---
1. The IRA Tax Credit Pull Architecture is the single highest-load node in the graph (36 connections, w=8.5), and it structurally generates its own destruction.
The node has outbound edges creating the fiscal cost explosion (`causes_when_successful`), triggering the global arms race (`triggered_by_scale_and_speed_of`), and crowding out construction capacity (`demand_pull_mechanism_accelerates_crowding_out`). It has inbound edges from OBBBA (`partially_dismantles`), the WTO DS623 ruling (`legally_invalidates`), and the IRA Clean Energy Investment Collapse (`proves_fragility_of_demand_pull`). The same mechanism that makes the IRA work — uncapped, fast, private-capital-leveraging — generates the fiscal, political, and legal pressures that subsequently undermine it.
2. Tacit Knowledge (31 connections, w=8) functions as the non-substitutable constraint across all industrial policy domains — semiconductor, battery, and critical minerals.
The node receives inbound edges from Korea's DRAM strategy (`accumulated_over_30_years`), China's mineral processing monopoly (`mirrors_mechanism_of_processing_expertise`), and the Agglomeration Dividend (`is_primary_transmission_mechanism_of`). It has outbound edges explaining failure in Made in China 2025 semiconductors, Rapidus's leapfrog risk, and the CHIPS Act workforce gap. Only two mechanisms in the graph claim to partially address it: SEMATECH's co-location model (`directly_addressed_via_colocation`) and TSMC Arizona's workforce transfer (`solved_via_workforce_transfer_not_domestic_training`). Subsidies appear consistently on the supply side of the problem, not the solution side.
3. The graph contains two "weight anomalies" that are analytically significant.
TSMC Geopolitical Chokepoint (w=1, 20 connections) and CHIPS Act Foundry Subsidy Mechanism (w=1, 18 connections) are among the most structurally connected nodes but carry the lowest node weights. Both appear as duplicate entries (one as `idea`, one as `thing` or with different weights). These nodes function as convergence anchors — most paths in the US semiconductor policy subgraph route through them — but their low assigned weights suggest they were treated as structural connectors rather than independently analyzed concepts. The graph's hub analysis understates their structural role.
4. The Additionality Problem (17 connections, w=8) has no complete solution in the US or EU policy subgraphs.
India PLI (`structurally_eliminates`), SEMATECH (`solves_via_matching_contribution_requirements`), and Wright's Law (`resolves_by_showing_spillover_benefits_exceed_additionality_concern`) each claim partial resolution. The CHIPS Act equity conversion (`attempts_to_solve_via_alignment`), CHIPS milestones (`partially_solves_via_milestone_verification`), and Mazzucato-Rodrik conditionality (`provides_conditionality_design_solution`) each claim alternative partial approaches. No single node in the US or EU context carries an unconditional `eliminates` or `solves` edge toward this problem.
5. The WTO enforcement vacuum creates a legal-but-unenforced environment that all major industrial policies now depend on.
WTO Appellate Body Void (`enables_by_removing_enforcement_constraint`) connects to Global Industrial Policy Subsidy Arms Race, IRA Tax Credit Pull Architecture, and CHIPS Act Foundry Subsidy Mechanism. The IRA WTO DS623 Domestic Content Ruling 2026 (`legally_invalidates`) the IRA domestic content tier — while simultaneously the WTO Appellate Body Void makes enforcement structurally impossible. The graph records an active ruling with no mechanism of compliance.
---
Loop 1: IRA Fiscal Success → Political Reversal → Investment Collapse → Undermines IRA
- IRA Tax Credit Pull Architecture `causes_when_successful` → IRA Fiscal Cost Explosion: $383B → $1.1T
- IRA Fiscal Cost Explosion `provided_fiscal_justification_for` → OBBBA IRA Rollback 2025
- OBBBA IRA Rollback `triggers` → IRA Clean Energy Investment Collapse 2025
- IRA Clean Energy Investment Collapse `proves_fragility_of_demand_pull` → IRA Tax Credit Pull Architecture
The mechanism's success (investment leverage) generates the fiscal data that enables its political dismantling. The loop runs in approximately 3 years (2022 IRA → 2025 OBBBA).
