 — Crowding. Bifurcation is a widely held institutional posture. If dispersion and positioning data show the halo-versus-legacy split already owned, the structural read has been converted into exit liquidity and the framework’s tradeable content is spent.

Risk Box 04 — Half-life. This report publishes inside the terminal Section 48D window. The pull-forward mechanics it describes are observable now and extinguish at the cliff — unless the cliff moves. Thesis half-life is bounded by a statutory date that is under active extension advocacy, which makes the bound conditional rather than fixed.

The Premise Beneath the Split

The consensus this report attacks is not the caricature of a universal rising tide. Sell-side coverage already discriminates: advanced-node and packaging exposure carries a structural premium, mature-node exposure carries cyclical caution, and the caution is attributed almost entirely to two externally sourced narratives — foreign capacity additions and an unresolved inventory cycle. That split is real, well-documented, and saturated. Writing against it would be writing against a strawman.

What both camps hold in common is the assumption that survives unexamined precisely because it is shared: spillover-neutrality. Sovereign capital is priced as positive for recipients and exogenous to non-recipients. The transfer is modelled; the externality is not. There is no line in a mature-node operator’s cost bridge labelled “subsidy-induced factor inflation,” no rating agency notch for “competing capacity financed at a 35% credit,” and no consensus estimate carrying a domestic-policy term in the mature-node ROIC path.

The Incumbency Tax is the name for that missing term. It is not a claim that subsidy is net-negative for the ecosystem, and not a claim that the CHIPS architecture failed. It is a claim about incidence: the cost of the transfer is borne partly by parties who receive none of it, through channels that are physical and local rather than fiscal and national, and the incidence is unpriced because the accounting system has no field for it.

The Rate Floor and the Anatomy of the Subsidy Stack

2.1 · The floor that descends too slowly. The monetary architecture underneath this framework is not a tightening cycle, and — on audit — it is not a flat plateau either. It is a shallow glide at an elevated level. The Fed funds target sits at 3.50–3.75% HOUSE-ANCHOR Range, with the effective rate observed near 3.6% inside that band, and the projection path — 3.8% end-2026, 3.6% end-2027, 3.4% end-2028 HOUSE-ANCHOR Confirmed, each a published SEP median — describes roughly twenty basis points of annual descent terminating above the pre-2022 neutral regime. The nominal policy floor a capital-intensive manufacturer funds against is therefore 3.50–3.75% today and materially unchanged in character three years out. The real-yield translation of that floor is held at N/A — 10y real yield, FF less expected inflation, % — an indicative band exists in house discussion but is carried Speculative and is not inserted as a value. Until it is resolved, the funding-penalty limb is argued on nominal terms only, and that limitation is disclosed rather than papered over.

What the glide does is remove the historical escape hatch without removing the appearance of relief. In prior cycles, a mature-node operator facing cost inflation could wait: the discount rate would fall far enough that refinancing cleared and deferred maintenance was caught up in the next easing window. A descent of this shape offers the form of that relief without the substance — the rate falls, the operator waits, and the arrival never comes before the maturity does. Waiting converts from a strategy into an accumulation of unfunded obligation, and it does so more insidiously under a visible glide than under a visible plateau, because the glide supplies a reason to keep waiting.

2.2 · Subsidy stack anatomy. The stack has three tiers with different eligibility geometry, and conflating them is the most common analytical error in this sector. Direct federal awards are discretionary, negotiated, milestone-paced, and — on audit — decisively not confined to leading-edge nodes. The dollar- and project-weighted composition runs 60–80% toward sub-5nm, AI architectures and advanced packaging, against 20–40% reaching mature and foundational capacity Counter-Trend Corrected Market Estimate. That is a skew, not a wall, and the distinction governs everything downstream: a categorical premise assigns the Incumbency Tax to a node class, whereas a gradient premise assigns it to a position on a distribution. Section 48D is the second tier and behaves differently again: a statutory investment credit operating at 35% HOUSE-ANCHOR Confirmed, the top of a 30–35% statutory band, with the start-of-construction deadline of 2026-12-31 unchanged, node-agnostic on its face, and monetizable by election for filers without sufficient tax capacity. Institutional and private co-investment alongside federal awards forms the third tier; where such structures operate at megasite scale, WACC compression at the apex compounds beyond the statutory credit alone.

The consequence for thesis construction is decisive, and it narrows the claim. Because Section 48D is node-agnostic with elective pay, and because one fifth to two fifths of direct award value reaches mature and foundational capacity, the halo is not a firm-level property and “legacy” is not the victim class. The share of mature-node capex genuinely outside qualifying scope is held at N/A — share of US mature-node capex outside 48D-qualifying scope, %. What can be said from audited data is adjacent but not identical: wholly unsubsidised funding — unsubsidised debt plus internal cash flow, with no grant and no credit — covers 30–60% of domestic legacy capacity Counter-Trend Corrected Market Estimate. The halo boundary therefore survives at capex-event level and at funding-regime level, not at node level. A firm with no qualifying construction earns nothing regardless of node class or distress; a firm building 300mm mature-node capacity earns the credit regardless of how legacy its output is. The tax falls on standing assets and harvest-mode operations. That is a narrower thesis than the anchor implied, and a more defensible one.

