Meridian Instruments (fictional) — Vendor Exposure Review

1. The one-page action list

The five findings a procurement leader would act on first, each with the number that justifies it.

1. 6 of 28 vendors (21%) sit behind the same upstream chokepointsAjinomoto, ASML, Carl Zeiss SMT, Lasertec, TRUMPF — each a monopoly-position supplier reachable within three tiers upstream of Diodes Incorporated, NXP Semiconductors, Renesas Electronics, WPG Holdings, Arrow Electronics, Avnet. One disruption at any of these touches 21% of this vendor list at once.
2. 12 companies sit upstream of BOTH Applied Materials and Lam ResearchA concentration invisible from either vendor relationship alone (11 of 12 individually researched; the rest ride a shared segment template — worth confirming before acting): Advanced Energy, Atlas Copco, Brooks Automation, Comet Group, Ferrotec, Horiba, Ichor Holdings, INFICON, +4 more.
3. 21 of 28 vendors (75%) are domiciled in just two countriesUnited States (11) and Japan (10). Domicile is not fabrication location — see Section 4 and the methodology caveat — but a shared home country still means shared regulatory, currency and export-control exposure.
4. 21 of 28 vendors (75%) sit in segments exposed to GalliumUSGS: China supplies ~99% of world production (US net import reliance: 100% (of reported consumption)). This is a segment-level exposure category, not a claim about any one company's bill of materials — see Section 4.
5. 14 of 42 submitted vendor names (33%) aren't in the map at allThat is the map's edge, not a failure of the analysis — see Section 5. This specimen's 67% match rate is HIGHER than a typical real engagement's (well under half, per the delivery checklist): this list was built to showcase real chain analysis, so it skews toward well-documented suppliers. Expect more tier-3 shops and distributors in an actual client list.

2. Chokepoint exposure

For each matched vendor: which single points of failure it sits behind, computed by the same reach model as wafergraph's own /risk page (suppliers, and their suppliers, up to three tiers upstream; a "chokepoint" is any company at monopoly market position found along that path).

Fragility (0-100) is wafergraph's own supplier-concentration score: a weighted sum over a vendor's direct suppliers by market position and scarcity of substitutes, normalized against the highest-scoring company in the full 565-company dataset. It is a relative ranking, not a probability.

3. Shared upstream

Intersecting the matched vendors' own supplier sets: any company that shows up as a direct supplier behind two or more of them is a concentration risk invisible from any single vendor relationship. Each edge is labeled with its provenance — researched (individually documented), shared (corroborated across a small 2-4 company set), or segment-inferred (a broad template) — exactly the split wafergraph's own /methodology page discloses for the full 565-company graph.

4. Country & mineral concentration

By vendor domicile

Caveat: country here is a vendor's HQ domicile (where it is legally incorporated), not where its fabs, mines or assembly lines actually sit. A US-domiciled vendor can still run 100% of its production through a single foreign fab; domicile concentration and production concentration are different exposures and this table only measures the first.

By mineral (USGS-tracked commodities relevant to matched vendors' segments)

Exposure here means a vendor sits in a market segment (materials, analog/power/RF components, front-end equipment) where this mineral is a documented input class — it is not a claim that any specific vendor sources that specific mineral. Source: USGS Mineral Commodity Summaries (see data/minerals.json for the full citation per commodity).

5. The unmatched names

14 of 42 submitted names (33%) did not match a company in the 565-company map. These are not a failure to hide — a real engagement's delivery checklist calls the unmatched list the single most valuable output of the whole exercise, because it names exactly which vendors the free public map is missing, and researching the important ones by hand is where a real engagement earns its fee.

• a regional PCB fabricator (Texas)
• a metals stamping & fabrication shop (Ohio)
• a contract machining house (Wisconsin)
• an injection-molded plastics supplier (Guangdong)
• a wire harness assembler (Baja California)
• a local contract manufacturer / EMS shop (Shenzhen)
• a specialty adhesives & potting-compound distributor
• a custom cable & connector assembly shop
• an independent calibration & test lab
• a freight forwarder / customs broker
• a regional fastener distributor
• a tooling & fixtures supplier
• a conformal-coating & surface-finishing service provider
• a small-batch SMT assembly house (Vietnam)

In a real engagement, each of these gets researched by hand against public disclosure (supplier press releases, trade-press coverage, the client's own SEC filings if applicable) — billable work, and the part a graph alone cannot do.

6. Limits

• Curated sample, not a census. wafergraph tracks 565 companies across the semiconductor & AI supply chain — a deliberately curated set of the largest, most-cited names in each segment, not every company on earth that touches this industry. A vendor list drawn from any real company's actual spend will always contain names outside that set (see Section 5).
• In-degree measures documentation as well as criticality. 15 of the 28 matched vendors here carry no recorded supplier list at all in the public map (mostly materials and passive-component producers — the upstream-most tier, which the graph tracks primarily as suppliers TO others, not what supplies them). A company showing "no chokepoints reachable" above may simply be under-documented, not genuinely unexposed. wafergraph discloses the same bias for the full graph at /methodology.
• Edge provenance is mixed, and labeled honestly rather than smoothed over. Of the 28 matched vendors, 15 carry no supplier list at all (previous bullet); of the 13 that do, 7 carry an individually researched list, 2 a list corroborated across a small group, and 4 a list that is a broad segment template or dominated by one — the same categories wafergraph publishes for its entire dataset at /methodology (which additionally breaks out template-dominated lists as their own line), applied here to this vendor list specifically.
• Snapshot, not a feed. Every figure in this document is as of Jul 26, 2026, wafergraph's own dataset stamp. Company facts, market caps and supplier edges change; this document does not update itself. A real annual re-scoring engagement is priced separately for exactly this reason.
• Chokepoint reach is capped at three tiers upstream. A dependency four or more tiers removed from a matched vendor will not appear in Section 2, even if it is real.