Scrutica
Supply Chain
Power capacity, hardware inventory, upstream suppliers, and downstream customers for the row owned by TSMC; any BIS Entity List exposure is walked through the canonical org graph rather than asserted at the facility row.
Capacity, ownership chain, and supply-chain position for TSMC Arizona Fab 21 (Phase 3), including any export-control exposure that surfaces when the BIS Entity List touches a node above the facility in the chain.
As of 2026-06-12
Operator (or owner of record) per the Scrutica facility row.
TSMC, the documented owner of TSMC Arizona Fab 21 (Phase 3), sits as a seed node in 6 pre-built cascade scenarios. Severity is the scenario's seed-impact value (per-node override where one is defined); decay reflects the curator's substitutability assessment for the disrupted role. Run any scenario to walk the propagation through the full supply-chain graph.
TSMC is deeply embedded in advanced-node logic, but some customers maintain Samsung Foundry and Intel Foundry relationships at lower nodes — a non-trivial fraction of impact is absorbed at the first hop.
Description
TSMC Arizona Phase 3: N2 (2nm) and A16 (1.6nm angstrom-class). Broke ground April 2025. Production of 2nm and A16 chips targeted by 2029.
Timeline
Announced
Apr 8, 2024
Construction Started
Apr 1, 2025
Expected Operational
Jan 1, 2029
Notes
Groundbreaking April 2025. N2/A16 process nodes. TrendForce (Sep 2025) reported timeline pulled ahead by one year (tier 3). Wikipedia and NIST CHIPS pages corroborate. Updated 2026-03-31.
Nearby Facilities (within 100 km)
Energy Profile
Grid Carbon
Carbon Intensity
Balancing Authority
NERC Region
eGRID Subregion
Chain terminates at a publicly-listed entity (TSMC). Subsequent ownership is observable via SEC filings or the equivalent national registry.
Taiwan (TW)
Not a Wassenaar participant (UN status); Strategic High-Tech Commodities (SHTC) controls administered by Bureau of Foreign Trade under the Foreign Trade Act.
Regimes: Taiwan SIPA-administered controls
Established via: TSMC
No flags found across the chain (BIS Entity List, designation-cascade closures, or compute-threshold reporting obligations) as of 2026-06-12. Cascade closures evaluated: 0.
No ownership-change events documented in the available data for any entity in this chain over the last 5 years. Events draw from a licensed corporate-ownership database’s deal records (M&A, buyout, divestiture, asset sale, JV formation); coverage is strongest for private US/EU transactions and partial for state-owned-enterprise restructurings or non-Western private deals. Subsequent ingestion runs may add events here.
No documented sovereign or state-linked LP investment touches any entity in this chain. The licensed fund/LP database’s LP-disclosure rate is partial; absence of evidence here is not evidence of absence.
The walk for TSMC Arizona Fab 21 (Phase 3)starts at the operator org id (or owner org id when no operator is recorded) and climbs upward through the unified affiliate graph, normalized to trace child-to-parent relationships. At each hop, the highest-authority parent edge wins (priority ordering: SEC EDGAR / 10-K / LEI registry > licensed corporate-ownership database > licensed supply-chain database > BIS / Federal Register > licensed affiliate graph). Confidence and recency break ties.
The walk terminates on (a) a node with a public-filing-grade ancestor edge — a SEC filer, an LEI-registered listed entity, or a 10-K-disclosed parent — (b) a government entity, (c) a 10-hop limit, or (d) a detected cycle. When termination is not at a public filer or government, the panel labels the chain's end-state as opaque beyond that point. Opacity is reported, never implied.
Per-source confidence (licensed corporate-ownership database): HIGH for Parent / Subsidiary edges; MEDIUM for JV and AFFILIATE edges. Known limitations: (1) shell companies with undisclosed principals terminate the public-record walk; (2) recent transactions reach the panel after the database's weekly refresh; (3) cross-jurisdiction private holding structures may carry intermediate hops that are not in any ingested source. Each row carries an edge-level as of date so freshness is auditable per hop.
Two further caveats are inherent to how the walker and the underlying data compose. The walker selects a single highest-authority parent per hop, so JV or multi-parent ownership structures collapse to one displayed path; alternate parents exist in the ingested sources but are not surfaced on this card. And some licensed-database edges encode parent-of relations against entity-resolved stub IDs whose display name reflects a sub-entity (a business unit, an acquired-then-renamed line) rather than the canonical parent — when a chain reads structurally surprising, the per-row source citation is the authoritative reference, not the chain's narrative shape.
Computed at: 2026-06-12. Walk depth observed: 1 hop. Sources combined: Scrutica facility record + licensed corporate-ownership database + SEC CIK cross-reference + LEI registry + (where applicable) licensed supply-chain database + Federal Register designation closures. The licensed databases are held under subscription and not redistributed; per-hop labels name no vendor.
Substitutability is unchanged from the blanket TSMC disruption (Samsung Foundry / Intel Foundry absorption capacity is identical regardless of who is restricted); the policy-vs-blanket distinction is carried by edgeFilter, not by per-hop decay.
Structural shortfall with no alternative CoWoS-L packaging at volume; OSATs cannot replicate Blackwell-class packaging at scale, and NVIDIA holds >60% of TSMC CoWoS allocation, so non-NVIDIA AI GPU production absorbs the constraint with limited dampening.
Compound packaging + EDA disruption hits chiplet-architecture chips (AMD MI300X, NVIDIA Blackwell) disproportionately; some single-die monolithic designs and trailing-node products are insulated, providing modest per-hop attenuation relative to the blanket scenarios.
US strategic helium reserve and alternative shipping routes (Australian, Algerian, US domestic LNG-byproduct streams) provide some buffer; helium boil-off limits stockpiling beyond ~45 days, so the buffer is real but bounded.
Multi-source packaging via OSAT partners (Amkor, SPIL handling 240–270K wafers/year overflow) now dampens downstream propagation; CoWoS capacity tripled to ~75K wafers/month by end-2025 per TrendForce.
TSMC Fab 18
Tainan, TW