2.3 · The cliff is a shape, not a level. The 2026-12-31 start-of-construction deadline HOUSE-ANCHOR Confirmed compresses every qualifying build into a single pre-cliff window. This is not a level distortion in the cost of capital; it is a temporal distortion in the demand for factors. Qualifying starts concentrate at a level held N/A — 48D pull-forward concentration, qualifying starts pre-cliff as % of announced pipeline — exhausting trade-labor and construction capacity precisely when a non-qualifying operator must also execute maintenance work. The mirror image follows: a post-cliff capex void, also held N/A — post-cliff 2027 sector capex decline vs 2026, % — that strands late movers and collapses the factor-price signal that justified deferral. One qualification is binding and is carried throughout: industry advocacy for statutory extension of the credit window is active and unresolved Scenario. The cliff is dated; it is not guaranteed.

Sovereign WACC Bifurcation

The chain runs: policy floor → refinancing cost for capex that cannot clear the subsidy stack → divergence in capex hurdle rates between halo-adjacent and non-halo activity → involuntary deleveraging or asset sale → migration of capacity ownership toward subsidy-eligible balance sheets. The structural break is not a price. It is a quantity. Lenders do not raise the coupon on a brownfield credit line at a fab whose interest coverage has fallen through a floor held at N/A — ICR floor at which brownfield credit lines are withdrawn, × — they withdraw the facility. Rationing, not repricing, is the observable regime marker.

Cohort Census Panel — the addressable asset base. The framework requires operators simultaneously levered, outside the subsidy perimeter, and independent. The audit resolves the aggregate and leaves the roster open, and this reverses the panel’s original risk direction. Wholly unsubsidised funding — unsubsidised debt plus internal cash flow, no grant, no credit — covers 30–60% of domestic legacy capacity Counter-Trend Corrected Market Estimate, distributed across three to four distinguishable funding regimes rather than the binary the draft assumed. The operator count remains held at N/A Speculative — count of US mature-node operators simultaneously levered, non-subsidised and independent — because the facility-level census (owner, rating, leverage, CHIPS grant and 48D status) is not built. Without this census, the actual target and intensity of the Incumbency Tax cannot be systematically verified, leaving the highest-kill-power predicate of the thesis structurally weak and operating without data. The cohort is therefore not thin; it is large, heterogeneous, and unidentified. Three constituencies populate the band: harvest-mode facilities executing no qualifying capex; sub-scale operators for whom application overhead exceeds the expected award; and private or sponsor-owned specialty fabs whose capital structures are not publicly observable — the reason the census cannot be completed from listed filings, and the reason availability bias toward large-cap names systematically understates the tail. The relocation clause is now a secondary rather than a primary contingency, since the aggregate is populated. It nonetheless stands: the same incidence logic applies inside a diversified investment-grade issuer, where a legacy division competes for internal capital against a segment whose builds earn a 35% credit. There the victim is a segment rather than a filer, the observable is internal capital rationing rather than credit distress, and the maintenance-capex deferral rate of the stressed cohort — held at N/A Speculative — becomes the primary tell.

Credit Evidence Panel — is the wedge priced, latent, or imagined? The proposition that public credit markets already price a halo premium is a testable claim and remains unverified. The audit returned no credit-spread observation of any kind: the required evidence — a rating-controlled spread differential between subsidy-linked and non-subsidy-linked semiconductor issuers — sits at N/A — rating-controlled spread wedge, bp. This is a stronger form of the original position: the question was put to the inventory and came back empty, rather than merely being left unasked. If the wedge is observable ex-ante, the mechanism is confirmed and largely priced, and analytical value collapses to timing. If the wedge is absent but the underlying cash-flow divergence is measurable, the mechanism is latent and the framework has genuine content. If neither is observable, the credit channel is imagined and C1 should be struck. No language in this report asserts the wedge exists.

Chain seam — capital scarcity versus factor scarcity. Global capital is elastic; a solvent US manufacturer with an acceptable project can raise money in almost any rate regime, and any claim that federal grants “used up” the capital available to non-recipients is analytically weak. That is not the crowding-out asserted here. The binding scarcity is physical and non-arbitrageable: a cleanroom-qualified electrician in a fab hub, a rigging crew with tool-install experience, a firm interconnection position in a constrained service territory. These cannot be imported at the margin on a project timeline, and no cost of capital purchases queue position. Capital crowding-out is discarded; factor crowding-out is retained as the operative channel and carries the entire C2 burden.

