Graph written 2026-09-06
Supply Chain
Supplier–customer relationships between named firms, and the layers with only one supplier.
Supplier–customer relationships between named firms, and the layers with only one supplier.
Herfindahl–Hirschman index of each layer’s supply market
Lithography, deposition, etch · extreme criticality
measured, EUV lithography sub-layer only · ASML 100% of EUV systems · 2025 · tier 1
AI accelerator memory · high criticality
measured · HBM revenue shares 57/22/21 · Q3 2025
Semiconductor foundries · extreme criticality
lower bound · foundry revenue, top-6 named · FY2025
Compute deployment · moderate criticality
lower bound · cloud infra revenue shares · Q4 2025
CoWoS, InFO, chiplet integration · high criticality
no tier-1/2 published shares · CoWoS capacity: 75K → ~130K wpm, ’25→’26
GPU & accelerator designers · high criticality
no tier-1/2 published shares · NVIDIA ~80–85% merchant accel. · tier-3 band
Electronic design automation · high criticality
no tier-1/2 published shares · Synopsys ~32 / Cadence ~30 / Siemens ~14 · 2024 · tier 3
Wafers, photoresists, specialty gases · high criticality
no tier-1/2 published shares · wafers: no public per-supplier breakdown (Omdia)
Frontier compute consumers · moderate criticality
not a supply market · labs buy compute and supply none
of 9 layers have a Herfindahl: 2 measured outright, 2 as a lower bound whose unattributed remainder can only push the figure higher.
have none. 4 have a supply market that no tier-1 or tier-2 source breaks into shares; 1 is a consumer layer, which buys compute and supplies none.
the highest measured Herfindahl in the figure above: Equipment, EUV lithography sub-layer. Ten thousand is the arithmetic ceiling, one supplier holding the whole market.
The measurement gap. One row per layer of the compute supply chain, ordered by what is known about each (measured figures first, then lower bounds, then the layers carrying no figure at all), on a Herfindahl–Hirschman axis running from zero to 10,000.
A solid bar is a figure recomputed in code from published market shares; a bar with an open edge is a lower bound, drawn short of a true value that lies somewhere to its right because part of the market is unattributed; a dotted span is a real supply market for which no tier-1 or tier-2 source publishes shares, so no Herfindahl can be computed; a bracketed span is a consumer layer that buys compute and supplies none.
The criticality beside each layer name is an editorial judgement of supply-chain importance; it is never plotted on the Herfindahl axis. Every figure that is plotted is a single-layer market-share Herfindahl. The stress index reports a portfolio-weighted Herfindahl across 93 CSET steps: a second construct, over a second population.
The hourglass below shows who occupies each step of AI chip production, from the machines that pattern the silicon to the packaging that assembles the finished part.
43 companies across 9 layers. Each node traces what depends on it downstream on hover, from the full dependency graph. Block width is the published share on HBM and Fabrication, whose full charted sets have one, and an editorial judgement of supply-chain importance everywhere else; no width is squared into any figure.
moderate criticality (editorial)
labs buy compute and supply none
context: Epoch (T2 est.) · Menlo survey (T3)
No HHI: AI labs are frontier compute consumers, not suppliers in a defined product market. Context figures, each covering only the market named beside it and none of them a share in the HHI sense: US-enterprise LLM API spend Anthropic 40 / OpenAI 27 / Google 21 (Menlo Ventures survey n=495, Nov 2025, tier 3); OpenAI ~10–15% of world operational AI compute and top-five labs combined <50% at end-2025 (Epoch AI estimate, tier 2).
No HHI — AI labs consume compute; they are not suppliers in a defined product market. Epoch AI (May 2026 estimate): OpenAI ~10–15% of world operational AI compute; top-five labs combined <50%.
moderate criticality (editorial)
cloud infra revenue shares · Q4 2025
Synergy Research · tier 2 · ~37% unattributed
LOWER BOUND (top-3 only): 28² + 21² + 14² = 1,421 from Synergy Research Group Q4 2025 worldwide cloud infrastructure revenue shares (tier 2). The remaining ~37% (Oracle, CoreWeave, Alibaba, long tail) is unattributed — the true HHI is strictly higher.
