Sovereign-AI program lead at national-government / research-council level.

Audience Profile

• Age / Experience: 15–30 years experience; senior career stage.
• Current role: Program Director / Senior Research Council / Policy Advisor (national lab / government R&D agency / sovereign fund).
• Top pain points: Export-control complexity (US BIS rules + EU / China parallels); Domestic / allied-nation AI chip supply assurance; IP-localisation requirements for sensitive workloads.
• Top decision blockers: Long-cycle procurement + multi-party approval chains; Sovereign capability gap vs market-leading benchmarks.

What This Segment Needs

• Information: Export-control exposure per vendor (BIS entity-list, China-charge history), foundry-jurisdiction map, and roadmap dates for allied-fabbed parts.
• Tools: A vendor-fit matrix scoring supply assurance, jurisdiction of fab/HBM, and contracted 2026 backlog visibility.
• Services: Multi-party procurement support, on-prem/air-gapped deployment, and IP-localisation / sovereign-cloud contractual guarantees.

Top 5 Companies for You (Fit Score)

| Rank | Company | Score | Why | |------|---------|-------|-----| | 1 | AMD | 76/100 | Record Q3 FY2025 ~$9.25B (+36% YoY); OpenAI 6GW (2025-10-06) + Oracle 50,000 MI450 (2025-10-14). But ~$800M MI308 China export charge (Q2 FY2025) = direct sovereign-supply risk. | | 2 | Cerebras Systems | 72/100 | $1.1B Series G at $8.1B (2025-09-30); wafer-scale build-out. Caution: revenue concentrated in G42-linked international capacity — assurance, not diversified. | | 3 | Tenstorrent | 71/100 | Samsung SF2 2nm Quasar (2025-07-22); Canadian HQ = low US export-control exposure. Zero disclosed customer wins; ~$693M Series D runway. | | 4 | Groq | 70/100 | $750M at $6.9B (2025-09-17); Helsinki + Dammam HUMAIN + Bell Canada multi-continent. Mega-deal concentration (HUMAIN ~$1.5B) and political-reversal risk. | | 5 | Rebellions | 69/100 | Series C ~$1.4B (2025-06-12); REBEL on Samsung 4nm + SK hynix HBM3E + Arm Total Design (2025-09-30). No customer/revenue signals disclosed. |

Deal-Breakers (Your Hard Preferences)

No hard preferences declared for this segment.

How to Evaluate Any Company in this Niche (Checklist)

• [ ] Check growth signals: contracted multi-gigawatt/2026 backlog with named counterparties (e.g. Oracle 50,000 units), not TAM slides.
• [ ] Check comp data: none supplied here — request public-sector pay-band benchmarks before relying on vendor claims.
• [ ] Check learning signals: silicon ladder depth (Principal/Staff RTL, P&R, UVM) and node (SF2 2nm, 4nm) versus software/ROCm gaps.
• [ ] Check stability signals: explicit export-control charges (AMD ~$800M MI308) and single-customer concentration (G42, HUMAIN).
• [ ] Check culture signals: ask for fab/HBM jurisdiction and whether an air-gapped / IP-localised SKU exists on roadmap with dates.
• [ ] Check supply assurance: confirm allied-nation foundry path (Samsung, TSMC-allied) and HQ jurisdiction (Canada lowers BIS reach).

Reverse-Hype Watch

• AMD: FAD 2025 '35%+ growth over 3-5 years' vs '$1T TAM by 2030' is aspirational, not booked; 5/5 reqs hardware while ROCm flagged as the software risk.
• Cerebras: 'tens of millions tokens/sec' rests on G42-linked capacity; diversification 'still early'.
• Groq: Bell 'up to 500 MW' target vs only ~7 MW live (~70x gap).
• Tenstorrent / Rebellions: scale-out and valuation claims with zero customer-win signals — investor-driven, not customer-backed.

Under-reported for sovereign buyers: export-control resilience and fab/HBM jurisdiction are rarely scored publicly. Vendors publicize valuation and token-throughput, but almost never disclose which jurisdiction fabs the part, whether an air-gapped SKU exists, or contractual IP-localisation terms — the dimensions that actually gate national procurement.