Factor-Market Crowding — and Its First Falsification Test

Megaproject concurrency inflates the wage and construction cost base in metros where fab clusters sit. Transmission runs: shared trade-labor pool → wage and construction escalation → unit-cost inflation in maintenance capex at co-located legacy facilities → deferral of maintenance → yield and reliability decay → margin erosion independent of pricing. The critical distinction is between expansion capex and maintenance capex. Deferred expansion is a demand signal and is priced by consensus. Deferred maintenance is a solvency signal and is not — yield capability lost to deferred maintenance is not recoverable on an economic timeline, and qualification lapses cannot be re-earned quickly at any price.

The chain is co-location-contingent, and this is its principal vulnerability. The audited magnitude is on the record for the hub side: in regions where subsidised fab load, hyperscaler datacenter build and grid-reinforcement projects overlap, unit costs for transformers, HV equipment, construction and specialist trades rose cumulatively 20–50% across 2020–2026 Scenario — a band spanning both the construction-premium and specialist-wage limbs, cumulative rather than year-over-year. The national comparator has since been supplied for one leg, and it does not support this chain. US producer prices for new industrial building construction rose +45.4% cumulatively from 2020-01 to 2026-06 (138.100 → 200.869) Confirmed — inside the hub band, not beneath it. On construction cost the geographies do not diverge, which is the condition this chain nominates below as a clean kill, and the construction sub-channel is therefore struck as subsidy-attributable rather than defended. The co-located share of domestic mature-node capacity, and the same-object comparator for equipment and specialist labour, both remain N/A — national comparator, transformer/HV-equipment and specialist-trade cost series, cumulative % 2020–2026. The test is unchanged in logic and narrowed in scope: if non-co-located fabs show equipment and trade cost paths converging with the national series while hub-located peers hold their escalation, the mechanism is confirmed on those legs and its damage set is bounded to the co-located share. If all operators show identical escalation regardless of geography across every leg, the driver is national input inflation and C2 is falsified entirely — the reason hub-versus-non-hub divergence heads the Blotter. A further condition binds: escalation must persist beyond the pre-cliff pull-forward window to be structural rather than a scheduling artefact. The causal direction is unchanged — where megaproject concurrency binds a physically fixed factor pool, the incumbent pays, and no cost of capital buys queue position. What has changed is that the most substitutable input in that pool, general construction, has been tested and shows no premium. C2 narrows from a broad cost-inflation claim to a claim about transformers, HV gear and cleanroom trades specifically, and it is now confirmed on those legs or not confirmed at all.

Pricing-Ceiling Collision and Shadow Liquidity

C3 · Pricing-ceiling collision (exogenous). Offshore sovereign-backed mature-node scaling presses global legacy ASPs downward by −10% to −30% Counter-Trend Corrected Market Estimate, closing the pass-through valve for cohorts without differentiation. This chain is a downstream input only; no primary node sits offshore. The audit forces one structural correction: the cap is not global and uniform. An effective Section 301 band of 50–75% on China-origin semiconductors Confirmed, with certain exclusions extended into late 2026, partially decouples domestic sales — domestic-facing revenue re-valved, export and global-facing revenue still capped. The same operator can therefore display divergent margin behaviour by geography of sale, and a screen assuming a single global cap over-prices the damage to US-domiciled legacy ROIC. Qualification stickiness modulates it further: AEC-Q and ITAR-accredited lines are not commodity-substitutable and retain repricing power regardless of offshore capacity. Consensus carries the China-supply narrative in full; no edge is claimed on the supply story. The edge claim, such as it is, sits on the bifurcation.

C4 · Shadow liquidity and systemic migration (dark-matter chain). The transmission required mechanical correction from the source framing, and the audit now supplies the pricing. Commercial paper is effectively investment-grade-gated and clears at 1–5% Confirmed; a stressed sub-scale fab does not enter that market, it is absent from it. The realistic conduits — private-credit direct lending, equipment-backed ABS, sale-leaseback and convertible or PIK structures — require 6–12% all-in coupons Confirmed. Each shares two properties that matter more than the label: covenant-heavy, and predominantly floating-rate. Under a policy floor descending only twenty basis points a year, a floating-rate liability share held at N/A — floating-rate share of stressed-cohort liabilities, % — converts the rate path directly into a cash-flow constraint with no fixed-rate insulation.

The chain’s status has changed shape: the audit resolved one half of it and left the other empty. The pricing regime is no longer speculative — the 500–1,100 basis point gap between IG commercial paper and private structured credit is documented at channel level, establishing that a fab pushed out of the IG channel faces a step-change rather than a gradient in funding cost. What remains wholly unevidenced is sector-specific instantiation, held at N/A — documented mature-node fab private-credit, sale-leaseback and equipment-ABS transactions, count 2024–2026. The channel exists and is priced; whether any US mature-node fab has actually used it is unknown. If no such transactions are documented, C4 is a correctly-priced channel with zero observed traffic, and should be struck as a live mechanism rather than softened into a latent one. A mechanism with confirmed pricing and no confirmed instances is a hypothesis with a price tag, not a risk.