Cloud infrastructure services: AWS 28% / Azure 21% / GCP 14% worldwide (Synergy Research Group, Q4 2025, tier 2). Top-3 HHI ≥ 1,421 — a lower bound: the remaining ~37% (Oracle, CoreWeave, Alibaba, long tail) is unattributed, so the true HHI is higher.
high criticality (editorial)
NVIDIA ~80–85% merchant accel. · tier-3 band
denominator-sensitive — see methodology
No HHI at tier-1/2. Share band: NVIDIA ~80–85% of merchant AI accelerators (tier-3 analyst band, 2025). Market-definition note: including captive ASICs (TPU, Trainium) moves NVIDIA toward ~75%; excluding ASICs entirely moves it above 90%. The band-squaring arithmetic appears in methodology text only, with caveats; the OECD "over 80%" line is a disclosed tier-3 citation chain (Farooque via OECD), not an OECD measurement.
Merchant AI accelerators: NVIDIA ~80–85% — a tier-3 analyst band, denominator-sensitive (captive ASICs in → ~75%; ASICs out → >90%). No tier-1/2 share publication exists, so no HHI is published for this layer.
Share band, tier 3 — no tier-1/2 publication. The denominator swings the number: including captive ASICs (Google TPU, AWS Trainium) moves NVIDIA toward ~75%; excluding ASICs entirely moves it above 90%. The OECD’s "over 80%" GPU line (Competition in Artificial Intelligence Infrastructure, Nov 2025, p.14 §2.3.1) is itself a tier-3 analyst estimate (Farooque, Yahoo Finance, ~85%) reported by the OECD — a citation chain, not an OECD measurement.
high criticality (editorial)
Synopsys ~32 / Cadence ~30 / Siemens ~14 · 2024 · tier 3
10-K revenue ÷ tier-3 market-size estimates
No HHI at tier-1/2. Shares on record, every one of them an estimate, vintage 2024: Synopsys ~32% (FY2024 10-K revenue $6.13B against tier-3 total-market estimates), Cadence ~30% (FY2024 10-K, 71% EDA segment), Siemens EDA ~14% (industry estimate, no named source). Three named firms sum to ~76%, and squaring them publishes no HHI: a figure derived from tier-3 estimates would assert precision its inputs cannot carry.
Three firms hold ~76% of EDA by revenue-derived estimates (tier 3), and export-control authority already reaches the layer: the FDPR extends US jurisdiction to chips designed with US EDA tools, and the regime register lists EDA software as US-restricted. No tier-1/2 share publication exists, so no HHI is published for this layer.
Revenue $6.13B (FY2024 10-K). Share is derived from revenue / total market estimate. Share is 10-K revenue against tier-3 total-market estimates — estimated, not published.
Revenue $4.64B (71% EDA segment, FY2024 10-K). Share is 10-K revenue against tier-3 total-market estimates — estimated, not published.
Siemens AG unit (ex-Mentor Graphics); share is an industry estimate with no named-source breakdown — the weakest-provenanced row in this layer, and the reason the layer publishes no HHI.
high criticality (editorial)
CoWoS capacity: 75K → ~130K wpm, ’25→’26
TrendForce · tier 2 · capacity ≠ share
No tier-1/2 share publication for advanced packaging (CoWoS) — HHI sourced-unavailable. Capacity context (TrendForce, tier 2): ~75K wafers/month end-2025 → ~130K projected end-2026; TSMC dominant with ASE (SPIL) and Amkor as overflow. Capacity is not a market share.
CoWoS capacity (TrendForce, tier 2): TSMC ~75K wafers/month (end 2025) → ~130K projected (end 2026); ASE (SPIL) and Amkor absorb overflow. Capacity ≠ market share — no tier-1/2 share publication exists, so packaging HHI is sourced-unavailable.
~75K wafers/month (end 2025, verbatim at source); ~130K (end 2026) is a TrendForce-tracked projection reported via syndicated trade coverage. Source: TrendForce (Tier 2).
SPIL handles ~60–80K CoWoS wafers/year overflow. Source: TrendForce (Tier 2).
