Data 2026-03

Scenarios

Five stress cases over the AI-compute supply chain, each dated by when it was last reviewed and by the age of its newest input.

Taiwan Strait: TSMC Disruption Cascade

No recorded reviewNewest input 2026-02Runs against the cascade engine

No review recorded. The analysis carries dated sources but no human has re-read it against current conditions since the currency machinery was introduced. The propagation figures on this page come from a live run against the supply-chain graph; the threat and capacity analysis around them is human-authored and sourced individually. Newest input dated from TSMC Arizona substitution entry, vintage 'Q4 2024 (Phase 1 launch); Feb 2026 (Apple commitment)' — src/lib/data/taiwan-strait-data.ts. Review state for all five analyses is on the index.

Taiwan’s leading-edge fabrication concentrates global AI compute on one island. The four scenarios here — blockade, targeted strike, power-grid failure, export restriction — quantify what changes downstream for US and allied labs when that single node is disturbed, propagated across 18,999 supply-chain edges.

Supply Chain Exposure Analysis

6 TSMC fabs, 5 alternative foundry sites. Cascade propagation across 0 supply chain edges in this view (public-provenance subset of 18,999); substitution gap 85-95% for advanced process node capacity. Active: Naval BlockadeTaiwan manufactures over 90% of the world's most advanced AI chips. Four disruption scenarios model how that concentration translates to downstream risk. Active: Naval Blockade

Choose scenario

Severity

Price correlation

Taiwan manufactures over 90% of the world's most advanced AI chips at TSMC fabs that no combination of alternative foundries can replace within 3–5 years. The cascade engine propagates each disruption through 0 real supply-chain edges from a licensed supply-chain database and public filings; substitution capacity outside Taiwan covers under 20% of advanced-node demand under any of the four scenarios below.

Methodology →Calibrated against 1999 Chi-Chi · 2011 Tohoku · 2016 Kumamoto precedents

Active scenarioNaval Blockade

Supply Chain CascadeAnimated propagation: how the disruption ripples outward from TSMC fabs through 0 supplier-customer edges in this view (public-provenance subset of 18,999) to AI labs.

Concentration AnalysisHow concentrated is the fab layer? Country-level share for leading-edge foundry, measured HHIs for total foundry revenue and EUV lithography, plus key customer dependencies.Taiwan ~92% of leading-edge (country-level)

Substitution PathsWhich alternative foundries can absorb TSMC demand at each process node, and where capacity falls short of viability.6 paths

Sovereign Program CascadeSovereign AI programs (Stargate, EU AI Factories, etc.) ranked by their TSMC-Taiwan dependency and substitution exposure.

Temporal CurvePer-week affected-compute trajectory with empirical buffer-range uncertainty band; inflection points (buffer exhaustion, substitution ramp).

Chinese SubstrateHow close SMIC + China-domestic alternatives sit to TSMC parity, by node generation and equipment-access constraints.

Crisis EnforcementBIS / OFAC / Section 232 enforcement levers and historical precedent for escalating Taiwan-related controls.

Scenario DetailsReference cards for each of the four disruption profiles — severity, affected fabs, time-to-impact, historical precedent.4 scenarios

MethodologyBFS propagation engine, severity / decay parameters, price-correlation weighting, calibration against historical precedents.

Rerun this disruption in the engine

This is the same full-density disruption the panels above run — a complete loss of TSMC. Opening it in the engine gives you the analyst controls this page doesn’t: hold the TSMC cascade side-by-side with another scenario or organization, step through the temporal-dynamics view as inventory buffers deplete, and adjust the correlation weighting. The four Taiwan scenarios here (blockade, strike, grid, export restriction) are severity/decay variations on this same seed.