Scenario × Asset × Impact Matrix

All impact cells carry the Scenario label pending further resolution. S1 Audited Glide — SEP path realised at 3.8% / 3.6% / 3.4%, cliff binds, no further trade action. S2 Shielded Domestic — the 50–75% Section 301 band holds and broadens. S3 Steepener + Demand Break — easing materially faster than the SEP glide, with automotive/industrial order deterioration. S4 Soft-Landing Corridor — easing beyond the glide arrives without end-market deterioration.

Impairment Ordering and the Falsification Corridor

The impairment sequence is governed by an intersection of five conditions, not by node class. A facility crosses first when it is simultaneously commodity-exposed in its output mix, co-located inside a megasite factor radius, in harvest mode with no qualifying capex, funded inside the 30–60% wholly unsubsidised band Market Estimate, and unmoated by qualification barriers. The fourth condition is now a regime rather than a rating: the audit finds three to four distinguishable funding structures across legacy capacity, so “outside investment grade” understates the discrimination available. Remove the fifth condition and the ordering inverts materially — AEC-Q automotive and ITAR / Trusted Foundry status create switching costs measured in years of requalification, preserving repricing power straight through the C3 cap. Moated cohorts holding a revenue share still held at N/A — qualification-moat revenue share, AEC-Q and ITAR/Trusted cohorts, % of segment revenue — do not sit at the front of the queue at all; they sit behind unmoated advanced-adjacent capacity in several plausible states. The operative split is intra-legacy along a moat axis, not the advanced-versus-legacy divide consensus already runs exhaustively.

The timing question is whether three curves intersect, and the audited rate path changes how that claim must be argued rather than whether it survives. Grant disbursement pace, the pre-cliff construction concurrency peak, and the 2026–2028 maturity wall for US semiconductor issuers ex-mega-caps all remain held at N/A pending disclosure. The original formulation leaned on flatness: no descent, therefore no relief. That formulation is wrong and is withdrawn. The audited path does descend — from 3.8% through 3.6% to 3.4%, roughly twenty basis points a year. The forcing logic must therefore rest on level and slope insufficiency rather than on immobility — and it does so more strongly than the flat version did. A 2027 maturity refinances into a mid-threes handle against a coupon struck in a sub-one regime; a 2028 maturity gains perhaps forty basis points of relief against a 500–1,100 basis point channel penalty if the issuer has by then been pushed out of the IG market. The descent is real and it is irrelevant at the scale of the problem — a harder claim to make and a more defensible one to hold, because it survives the audit that killed the plateau.

The 2026-12-31 cliff cuts both ways on the framework’s own shelf life, and the audit removes the certainty that made that bound clean. Pull-forward exhausts factor markets now, which is what makes C2 observable in current filings; the post-cliff void then opens a 2027 air-pocket (magnitude held N/A) in which factor costs deflate and funding stress migrates from cost inflation to demand contraction. Active industry advocacy for extending the credit window Scenario makes the “statute not cycle” assumption unsafe: the cost-push mechanism is bounded by whichever date the deadline finally lands on, while the incumbency-depreciation mechanism is bounded by nothing at all. Extension would not weaken this framework; it would lengthen the cost-push limb and leave the depreciation limb untouched.

The zombie criterion is a structural condition rather than a valuation, and under a descending floor it is duration-bounded rather than open-ended. A facility has crossed when interest coverage sits inside an ICR band (held N/A) for a continuous lockout duration (held N/A) while maintenance capex — not expansion capex — is deferred. A lockout is survivable if the operator can hold out until rates fall enough to reopen the window, so the criterion is really a race between the lockout clock and the descent. At roughly twenty basis points a year against a 500–1,100 basis point channel penalty, the descent loses that race by an order of magnitude, and lockout duration is set by covenant and cash burn rather than by monetary policy. The single confirming datapoint is not a downgrade and not a covenant breach: it is the withdrawal or non-renewal of a brownfield revolving facility disclosed in a periodic filing. Price-based markers lag it by construction, because the lender exits before it reprices.

Risk Parameter Translation & Monitoring Blotter

Reader-side risk parameters, monitoring triggers and invalidation conditions. This is not a recommendation sheet and contains no execution guidance.