~180–190K CoWoS wafers/year from TSMC overflow. Source: TrendForce (Tier 2).
high criticality (editorial)
HBM revenue shares 57/22/21 · Q3 2025
Counterpoint Research · tier 2
Measured: 57² + 22² + 21² = 4,174 from Counterpoint Research Q3 2025 HBM revenue shares (SK Hynix / Samsung / Micron — all suppliers named; tier 2, via Semiecosystem). Samsung recovered 15% → 22% Q2→Q3 2025 on HBM3E qualification; refresh on Q4 publication.
HBM revenue shares: SK Hynix ~57%, Samsung ~22%, Micron ~21% (Q3 2025, Counterpoint Research, tier 2) — HHI 4,174 = 57² + 22² + 21². HBM4 mass production began Feb 2026.
HBM3E 12-Hi 36GB in volume. HBM4 paid samples to NVIDIA Q4 2025; mass production Feb 2026 at M16/M15X. Prices raised ~20% for 2026. Source: Counterpoint Research (Tier 2).
~22% HBM market share (Q3 2025, Counterpoint Research) — recovered from ~15% (Q2 2025) on HBM3E NVIDIA qualification. 50% HBM capacity surge in 2026, targeting ~250K wafers/month (from ~170K). HBM4 paid samples to NVIDIA Q4 2025.
~21% HBM market share (Q3 2025, Counterpoint Research). HBM4E roadmap targets late 2027.
extreme criticality (editorial)
foundry revenue, top-6 named · FY2025
TrendForce · tier 2 · adv-node co. HHI: none
Named-share lower bound: 69.9² + 7.2² + 5.32² + 4.35² + 3.87² + 2.6² ≈ 5,007, recomputed in code from TrendForce FY2025 foundry revenue shares (tier 2, via Taipei Times 2026-03-14). Advanced node: Taiwan ~92% of leading-edge pure-play foundry manufacturing (CSET/VerWey 2021, COUNTRY-level incl. UMC + PSMC); no company-level advanced-node HHI exists at tier-1/2.
Overall foundry revenue FY2025 (TrendForce, tier 2): TSMC 69.9 / Samsung 7.2 / SMIC 5.32 / UMC 4.35 / GF 3.87 / Hua Hong 2.6 — named-share HHI 5,007 (lower bound). Advanced node: Taiwan ~92% of leading-edge pure-play foundry manufacturing (CSET/VerWey 2021) — COUNTRY-level, includes UMC and PSMC; no tier-1/2 company-level share exists.
TrendForce FY2025: $122.54B foundry revenue, 69.9% share (up from 64.4% in 2024). Advanced node: Taiwan ~92% of leading-edge pure-play foundry manufacturing (CSET "No Permits, No Fabs", VerWey 2021) — a COUNTRY share including UMC and PSMC, NOT a TSMC company share.
BIS Entity List addition Dec 2020 (Federal Register 2020-28031); advanced-tool shipments under presumption of denial.
extreme criticality (editorial)
ASML 100% of EUV systems · 2025 · tier 1
ASML AR + 6-K · broader equip: unavailable
EUV sub-layer (tier 1): ASML 100% of EUV lithography systems → HHI 10,000, the ceiling of the scale. Source: ASML 2025 Annual Report, financial-performance section + Q4-2025 investor presentation (Form 6-K, Jan 28, 2026) (48 EUV systems recognized 2025, 4 EXE + 44 NXE; EUV system sales +39% to €11.6B). Broader wafer-fab equipment market: sourced-unavailable — no HHI published. The prior blended equipment figure is retired.
EUV lithography: 100% ASML — HHI 10,000, tier 1 (ASML 2025 Annual Report, financial-performance section + Q4-2025 investor presentation (Form 6-K, Jan 28, 2026)). Broader wafer-fab equipment market (deposition, etch, metrology): no sourced HHI.
