Cascade Simulation — Disruption Propagation Through the AI Compute Supply Chain

Gringras, David

Cascade

2 corrections · latest May 19, 2026Fresh10d

If a node, region, or supply-chain layer fails, what stops — where, and how fast. A quantitative propagation model that walks the supplier–customer graph one edge at a time, weighting each hop by trade-flow share, editorial replaceability, and the stock-price correlation between coupled firms. The Engine runs any disruption you build; Scenarios pins it to named geopolitical stress cases, each carrying its own dated threat analysis.

Cascade

2 corrections · latest May 19, 2026Fresh10d

Engine

Remove any node, region, or supply-chain layer and watch the disruption propagate forward through every structurally exposed firm — weighted by trade-flow share, editorial replaceability, and the stock-price correlation between coupled firms. The graph merges licensed supplier–customer relationships, licensed competitor edges, and CSET ETO chip-deployment chains into a single firm-edge graph (21,037bilateral edges, full substrate). Per-edge criticality is the product of editorial replaceability (1–10) and 3-month stock-price correlation on the ~40% of edges where the price series is available (8,392 of 21,037); sole-source edges are floored at criticality 9 regardless of the correlation reading. Every weight carries its methodology trail inline; for the static hourglass view of layer-by-layer concentration, without temporal propagation, see Chokepoint Anatomy.

2 corrections · latest May 19, 2026

Show all 20 cascade scenarios as a table (alternative to the interactive Cytoscape graph)

All 20 pre-built cascade scenarios with seed nodes, severity, and propagation decay. Severity is the initial impact magnitude at the seed; propagation decay attenuates impact per supply-chain hop. Sourced from SEC 10-K filings, USGS Mineral Commodity Summaries, TrendForce / IDC / SemiAnalysis market-share reports, and editorial assessments where primary sources are unavailable. Each row links to the interactive cascade run.
Scenario	Category	Seed nodes	Severity	Sev. tier	Propagation decay	Direction
TSMC Taiwan Disruption	natural disaster	org-tsmc	1.00	T4	0.90	downstream
ASML Export Halt	export control	org-asml	1.00	T1	0.95	downstream
US–China Full Decouple: Revenue Impact	geopolitical	org-nvidia, org-amd, org-intel, org-qualcomm	0.15	T1	0.70	upstream
US–China Full Decouple: Access Denial	geopolitical	org-smic, org-huawei, org-baidu, org-bytedance, org-alibaba, org-tencent	0.55	T3	0.80	downstream
CoWoS Advanced Packaging Bottleneck	supply bottleneck	org-tsmc	0.40	T2	0.70	downstream
SK Hynix HBM Disruption	supply bottleneck	org-sk-hynix	0.60	T2	0.80	downstream
EDA Export Restriction	export control	org-synopsys, org-cadence	0.90	T3	0.80	downstream
Chiplet Assembly Disruption	supply bottleneck	org-tsmc, org-synopsys, org-cadence	0.80	T4	0.75	downstream
Helium Supply Crisis (Hormuz)	supply bottleneck	org-samsung, org-sk-hynix, org-tsmc	0.65	T1	0.75	downstream
InP/EML Laser Shortage	supply bottleneck	org-coherent, org-lumentum	0.85	T2	0.80	downstream
ABF Substrate Monopoly (Ajinomoto)	supply bottleneck	org-ajinomoto	0.90	T1	0.95	downstream
Gallium/Germanium Restriction (Nov 2026)	geopolitical	org-coherent, org-skyworks-solutions, org-qorvo, org-wolfspeed, org-infineon	0.70	T1	0.65	downstream
CoWoS 2026 Allocation Constraint	supply bottleneck	org-tsmc	0.85	T1	0.85	downstream
Japan Semiconductor-Input Seismic Shock (Korea coupling)	natural disaster	org-shin-etsu, org-sumco, org-tokyo-electron, org-advantest, org-disco, org-lasertec, org-tokyo-ohka-kogyo, org-resonac, org-kioxia	0.70	T4	0.82	downstream
Allied Coordination: TSMC Advanced Node Restriction	policy intervention	org-tsmc	1.00	T4	0.90	downstream
EDA License Restriction: China Entity List Expansion	policy intervention	org-synopsys, org-cadence	0.95	T4	0.80	downstream
Netherlands DUV Export Restriction to China	policy intervention	org-asml	0.70	T4	0.85	downstream
China Critical Minerals Retaliatory Restriction	policy intervention	org-coherent, org-skyworks-solutions, org-qorvo, org-wolfspeed, org-infineon	0.70	T4	0.65	downstream
Multilateral Cloud Compute Access Restriction	policy intervention	org-amazon, org-microsoft, org-google-cloud, org-alphabet, org-oracle	0.80	T4	0.75	downstream
Netherlands EUV Restriction: Allied Countries Only	policy intervention	org-asml	1.00	T4	0.95	downstream

Peer attribution

Data Provenance

Primary35%(7,336)Research65%(13,702)

Sources

Source	Tier	Records
Licensed supply-chain database (bilateral + segment pulls; held under subscription, not redistributed)	T2	13,498
SEC EDGAR (10-K/20-F disclosures re-sourced to primary filings)	T1	3,170
Re-derived from public primary sources (SEC EDGAR / public registries / company disclosure)	T1	4,166
Other (BIS crossref + CSET ETO + research-residue)	T2	203
Epoch AI Facilities	T2	—

Processing

BFS propagation across the supply chain graph with weighted decay (supply share × substitutability × correlation-blend); per-edge 3-month stock-price correlation augments the editorial 1–10 criticality where available, with sole-source edges floored at criticality 9. Severity parameters from SEC 10-K filings (Tier 1–2) and structural assessments (Tier 3–4). All propagation-decay values are editorial assessments (Tier 4).

Reading this data

~valueEstimated — derived, modeled, or inferred. Hover for method and source.

valueDerived from another measured field (e.g. power from GPU count).

Source conflict — multiple sources disagree. Click for competing values.

hoverHover any annotated value to see authority tier, method, source, and vintage.

Full methodology

Scenario Analysis

Pre-built scenarios using this simulation engine

Chokepoint Anatomy

Static concentration behind the dynamics

Stress Index

Current fragility composite from 4 sub-indices

Supply ChainControls & Supply

The graph this simulation runs on

Chokepoint Activity Feed

Live activity feed for the chokepoint layers in these scenarios

Sovereign AICompute & Capability

Programs whose deployment timelines are bounded by the chip-supply graph this propagates over

Impact Pre-Scorer

Single-organization removal estimator (BFS over the same graph)

Built for the AI governance research community