The Attribution Problem, Cross-Asset Leakage, and the Terminal Caveat

The attribution problem — lead objection. The strongest position against this framework is that mature-node ROIC compression is fully explained by foreign capacity additions and an unresolved inventory cycle, and that domestic subsidy incidence is second-order noise dressed as structure. The objection is correct unless the third term is separably observable, and the framework concedes the burden of proof entirely. Four discriminators isolate the domestic externality channel. Foreign supply and inventory effects are national and product-specific; they do not vary by the claimant’s zip code, whereas subsidy-induced factor inflation is geographic. The first discriminator is hub-versus-non-hub cost divergence, and it has now partially fired — against this framework. The hub leg stands at 20–50% cumulative escalation Scenario. The national comparator is resolved for the construction leg: US producer prices for new industrial building construction rose +45.4% cumulatively from 2020-01 to 2026-06, 138.100 to 200.869 Confirmed. That figure sits inside the hub band rather than beneath it — and beneath the band’s own midpoint, which would put the hub at a discount to the nation. On construction cost there is no observable geographic gradient. That is what a national-inflation explanation predicts and not what this framework predicts, and the house records it as evidence against its own channel rather than around it. Two qualifications bound the finding without rescuing it: the comparison is object-asymmetric (a mixed hub band against a construction-only comparator, so the equipment/labour legs remain outstanding and could still restore a gradient), and tier-asymmetric (a Tier 1 confirmed national figure read against a scenario-tier hub band). Status: partially fired — no gap on construction, equipment and labour unverified. Second, capex deferrals whose stated cause is input cost rather than demand. Third, the rating-controlled spread wedge, which the audit returned empty. Fourth, cluster utility tariff filings citing concurrent large-load additions. If these four cannot be assembled, the framework has no alpha claim and should be read as a taxonomy rather than a thesis. The house position is unchanged in kind and materially worse in standing: the first discriminator has reported and reported against the framework on the one leg it can currently measure, and survival now rests on the equipment and labour legs plus the three discriminators that have not yet reported.

Cross-asset leakage — where the framework escapes the sector. Four channels carry the incidence outside semiconductor equity, and the audit forces a downgrade of the house’s own claimed advantage on the first of them. Utility ratebase and industrial tariff feedback in fab clusters is the most tractable channel, but the draft’s claim that the house grid and transformer time-series constitutes a differentiating observation base does not survive audit — that data has not been systematically collected, and the series is designated rather than populated. That gap is now a binding constraint rather than a housekeeping item, since the construction-leg comparator returned no gradient and the equipment/labour legs are the only remaining route to a positive reading on the first discriminator. Municipal and state incentive exposure is the second channel; industrial commercial real estate around megasites is the third and carries the sharpest reflexivity — the same pull-forward that inflates construction demand builds the vacancy that follows the cliff. The fourth is WFE order-book skew, where legacy-tool lead-time inflation (held N/A) reads directly on whether apex orders crowd foundational-node maintenance.

Terminal caveat — is this already owned? Bifurcation as a posture is not novel and has been institutionally held for several quarters. If positioning data (held N/A) show the split crowded, the incremental content of this report is not the direction of the split but the axis along which it is drawn — intra-legacy by moat, co-location and capex-event eligibility, rather than inter-tier by node. A crowded position on the wrong axis is more dangerous than no position, because it unwinds on the first datapoint showing a moated mature-node operator repricing while an unmoated advanced-adjacent one does not. Readers already holding the tier-level view should treat this as a re-sorting instruction rather than an entry thesis — and should note that co-location, one component of the axis, has just failed its first measured test on the construction leg and is currently carried on the strength of the equipment and labour legs alone.

The Exogenous Counterforce, and What the House Is Prepared to State

Collision note (exogenous). Sovereign-backed mature-node capacity additions outside the United States enter strictly as a downstream input to C3 and generate no primary node. The capacity-cap mechanism is fully carried by consensus and no edge is claimed on it; the audited magnitude of −10% to −30% pressure on global legacy ASPs Market Estimate is reported for calibration, not differentiation. The cap is not uniform once trade policy is admitted: an effective Section 301 band of 50–75% on China-origin semiconductors is in force Confirmed, with exclusions running into late 2026. The pricing regime therefore splits — domestic-facing revenue partially re-valved, export-facing revenue still capped — and the same operator can show divergent margin behaviour by geography of sale. A framework assuming uniform commodity pricing across a tariff wall is mis-specified in exactly the way a framework assuming a firm-level subsidy halo is mis-specified: both replace a gradient with a boundary.

House View. The Rent Ladder framework established in BP-10 measured front-end ROIC as a function of position on a single rent curve. The Incumbency Tax bifurcates that curve. Under a node-agnostic investment credit with elective monetisation, and with one fifth to two fifths of direct award value reaching mature and foundational capacity, the determinant of rung position is no longer node class but capex-event eligibility interacting with three modifiers: physical co-location, funding regime, and qualification moat depth. Two operators at identical nodes producing identical parts can sit on different ladders entirely.