Loop 2: AI Demand → TSMC Dependency → Reshoring → AI Demand Exceeds Timeline
- AI Compute Hypercapex TSMC Dependency Paradox `paradoxically_deepens_strategic_dependency_on_via_600B_AI_capex` → TSMC Geopolitical Chokepoint
- TSMC Geopolitical Chokepoint motivates (via multiple paths) → CHIPS Act Foundry Subsidy Mechanism
- CHIPS Act Foundry Subsidy Mechanism aims to reduce TSMC dependency
- AI Compute Hypercapex TSMC Dependency Paradox `undermines_TSMC_independence_goal_by_outrunning_domestic_fab_timeline` → CHIPS Act Foundry Subsidy Mechanism
The AI investment surge simultaneously provides the economic justification for reshoring (national security / economic concentration risk) and accelerates TSMC's economic indispensability faster than the reshoring timeline can close.
Loop 3: MIC 2025 EV Success → Western Green Subsidies → Arms Race → MIC 2025 EV Reinforcement
- Made in China 2025 Asymmetric Track Record `EV_and_clean_energy_success_directly_triggered` → Global Green Subsidy Arms Race
- Global Green Subsidy Arms Race `amplifies` → EV Policy Divergence Spiral
- China MIC 2025 Asymmetric Scorecard `constitutes_supply_side_manufacturing_engine_of` → EV Policy Divergence Spiral
- EV Policy Divergence Spiral reinforces China's structural lead (`CATL_cost_gap_is_structurally_permanent` per New Trade Theory edge)
Western subsidy responses to China's EV dominance accelerate market bifurcation without closing the underlying cost-curve gap, because the Wright's Law learning curve advantage China holds predates the policy response.
Loop 4: Arms Race → EU Fragmentation → EU Falls Further Behind → Arms Race Pressure Increases
- Global Industrial Policy Subsidy Arms Race `worsens_member_state_competition_within` → EU Chips Act Sovereignty Fragmentation Trap
- EU Chips Act Sovereignty Fragmentation Trap `competitive_pressure_makes_more_urgent_and_larger` → Draghi Report EU €800B Investment Gap
- Draghi Report `competitive_pressure_makes_more_urgent_and_larger` → Global Industrial Policy Subsidy Arms Race (per `EU_competitive_response_to_losing_in`)
The arms race widens the EU's relative investment gap, which increases the diagnosed urgency, which cannot be addressed due to structural fragmentation — which is itself worsened by the arms race.
---
Solyndra (2011) → IRA Tax Credit Design (2022)
The edge `directly_caused_tax_credit_over_grant_design_of` between Solyndra Loan Guarantee Political Trauma and IRA Tax Credit Pull Architecture is non-obvious and consequential. A single loan guarantee failure — which the graph elsewhere notes was `actually_caused_by_cheap_solar_dumping_from` Made in China 2025 — structurally precluded the grant-based design that would have preserved accountability. The fiscal cost explosion ($383B → $1.1T) is, in part, a second-order consequence of Chinese solar dumping in 2011.
Wright's Law as Resolution to the Additionality Problem
Wright's Law Learning Curve `resolves_by_showing_spillover_benefits_exceed_additionality_concern` → Industrial Policy Additionality Problem. This is non-obvious because additionality is typically treated as a deadweight problem (subsidizing what would happen anyway). Wright's Law reframes it: even if the subsidized firm would have invested without support, the social learning curve spillovers (cost reductions that benefit the entire sector) may exceed the deadweight cost. This edge provides a theoretical escape from the additionality critique without requiring incremental production design.
EU Defense Fiscal Exception as de facto Industrial Policy
EU Defense Fiscal Exception as Industrial Policy Wedge `funds_track_3_defense_while_starving_tracks_1_and_2` → EU Strategic Autonomy 3-Track Divergence. The connection reveals that the EU's largest near-term industrial policy instrument is not the NZIA, CISAF, or Chips Act — it is a budget rule exception for defense spending. The graph records this as a wedge that inadvertently defunds civilian industrial policy through fiscal space competition.
Tacit Knowledge as the Underlying Mechanism of Agglomeration Dividends
Industrial Policy Agglomeration Dividend `is_primary_transmission_mechanism_of` → Tacit Knowledge Bottleneck. Geographic clustering works not primarily through supply chain efficiency or labor pooling but through tacit knowledge transmission. The implication is that single-facility subsidies (CHIPS Act individual fab grants) structurally cannot capture agglomeration dividends even if they successfully build the facility.