Sole supplier of EUV lithography systems (tier 1). 2025: 48 EUV systems recognized in revenue (4 High-NA EXE + 44 NXE); EUV system sales +39% YoY to €11.6B; ~83% of ALL lithography systems by unit count. Source: ASML 2025 Annual Report, financial-performance section + Q4-2025 investor presentation (Form 6-K, Jan 28, 2026).
high criticality (editorial)
wafers: no public per-supplier breakdown (Omdia)
Japan ~56% of wafers · CSET 2024 · tier 2
Sourced-unavailable, split layer: no public source breaks silicon wafers down by supplier (Omdia holds a breakdown behind a subscription; ledger 2026-06-09), so no HHI by supplier can be computed for wafers, and photoresist and specialty-chemical shares have no tier-1/2 publication either. Country-level context: Japan ~56% of global silicon wafers (CSET TechInsights 2024, tier 2).
Silicon wafers: Japan ~56% COUNTRY-level (CSET TechInsights 2024, tier 2). Per-supplier shares: sourced-unavailable — Omdia's breakdown sits behind a subscription and appears in no public source, so no HHI by supplier can be computed for wafers. Photoresist/specialty chemicals: sourced-unavailable.
CSET TechInsights 2024: Japan = 56% of global silicon wafers (country-level). Per-supplier share: sourced-unavailable — Omdia's breakdown appears in no public source (ledger 2026-06-09).
Acquired by JIP (Japan Industrial Partners) 2023.
Measured shares: Counterpoint Q3 2025 (HBM) · TrendForce FY2025 (foundry, named top-6) · Synergy Q4 2025 (cloud top-3, lower bound) · ASML 2025 AR (EUV, tier 1) · CSET/VerWey 2021 (Taiwan country-level, advanced node). Block widths: measured shares on HBM + Fabrication; editorial importance weights elsewhere, and neither is squared into any displayed figure. Link graph: CSET TechInsights 2024, licensed supply-chain database, SEC EDGAR, BIS Entity List.
ASML is the sole manufacturer of EUV lithography machines, so the EUV sub-layer scores 10,000, the ceiling of the Herfindahl scale.
The right-hand column carries two judgments per layer: an editorial chokepoint-criticality level, which sorts and colors the rows and is computed from no data (tier 4), and the measured concentration state, with the shares it was recomputed from, their source, their vintage and its authority tier. EDA Tools and Chip Design have tier-3 share figures only, which is why both read sourced-unavailable. A measured Herfindahl above 2,500 clears the “highly concentrated” line of the 2010 DOJ/FTC Horizontal Merger Guidelines, which the 2023 Guidelines lower to 1,800; the comparison applies to measured figures only.Each layer carries two judgments. The colored level is our editorial assessment of how critical that supply-chain step is; no data computes it. The HHI, where one appears, is a market-concentration figure built from published market shares, naming its source, year and authority tier inline. A measured HHI above 2,500 counts as “highly concentrated” under the 2010 US merger guidelines (the 2023 guidelines use 1,800). Three measured layers clear that bar: HBM memory, chip fabrication, and EUV lithography equipment, where one company, ASML, is the only supplier on Earth.
Colossus 1 sits outside the cloud layer. Anthropic holds a 100% lease on the whole facility (SpaceX-owned post-merger; Memphis, >300 MW, >220K NVIDIA GPUs), announced 2026-05-06. It is single-tenant, so it falls outside the Cloud & Infrastructure layer, which counts cloud providers. Sources: xAI announcement (Tier 1), DataCenterDynamics 2026-05-06 (Tier 2).Anthropic has exclusive access to the whole Colossus 1 facility in Memphis (over 220,000 NVIDIA GPUs, >300 MW), owned by SpaceX since its 2026 merger with xAI and leased 100% to Anthropic from 2026-05-06. It is a single-tenant lease, so it falls outside the cloud-provider concentration layer, which counts shared cloud services. Sources: xAI and DataCenterDynamics.
Methodology
(1) Measured concentration. Where a tier-1/2 source publishes market shares, the HHI is recomputed in code from those shares: HBM 4,174 (Counterpoint Q3 2025: 57² + 22² + 21²); Fabrication 5,007 as a named-share lower bound (TrendForce FY2025: 69.9² + 7.2² + 5.32² + 4.35² + 3.87² + 2.6²); EUV sub-layer 10,000 (ASML 2025 Annual Report + Q4-2025 6-K, tier 1; ASML is the sole EUV supplier: 48 systems recognized 2025, sales +39% to €11.6B); Cloud ≥ 1,421 as a top-3 lower bound (Synergy Q4 2025: 28² + 21² + 14²; ~37% of the market is unattributed, so the true value is strictly higher). Where no such source exists the display says sourced-unavailable: Advanced Packaging (TrendForce publishes a capacity trajectory; the circulating “40% revenue share” is paywalled and unverifiable), Materials (Omdia’s per-supplier wafer breakdown appears in no public source), and the broader Equipment market beyond EUV.