The structural statement the house is prepared to make is conditional and narrow. If the hub-versus-non-hub cost gradient is observable — and on the construction leg, now measured on both sides, it is not — and if capex-deferral attribution runs to input cost rather than demand, then domestic subsidy incidence is a separable term in mature-node ROIC and the correct sorting axis is moat depth crossed with co-location, not node class. If either condition fails, the framework reduces to a well-organised restatement of the China-supply narrative and should be discarded rather than defended. Three audited corrections bind the statement further, and the third is sharpest against the house: the first discriminator has partially fired toward falsification, so the separability claim now stands or falls on equipment and specialist-labour series that have not been assembled. Additionally, the policy floor descends (so the funding-penalty limb rests on insufficiency of descent, not its absence), and the pricing cap is regional rather than global (so the damage estimate must be discounted for the tariff shield). No single-name conclusion follows from either branch; what follows is a re-specification of the screen. The house product here is the axis, not the position on it — and the axis is currently one measured leg short of being defensible.

Dark Matter Map

Sovereign transmission chain — from a local queue to the dot plot. (1) Subsidised fab load and hyperscaler datacenter load bid concurrently into overlapping utility service territories. (2) Interconnection queues and rate-case procedure impose a lag no cost of capital can shorten. (3) Utilities recover concurrent build-out through ratebase; cost allocation between new large load and incumbent industrial load is decided in contested proceedings, not markets. (4) Unit costs for transformers, HV equipment, construction and specialist trades in overlapping hub regions rise cumulatively 20–50% across 2020–2026 Scenario, while cluster tariff trajectories and queue durations remain N/A — though the construction component of that escalation is now known to be national rather than hub-specific Confirmed, concentrating the chain’s remaining weight on equipment and specialist trades. (5) The incumbent mature-node fab absorbs a power and equipment cost increase it did not cause and cannot pass through under the C3 cap. (6) Escalating industrial energy costs in high-concentration metros feed regional price paths, entering Core PCE as a supply-side, policy-generated component — marginal and regionally concentrated; no national index sensitivity is claimed or sized. (7) A supply-side inflation component monetary policy cannot address without demand destruction argues at the margin for a firmer terminal band in the SEP dot plot — the direction of pressure is the claim, not the elasticity. (8) The loop closes in direction if not in measured magnitude: industrial policy raises the incumbent’s cost base, the cost base contributes to the price index, the price index bears on the terminal rate, and the terminal rate prices the incumbent’s refinancing Scenario.

Rate Transmission

N-RT1 — Glide Mistaken for Plateau. The SEP path descends from 3.8% to 3.6% to 3.4% Confirmed — a shallow glide, not a fixed floor. Where a desk internalises it as a plateau, the risk premium applied to legacy fabs is set too high and distress scenarios are designed more pessimistically than the data supports. This node is entered against the house’s own prior framing: an earlier draft of this report made exactly that error, and the correction narrows the damage estimate rather than widening it.

Infrastructure Transmission

N-I1 — Factor-Pool Contention (Consensus Blindspot: Partial — Construction Leg Falsified). Megaproject general contractors outbid for the same trades, at cumulative hub escalation of 20–50% since 2020 Scenario. The node is downgraded on measurement, not logic: national construction prices rose +45.4% over the same window Confirmed, so general construction shows no hub premium. What survives is the non-substitutable residue — cleanroom trades, rigging crews, transformer and HV equipment — where national comparators are not yet assembled.

N-I2 — Power-Procurement Collision (Consensus Blindspot: Yes). Utilities welcome concurrent load because it grows ratebase. Rate-case lag and interconnection queues create a bottleneck that is entirely WACC-irrelevant. The house grid and transformer series (backlog and cost bands) remains N/A Speculative and is not systematically populated. Without these bands, the transmission strength from power-procurement cost inflation and grid bottlenecks into legacy-node ROIC cannot be measured — the claim stands as mechanism without magnitude. Following the construction-leg result, this node now carries a larger share of the framework’s evidentiary burden than it did at drafting.

N-I3 — WFE Order-Book Skew (Consensus Blindspot: Partial). Toolmakers allocate capacity to apex-margin orders; legacy tool lead times and prices inflate as a residual. Growing refurbished-tool dependence produces capability stagnation in exactly the foundational nodes designated as strategically critical.

Product Transmission

N-P1 — Incumbency Tax of Investment Credits (Consensus Blindspot: Yes). The sharpest node in the framework and the one that names it. Section 48D subsidises new competing capacity at 35% HOUSE-ANCHOR Confirmed while a standing, fully-depreciated asset producing the same parts earns nothing. The node survives the gradient correction intact and is strengthened by it: because the credit reaches mature capacity at 20–40% of award value Market Estimate, the incumbent’s competitor is frequently a subsidised build at its own node. It also survives the construction-leg falsification untouched, because this limb operates through asset-depreciation rather than factor prices.

N-P2 — Qualification-Moat Bifurcation (Consensus Blindspot: Yes). AEC-Q automotive and ITAR / Trusted Foundry accreditation preserve repricing power through the C3 cap. Commodity cohorts absorb it in full. This splits “legacy” into survivors and stranded assets along an axis the sell-side does not screen, because the moat is a qualification record rather than a financial metric.