Japan's IDM-to-Fabless Miss as Origin of TSMC Dependency
1986 US-Japan Semiconductor Trade Agreement `fatally_wounded_Japan_which_was_finished_by` → Japan Semiconductor IDM-to-Fabless Miss → `Japan_failure_created_structural_vacancy_filled_by` → TSMC Geopolitical Chokepoint. The specific chokepoint that motivates ~$200B in combined US/EU/Japan/Korea policy responses traces through Japan's organizational failure to adapt to the fabless model — itself triggered by a bilateral trade agreement. The current global policy response is addressing a path-dependent outcome from 1986-1990s organizational decisions.
---
IRA Tax Credit Pull Architecture (36 connections) functions as the graph's primary policy innovation node. It receives causal inflows from design decisions (Solyndra trauma, 45X credit), legitimating theories (Wright's Law, crowding-in evidence), and competitive pressures (Canada, EU responses). It generates outflows to the arms race, fiscal explosion, battery manufacturing boom, and its own political vulnerability. Its structural role is as a policy mechanism that simultaneously achieves deployment speed, creates fiscal exposure, generates geopolitical competition, and produces political coalition (IRA Red State Paradox) — making it both the most effective and most contested mechanism in the graph.
Tacit Knowledge Bottleneck (31 connections) functions as the graph's primary structural constraint. It explains both the durability of existing advantages (TSMC, Korea's DRAM lead, China's mineral processing) and the limits of policy interventions (Rapidus risk, workforce gap, agglomeration necessity). Its high connection count reflects the fact that it operates as a hidden variable in explanations that appear to be about cost, capital, or policy design but are actually about accumulated operational knowledge.
Rodrik "Let Losers Go" Principle (29 connections) functions as the graph's primary evaluative criterion. Most industrial policy failures in the graph carry edges connecting them to violations of this principle (EU champion model, Germany Energiewende, Trump equity conversion), while most successes carry validating edges (Korea DRAM, Taiwan ITRI, India PLI, SEMATECH). Its high connection count reflects its role as the convergence point for the "what distinguishes success from failure" question across the graph.
Global Industrial Policy Subsidy Arms Race (24 connections) functions as the graph's systemic-level node — the aggregate outcome of individual country decisions, fed by the WTO enforcement vacuum and legitimized by Washington Consensus Inversion. It receives inflows from the IRA, MIC 2025, and Airbus precedents, and generates outflows to EU fragmentation, subsidy cliff exposure, and Global South disinvestment. Its role is as an emergent coordination problem: individually rational national decisions produce a collectively suboptimal escalation.
---
1. Ally-Shoring vs. Domestic Capability
The graph records a structural tension between two goals embedded in the CHIPS Act: reducing geographic concentration (by building capacity in allied nations) and building US domestic technological sovereignty. TSMC Arizona `partially_mitigates` TSMC Geopolitical Chokepoint but `explains_why_arizona_fabs_remain_taiwan_dependent_despite` the subsidy investment (via Tacit Knowledge Bottleneck). The graph does not resolve whether the CHIPS Act is primarily a sovereignty program or a geographic diversification program — these goals require different success metrics.
2. Silicon Shield Erosion Paradox — Direction of Effect
The graph records two competing inferences from TSMC Arizona production: it `weakens_taiwan_deterrent_as_side_effect_of` TSMC Arizona (silicon shield erosion) and simultaneously `AI Compute Hypercapex TSMC Dependency Paradox contradicts_by_making_TSMC_more_economically_irreplaceable_than_ever`. The deterrence logic (Taiwan matters strategically because it produces critical chips) is being simultaneously weakened (by geographic diversification) and strengthened (by AI-driven demand growth). The graph identifies both effects but does not resolve their relative magnitude.
3. Tax Credit vs. Grant — No Mechanism Solves the Full Trilemma
Industrial Policy Grant vs Tax Credit Design Tradeoff `operationalizes` the Speed-Accountability-Scale Trilemma. The graph records three partial solutions: Operation Warp Speed (advance market commitment — fast and scalable but sector-specific), India PLI (accountable and scalable but slower deployment), and SEMATECH (accountable and fast but pre-competitive scope only). No mechanism in the graph achieves all three simultaneously at full industrial scale.