Chip Design publishes a share band only: NVIDIA ~80–85% of merchant AI accelerators (tier-3 analyst band). Squaring a mid-band share (~82%) would alone contribute ≈ 6,700 HHI points, far above any concentration threshold, but that arithmetic inherits the band’s tier-3 status and its denominator sensitivity (counting captive ASICs like TPU and Trainium moves NVIDIA toward ~75%; excluding ASICs entirely moves it above 90%), so no HHI is published for this layer. The OECD’s “over 80%” GPU line (Competition in AI Infrastructure, Nov 2025, p.14 §2.3.1) is itself a tier-3 analyst estimate (Farooque, Yahoo Finance, ~85%) reported by the OECD, which makes it a citation chain running through the OECD. AI Labs have no HHI: labs buy compute and supply no market (context: Menlo Nov 2025 enterprise API-spend survey, n=495, tier 3; Epoch AI compute-share estimates, tier 2). Advanced node: Taiwan ~92% of leading-edge pure-play foundry manufacturing (CSET “No Permits, No Fabs”, VerWey 2021) is a country-level share including UMC and PSMC; no tier-1/2 company-level advanced-node share exists.
(2) Editorial importance weights. The weights that size entity blocks outside HBM and Fabrication are tier-4 assessments of supply-chain importance, informed by CSET TechInsights 2024, a licensed supply-chain database and SEC EDGAR. They order the visual and enter no displayed or exported number.
| Layer | Criticality (editorial) — Editorial chokepoint-criticality assessment (moderate / high / extreme): assessed supply-chain importance, authority tier 4. The measured concentration state is in the next column. | Measured concentration — Herfindahl–Hirschman Index recomputed in code from published market shares (range 0–10,000). "≥" marks lower bounds; "sourced-unavailable" means no tier-1/2 share publication exists; AI labs have no figure because they buy compute and supply none. Never computed from editorial weights. | Basis (metric · source · vintage · tier) | Top nodes (editorial importance order) — Ordered by block-sizing weight: the measured share where the layer’s full charted set is published (HBM, Fabrication), otherwise editorial supply-chain-importance weight. Ordering only; weights are never rendered as market-share percentages. |
|---|---|---|---|---|
| AI Labs · Frontier compute consumers | moderate | NO HHI — consumer layer | No HHI: AI labs are frontier compute consumers, not suppliers in a defined product market. Context figures, each covering only the market named beside it and none of them a share in the HHI sense: US-enterprise LLM API spend Anthropic 40 / OpenAI 27 / Google 21 (Menlo Ventures survey n=495, Nov 2025, tier 3); OpenAI ~10–15% of world operational AI compute and top-five labs combined <50% at end-2025 (Epoch AI estimate, tier 2). | OpenAI · US; Google DeepMind · GB; Anthropic · US |
| Cloud & Infrastructure · Compute deployment | moderate | HHI ≥ 1,421 — top-3 lower bound | LOWER BOUND (top-3 only): 28² + 21² + 14² = 1,421 from Synergy Research Group Q4 2025 worldwide cloud infrastructure revenue shares (tier 2). The remaining ~37% (Oracle, CoreWeave, Alibaba, long tail) is unattributed — the true HHI is strictly higher. | Amazon Web Services · US; Microsoft Azure · US; Google Cloud · US |
| Chip Design · GPU & accelerator designers | high | HHI — sourced-unavailable (tier-1/2) | No HHI at tier-1/2. Share band: NVIDIA ~80–85% of merchant AI accelerators (tier-3 analyst band, 2025). Market-definition note: including captive ASICs (TPU, Trainium) moves NVIDIA toward ~75%; excluding ASICs entirely moves it above 90%. The band-squaring arithmetic appears in methodology text only, with caveats; the OECD "over 80%" line is a disclosed tier-3 citation chain (Farooque via OECD), not an OECD measurement. | NVIDIA Corporation · US; Advanced Micro Devices · US; Broadcom Inc · US |
