SSCIM — the Semiconductor Supply Chain Intelligence Map — is a live intelligence layer over the global chip ecosystem. This page explains what it is, how it thinks, and how to use every part of it in about five minutes.
Semiconductors are the most geopolitically concentrated industry on earth: one company in the Netherlands makes every EUV lithography machine, one island fabricates most leading-edge logic, one country dominates HBM memory. Excellent static maps of this structure exist — but they can't tell you what changed today, which nodes are newly exposed, and how a shock will travel.
SSCIM fills that gap with three synchronized layers over one computational engine:
Layer 1 — World Map. Real OpenStreetMap geography with all 16 countries participating in the chain. Node size shows how much of the chain a country touches; color shows its current risk level. Every country score is derived from its stage participation — nothing is hand-set.
Layer 2 — Industry Flow. The complete production pipeline: 24 stages from research/IP and EDA through wafers, chemicals, five equipment categories, three fab types, chip products, packaging, and end markets — connected by 34 value-weighted dependency edges.
Layer 3 — Intelligence Panel. The event feed, the company impact ranking, and a detail view that always shows its work: score breakdowns, engine arithmetic, first- and second-order effects, and hop-by-hop impact spread.
Underneath sits one propagation engine with three uses: live events, hypothetical scenarios, and company-disruption simulations all run through the identical code path. When an event lands, it decays over time (half-strength in about twelve days), travels downstream along value-weighted edges, and echoes one hop upstream — and every affected node, company, and country updates at once.
In the flow graph, edge thickness is the share of value flowing down that path, and each stage's dot and border size is its computed importance (structural centrality blended with economic value). The WHAT CHANGED strip at the top always summarizes the current state in one line — it turns copper when a scenario is active.
Start with the top card — the U.S. export-control expansion. The map highlights affected countries, the flow highlights affected stages, and the detail view shows the engine arithmetic (severity × confidence × time decay), the chain-impact index, and the impact spread tree: source companies at hop 0, their direct downstream at hop 1, second-order at hop 2, each with an exposure number.
Tap any stage in the flow — say Deposition. A subsection opens beneath the graph listing the major companies with their market shares (Applied Materials 30%, Tokyo Electron 25%, Lam 15%…), their current shock exposure, and each one's top customers with percentages.
Tap a company card. You get its Company Impact Index — a simulated full disruption injected as 10 × market share at every stage it occupies and propagated through the engine — plus its production footprint, its customers and suppliers with sales shares, and two spread trees: the customer-graph view (ASML → TSMC 35% → Apple/NVIDIA…) and the structural stage-level view.
The second tab of the intelligence panel ranks every company by chain impact. TSMC, ASML and NVIDIA emerge at the top from the mathematics, not by assertion — tap any name to see why.
The header buttons inject simulated events — Taiwan Strait crisis, China materials ban, Export controls max — into the same engine. Copper +deltas appear on every affected stage and country; company exposures recompute. Baseline resets everything.
The ⚡ GP Briefing button composes a full daily intelligence briefing from the current model state — what changed, most-shocked nodes, company exposure leaders, country risk board, watch-next — ready to copy. Run it in a scenario and you get the scenario briefing. This is the product GP News subscribers receive each morning.
The ⓘ Methodology button documents every formula, every propagation factor, which components are computed versus analyst judgment, and the known limitations. The ? Guide button inside the app repeats this walkthrough in short form.
Four of six risk components are computed; two are declared analyst inputs, tagged as such on every breakdown bar. Explainability isn't a feature here — it's the entire trust model.
The current build runs on curated sample data: every market share, customer percentage, policy entry and event is illustrative, chosen to be directionally realistic but not cited. Production data ships with a citation and date on every figure, and propagation parameters get calibrated against documented historical episodes before scores appear without the sample label. Everything SSCIM produces is descriptive supply-chain analysis — never investment advice.