N-P3 — Criticality Without Cashflow (Consensus Blindspot: Partial). Defence demand requires capacity it is far too small to fund. Automotive OEMs have qualified themselves into sole-source dependencies they decline to pre-pay for. The dependency is discovered ex-post, at a line-down event or supply audit — the trigger geometry for the intervention pressure in C4.

Regulatory Transmission

N-R1 — The 48D Cliff as Time-Shaped Distortion (Consensus Blindspot: Yes). The 2026-12-31 deadline HOUSE-ANCHOR pulls 2023–2026 capex forward and strands late movers after it. Consensus models the credit as a level effect on recipient returns; it is a shape effect on everyone’s input costs. Industry advocacy for statutory extension is active, and an extension would re-phase capex, employment and power demand across both mature and advanced nodes — a policy-cliff risk priced by no one as a two-sided variable.

N-R2 — Milestone and Renegotiation Risk (Consensus Blindspot: Partial). Disbursement pacing is administratively and politically contingent. The halo is a schedule, not a contract, and a recipient carrying announced-award assumptions in its capital plan holds an unhedged policy-execution exposure.

N-R4 — Criticality Has No Operational Definition (Consensus Blindspot: Partial). “Systemically critical” is invoked constantly and defined nowhere. An operational definition is available: a facility is critical when it holds sole-source qualification on a defence platform under Trusted Foundry accreditation, or on an automotive safety component where requalification exceeds the OEM’s inventory buffer. Against that definition, the realistic resolution path is not a public backstop — it is quiet acquisition by an investment-grade incumbent at a discount to replacement cost (held N/A), assets surviving under new ownership and only the equity extinguished. Any escalation beyond that is carried at Scenario tier throughout and is never upgraded.

N-R5 — Recipient Guardrails as a Hidden Exposure Axis (Consensus Blindspot: Partial). Award conditions restrict recipients’ expansion optionality in China. Non-recipients carry no such restriction, making their differential China exposure a risk axis no subsidy-based screen captures. The China Validated End-User framework moved to an annual licensing regime after its 2025-12-31 expiry HOUSE-ANCHOR Confirmed, converting a standing status into a binary renewal risk at every cycle.

N-R6 — Trade Shield and Domestic Price Bifurcation (Consensus Blindspot: Yes, on the bifurcation, not the supply narrative). The raw China-supply narrative is comprehensively carried by consensus; no edge is claimed on it. What is not carried is the consequence: an effective 50–75% Section 301 band Confirmed partially protects domestic sales while offshore scaling presses global legacy ASPs by −10% to −30% Market Estimate. A screen assuming a single global price cap systematically over-prices the pessimism on US-domiciled legacy ROIC. The blindspot runs in the direction of over-stated damage, not under-stated.

Consensus & Usage

Consensus under interrogation. The observable consensus is not that industrial policy lifts everything. It is that industrial policy has a boundary and stops at it — accretive to recipients, neutral to everyone else. Both the leading-edge bulls and the legacy bears hold that premise; they differ only on which side of the line to own. This report attacks the premise rather than either camp: sovereign capital deployed into a physically constrained industrial base imposes measurable costs on parties who receive none of it, through shared labour, construction and power markets, and through a credit design that finances an incumbent’s replacement. For a global or emerging-market allocator with no US semiconductor exposure, the transferable object is the mechanism, not the sector: any jurisdiction running large-scale industrial subsidy into a supply-constrained factor market generates the same incidence pattern, and the same four discriminators — geographic cost gradient, capex-deferral attribution, rating-controlled spread wedge, and utility tariff filings — test for it anywhere. The first of those has now been run against US data and returned no gradient on general construction, which is itself the transferable lesson: run the test on the non-substitutable inputs, not on the ones a national contractor market arbitrages away.

Attribution risk — single regime-invalidating line. If mature-node ROIC compression cannot be decomposed to separate domestic subsidy incidence from foreign supply glut and inventory cycle, this framework has no incremental claim and should be read as taxonomy. The first decomposition test has returned negative on the construction leg; the framework’s standing now depends on the equipment and specialist-labour legs.

Reader application lines. Long-only — re-sort mature-node exposure by moat depth crossed with co-location rather than by node class; treat maintenance-capex deferral language in filings as a solvency signal rather than a cycle signal, and weight the moat leg above the co-location leg pending the equipment/labour comparators. Event-hedge — the 2026-12-31 cliff and the 2027 capex air-pocket are dated and observable, but the date is under active extension advocacy and should be monitored as a two-sided variable. Macro-allocator — use the dark-matter chain as a template: industrial policy into a constrained factor market produces a supply-side price component that firms the terminal band, a cross-jurisdictional mechanism rather than a US-specific one; and verify whether an observed rate path is a plateau or a shallow glide before setting the risk premium, since the two imply materially different distress geometry.

Observation Markers

Dated, falsifiable, checkable by any reader against public sources. No positioning language appears in this block by design.