| EDA Tools · Electronic design automation | high | HHI — sourced-unavailable (tier-1/2) | No HHI at tier-1/2. Shares on record, every one of them an estimate, vintage 2024: Synopsys ~32% (FY2024 10-K revenue $6.13B against tier-3 total-market estimates), Cadence ~30% (FY2024 10-K, 71% EDA segment), Siemens EDA ~14% (industry estimate, no named source). Three named firms sum to ~76%, and squaring them publishes no HHI: a figure derived from tier-3 estimates would assert precision its inputs cannot carry. | Synopsys · US; Cadence Design Systems · US; Siemens EDA · DE |
| Advanced Packaging · CoWoS, InFO, chiplet integration | high | HHI — sourced-unavailable | No tier-1/2 share publication for advanced packaging (CoWoS) — HHI sourced-unavailable. Capacity context (TrendForce, tier 2): ~75K wafers/month end-2025 → ~130K projected end-2026; TSMC dominant with ASE (SPIL) and Amkor as overflow. Capacity is not a market share. | TSMC (CoWoS) · TW; ASE Technology · TW; Amkor Technology · US |
| HBM Memory · AI accelerator memory | high | HHI 4,174 — measured | Measured: 57² + 22² + 21² = 4,174 from Counterpoint Research Q3 2025 HBM revenue shares (SK Hynix / Samsung / Micron — all suppliers named; tier 2, via Semiecosystem). Samsung recovered 15% → 22% Q2→Q3 2025 on HBM3E qualification; refresh on Q4 publication. | SK Hynix · KR; Samsung · KR; Micron Technology · US |
| Fabrication · Semiconductor foundries | extreme | HHI 5,007 — named-share lower bound | Named-share lower bound: 69.9² + 7.2² + 5.32² + 4.35² + 3.87² + 2.6² ≈ 5,007, recomputed in code from TrendForce FY2025 foundry revenue shares (tier 2, via Taipei Times 2026-03-14). Advanced node: Taiwan ~92% of leading-edge pure-play foundry manufacturing (CSET/VerWey 2021, COUNTRY-level incl. UMC + PSMC); no company-level advanced-node HHI exists at tier-1/2. | TSMC · TW; Samsung Foundry · KR; SMIC · CN |
| Equipment · Lithography, deposition, etch | extreme | HHI 10,000 — EUV sub-layer only | EUV sub-layer (tier 1): ASML 100% of EUV lithography systems → HHI 10,000, the ceiling of the scale. Source: ASML 2025 Annual Report, financial-performance section + Q4-2025 investor presentation (Form 6-K, Jan 28, 2026) (48 EUV systems recognized 2025, 4 EXE + 44 NXE; EUV system sales +39% to €11.6B). Broader wafer-fab equipment market: sourced-unavailable — no HHI published. The prior blended equipment figure is retired. | ASML Holding · NL; Applied Materials · US; Lam Research · US |
| Materials & Chemicals · Wafers, photoresists, specialty gases | high | HHI — sourced-unavailable | Sourced-unavailable, split layer: no public source breaks silicon wafers down by supplier (Omdia holds a breakdown behind a subscription; ledger 2026-06-09), so no HHI by supplier can be computed for wafers, and photoresist and specialty-chemical shares have no tier-1/2 publication either. Country-level context: Japan ~56% of global silicon wafers (CSET TechInsights 2024, tier 2). | Shin-Etsu Chemical · JP; SUMCO Corporation · JP; JSR Corporation · JP |
A Herfindahl–Hirschman index is the sum of every supplier’s squared percentage share of one market. Squaring is what makes it useful: it weights a seventy-per-cent supplier far more heavily than seven ten-per-cent ones, so the index separates a market with one dominant firm from a market of several mid-sized ones.
Where only part of a market is attributed, squaring the named shares alone yields a floor: the unattributed remainder can only add to the sum. Those rows are labelled lower bounds and drawn with an open edge. The editorial criticality levels beside the layer names are assessments of supply-chain importance at authority tier 4.