AI Chip DesignMet
Three independent training-class accelerator programmes — NVIDIA Blackwell Ultra, AMD Instinct, Trainium — under US IP control
▸ Qualification
NVIDIA Data Centre compute revenue reached US$43.0 billion (Q3 FY2026, up 56%) with Blackwell Ultra as leading architecture. Amazon deployed 1.4 million Trainium chips; combined custom silicon revenue exceeds US$10 billion annually. AMD ships the Instinct MI series. No other jurisdiction hosts more than one competitive programme.
ⓘ Sources
Primary Grade 2NVIDIA (SEC 10-Q Q3 FY2026) · 2025-Q3
↗Data Centre compute revenue $43.0 billion, up 56%. Blackwell Ultra leading architecture across all customer categories. Corroborating Grade 2Amazon (SEC 8-K FY2025 Q4) · 2025-Q4
↗Trainium fully subscribed with 1.4 million chips landed across all generations. Combined custom silicon annual revenue run rate over US$10 billion (Trainium, Graviton, Nitro). Advanced FabricationMet
Intel Fab 52 manufactures Intel 18A wafers and TSMC Arizona is in volume production, both operating on US territory
▸ Qualification
Intel's Fab 52 is fully operational producing Intel 18A wafers — the most advanced logic process manufactured in the United States at time of assessment. TSMC's Arizona facility entered volume production in late 2024. Both sites operate on US soil under territorial jurisdiction regardless of corporate parent.
ⓘ Sources
Primary Grade 2Intel (SEC 8-K Q3 FY2025) · 2025-Q3
↗Fab 52 fully operational, manufacturing Intel 18A wafers — most advanced logic wafers produced in the United States at time of assessment. Corroborating Grade 2TSMC · 2025-03
↗TSMC Arizona site in volume production since late 2024. Critical EquipmentMet
KLA holds 73.8% of the global metrology and inspection market, described as indispensable to advanced process control
▸ Qualification
KLA increased its metrology and inspection market share from 72.6% in 2024 to 73.8% in 2025. Applied Materials and Lam Research hold dominant positions in deposition and etch respectively. No advanced fabrication facility at any node operates without US-designed process control equipment.
ⓘ Sources
Primary Grade 3247 Wall St citing SEMI · 2026-05
↗KLA increased metrology and inspection share from 72.6% in 2024 to 73.8% in 2025. Corroborating Grade 2KLA 2026 Investor Day · 2026-03
↗CEO: KLA is indispensable to advanced process control regardless of fab operator. AI ComputeMet
xAI Colossus comprises 100,000 GPUs expanded to 555,000 at 2 GW; multiple US hyperscalers train frontier models on US territory
▸ Qualification
xAI's Colossus cluster operates 100,000 NVIDIA Hopper GPUs, built in 122 days, and expanded to 555,000 GPUs at 2 GW capacity. Anthropic, xAI, Google, and OpenAI all occupy the top tier on Arena Elo rankings. US territory hosts the largest concentration of frontier-class AI training infrastructure globally at time of assessment.
ⓘ Sources
Primary Grade 2NVIDIA Newsroom (Colossus/Spectrum-X) · 2026-01
↗xAI Colossus: 100,000 NVIDIA Hopper GPUs, built in 122 days. Expanded to 555,000 GPUs at 2 GW. Corroborating Grade 3Stanford HAI AI Index 2026 · 2026
↗Anthropic, xAI, Google, OpenAI all occupy the top tier on Arena Elo. Frontier ModelsMet
Anthropic, OpenAI, Google, and xAI all occupy the top benchmark tier; no other jurisdiction hosts four frontier labs
▸ Qualification
Stanford HAI's 2026 AI Index confirms Anthropic, xAI, Google, and OpenAI in the top Arena Elo tier. Frontier models gained 30 percentage points on Humanity's Last Exam in a single year. Multiple US-headquartered organisations hold frontier positions across reasoning, coding, and knowledge benchmarks.
ⓘ Sources
Primary Grade 3Stanford HAI AI Index 2026 · 2026
↗Frontier models gained 30 percentage points in a single year on Humanity's Last Exam. Corroborating Grade 3Artificial Analysis / LMArena · 2026-05
↗Multiple US-headquartered models (Anthropic, OpenAI, Google, xAI) at global frontier across reasoning, coding, and knowledge benchmarks. AI RegulationMet
Export controls enforced extraterritorially — Operation Gatekeeper prosecuted US$160 million GPU smuggling conspiracy
▸ Qualification
The Department of Justice prosecuted conspiracy to export at least US$160 million in controlled GPUs destined for China. Operation Gatekeeper seized over US$50 million in NVIDIA H100 and H200 GPUs and obtained guilty pleas, demonstrating active extraterritorial enforcement of AI-relevant export control instruments.
ⓘ Sources
Primary Grade 1US Department of Justice · 2025-11
↗Prosecution for conspiracy to violate export controls by illegally exporting advanced GPUs to the PRC. Corroborating Grade 1US Department of Justice (Operation Gatekeeper) · 2025-12
↗Conspiracy to export at least US$160 million in export-controlled NVIDIA H100 and H200 GPUs. Over US$50 million seized. Guilty pleas obtained. Sovereign InvestmentMet
CHIPS for America awarded over US$33 billion in fabrication incentives; Intel alone received US$5.7 billion
▸ Qualification
The CHIPS for America programme awarded over US$33 billion of the US$36 billion in proposed incentives, directed at semiconductor fabrication on US territory. Intel received US$5.695 billion in CHIPS incentives and issued equity instruments. The programme is the largest single sovereign investment in domestic semiconductor manufacturing at time of assessment.
ⓘ Sources
Primary Grade 1US Department of Commerce (CHIPS for America) · 2025-01
↗CHIPS for America awarded over $33 billion of the over $36 billion in proposed incentives funding. Corroborating Grade 2Intel (SEC 8-K) · 2025-08-27
↗Intel received $5.695 billion in CHIPS incentives; issued 274.6M shares of common stock and a warrant. Severance: All 7 dimensions sustainable under severance. Full-stack ecosystem anchor.