4. WTO Ruling with No Enforcement
IRA WTO DS623 Domestic Content Ruling 2026 `legally_invalidates` the domestic content bonus tier while WTO Appellate Body Void `shields_from_enforceable_WTO_challenge` the same policy. The graph records a legal ruling that cannot be enforced and an enforcement vacuum that makes the ruling moot — but does not record what the medium-term equilibrium is. Whether WTO DS623 creates diplomatic leverage, future retaliatory risk, or is simply absorbed without consequence remains unresolved.
5. Democracy Time Horizon vs. Industrial Policy Durability
Industrial Policy Democracy Time Horizon Problem `adds_sixth_condition_to` the 5-condition Grand Synthesis. The IRA Tax Credit Constituency Lock-In Mechanism `partially_solved` this, but IRA Red State Paradox `partially_failed_to_prevent` OBBBA rollback. UK IS-8 `attempts_to_solve_via_10_year_commitment_and_statutory_ISC`. The graph records multiple partial solutions but no mechanism has been tested through a full political cycle — the OBBBA event is the only data point on survivability, and it shows partial dismantling.
---
H1: Programs that transfer existing tacit knowledge outperform programs that require building it.
Derivation: Tacit Knowledge Bottleneck has two partial solutions — SEMATECH co-location and TSMC Arizona workforce transfer. All "build from scratch" programs (Rapidus, new Intel nodes, EU Chips Act greenfield) carry `most_severely_constrained_by` edges to the same node. Testable prediction: TSMC Arizona Phase 2 will achieve yield targets faster and at lower cost overrun than Samsung Austin or Intel Ohio on equivalent nodes.
H2: Industrial policies with explicit exit mechanisms survive political cycles longer than those without.
Derivation: Rodrik "Let Losers Go" Principle has the highest evaluative connectivity in the graph. CHIPS Act Milestone-Based Disbursement Gap `provides_exit_mechanism_required_by` this principle. OBBBA partially dismantled the IRA but did not roll back CHIPS Act grants already disbursed. Testable prediction: CHIPS Act commitments with milestone-based structures will be honored through a second political cycle; uncapped IRA credits for new projects will continue to face rollback risk.
H3: EU industrial policy deployment speed is structurally capped by the grant/member-state architecture regardless of policy ambition level.
Derivation: EU CISAF `institutionally_incapable_of_replicating_mechanism_of` IRA Tax Credit Pull Architecture. EU NZIA Tax Credit Sovereignty Gap `cannot_replicate_due_to_constitutional_constraint`. EU Chips Act targets have slipped. Testable prediction: comparing deployment velocity (private investment committed per euro of announced public support, within 24 months of policy announcement) will consistently show the US tax credit model achieving 3-5x faster private capital mobilization than EU grant/state-aid models.
H4: The global subsidy arms race will converge toward tax credit models and away from grant models under competitive pressure.
Derivation: Korea K-CHIPS Act `empirically_demonstrates_speed_advantage_of_credit_side_of` the grant-vs-credit tradeoff. Canada's IRA competitive response is grant-based but was triggered specifically by the tax credit mechanism's speed. Testable prediction: within the current subsidy arms race cycle, nations that have not yet committed to a mechanism will adopt credit-or-credit-equivalent instruments at higher rates than grant instruments.
H5: Programs that solve the additionality problem by design (India PLI-type) will show measurably different cost-per-unit-of-incremental-output ratios than programs without this design.
Derivation: India PLI Additionality-by-Design `structurally_eliminates` the Industrial Policy Additionality Problem via incremental production payment. IRA 45X and CHIPS grants do not contain this mechanism. Wright's Law provides a theoretical escape but not a measurement solution. Testable prediction: India PLI mobile manufacturing targets will show higher correlation between subsidy disbursement and attributable production increment (net of trend) than IRA solar manufacturing credits over comparable periods.
H6: The tacit knowledge accumulation timeline, not the capital subsidy level, is the binding constraint on semiconductor reshoring speed.
Derivation: CHIPS Act 2026 Production Reality `confirms_60pct_cost_premium_persists` and `shows_progress_too_slow_to_resolve_AI_era_dimension`. TSMC Arizona resolved workforce constraints via workforce transfer (not domestic training). Tacit Knowledge Bottleneck `root_cause_of` the workforce gap. Testable prediction: the yield gap between TSMC Arizona and TSMC Taiwan on equivalent nodes will narrow at a rate consistent with workforce experience accumulation (5-7 years) rather than with capital deployment timelines.