Marker 01 · Deadline 2026-12-31. Observable: qualifying construction-start announcements against the announced US fab pipeline, ahead of the Section 48D deadline HOUSE-ANCHOR Confirmed; and, jointly, the legislative status of extension proposals. Confirms — pre-cliff starts concentrate above a threshold held N/A, evidencing the statutory date shapes factor demand. Invalidates — starts remain dispersed with no pre-deadline concentration; the cliff is then a financing detail, and the time-shape argument falls. Re-phases — statutory extension before the deadline neither confirms nor invalidates the marker; it moves the observation date, lengthens the cost-push limb, and leaves the depreciation limb untouched. The marker is to be re-dated, not retired.

Marker 02 · Deadline 2027-03-31. Observable: FY2026 annual filings of US mature-node operators — stated attribution of any capex deferral, and any disclosed withdrawal or non-renewal of a brownfield credit facility. Confirms — deferrals attributed to input-cost escalation rather than end-demand, and/or at least one disclosed facility withdrawal inside the 30–60% wholly unsubsidised band Market Estimate; either constitutes a first-cohort crossing under the zombie criterion. Deferrals citing equipment, transformer or specialist-trade costs specifically carry more weight than those citing construction, since the construction leg is now known to be nationally uniform. Invalidates — deferrals attributed uniformly to demand conditions with no input-cost language, and no facility withdrawals disclosed.

Marker 03 · Deadline 2027-06-30. Observable: state PUC rate-case dockets in fab-cluster jurisdictions — industrial tariff determinations and cost-allocation treatment of concurrent large-load additions, specified as a differential test against non-cluster jurisdictions since the absolute tariff threshold is not yet set (held N/A). Confirms — cluster tariffs diverge measurably from non-cluster jurisdictions, cost allocation falling on incumbent industrial load, consistent with the 20–50% hub escalation already on the record Scenario; completes the fiscal-to-physical-to-price trace. Invalidates — cluster tariffs track non-cluster jurisdictions with no measurable divergence, or allocation falls on new large load rather than incumbents; combined with the construction-leg result, that outcome would leave the factor-crowding thesis without an evidenced channel.

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

The four layers. Surface Narrative — the consensus as actually held, stated in its strongest form. Organ · Macro Vectors — the structural bodies that generate the condition: the policy rate path and the subsidy stack’s eligibility geometry. Nerves · Transmission — the named chains by which the macro vectors reach a specific operator’s cash flow, each with a stated falsification test. Scars · Regime Shifts — impairment ordering, threshold crossings, and the corridor in which the whole framework dies.

Data source hierarchy. Tier 1 · Sovereign / Regulator — statutory text, Federal Reserve policy statements and SEP, SEC filings (10-K/10-Q/20-F/8-K), Federal Register/USTR/BIS records, state PUC dockets, ISO/RTO interconnection queue data, federal statistical series including producer price indices. Tier 2 · Listed IR / Consensus — investor-relations disclosure, earnings materials, rating-agency methodology, sell-side consensus used as a positioning read. Tier 3 · Macro Proxies — construction/wage indices, tariff schedules, equipment lead-time surveys, private-credit and ABS transaction records, house-maintained grid and transformer series where populated.

Analytical labelling. Confirmed — Tier 1 verified, single value or bounded statutory band. Range — Tier 1/2 verified but expressed as a band or projection distribution. Market Estimate — consensus, third-party or audit-reconstructed estimate, never a house forecast. Scenario — house-constructed conditional state; every matrix cell and backstop proposition sits here by default. N/A — not observable from public disclosure; recorded as a census or data gap and never estimated.

Standing constraints. Zero fabrication — every unresolved quantity is carried as a flagged gap; no numeric value appears unless it originates in the house anchor library, the audit inventory, a Master override, or a primary-source exhibit. Unit and context match — no audited band is substituted into a placeholder without verifying the unit matches context; mismatches are logged. Credibility inheritance — predicates flagged as suspect are written in conditional register and never stated as fact. Counter-trend integrity — where audit correction inverted a premise, the reasoning was rewritten to defend the corrected quantity rather than substituting numbers into unchanged logic. Adverse evidence disclosure — where a resolved datapoint reports against this framework, it is recorded at the point of the claim it undermines and propagated to every dependent section. Directional call suppression — no single-name recommendation, price target, ratio or trade list appears anywhere. Operating-entity precision — metrics are attributed to the filing registrant, not to a group or brand; no named natural person appears in this document. Scope discipline — primary analytical nodes are US-domiciled only. Disclosed bias — availability bias toward large-cap registrants understates the private and sponsor-owned fab tail, recorded as a bounded census gap. Self-correction on the record — where audit or override invalidated a claim made in an earlier draft — the rate-path characterisation, the populated status of the house grid and transformer series, and the geographic gradient in construction cost — the correction is stated in the body rather than silently absorbed.