Scrutica
Scenario-based exposure modeling mapped against verified facility data and supply chain topology. Each assessment combines scenario characterization with infrastructure exposure quantification; dependency propagation is modeled through the cascade simulator. Exposure scoring uses published, open-source data only (CSIS, IISS, CRS, JINSA). Assessments are analytical tools, not intelligence products.
How do geopolitical and regulatory scenarios reshape access to AI infrastructure? Each assessment maps a specific scenario against real facility data to identify which countries and companies face the highest exposure, and traces downstream consequences through the supply chain graph. All exposure scoring draws on published, open-source data (CSIS, IISS, CRS, JINSA); these are analytical tools, not intelligence products.
Which Gulf data centers are within range, and how defended are they?
Exposure
14 missile systems (CSIS)
Region
Gulf states, Israel, Turkey
Key Facilities
AWS, Azure, Google Cloud, Stargate UAE
Threat scoring from published CSIS specs + empirical CEP from Oct 2024 satellite imagery. IRGC published a 29-target tech company list (RUSI). AWS DCs struck March 2026.
Which facilities can train models exceeding the 10^25 FLOP regulatory trigger?
Exposure
Regulatory jurisdiction: EU, US, China thresholds
Region
EU, US, UAE, China
Key Facilities
Hyperscale AI training clusters
Multi-regime threshold analysis (EU 10²⁵, US 10²⁶, China 10²⁴). 15 frontier models plotted against thresholds. Facility-level FLOP estimation with jurisdiction exposure mapping.
What happens to global AI compute if TSMC goes offline?
Exposure
Blockade, targeted strike, power grid, export restriction
Region
Taiwan, global downstream
Key Facilities
TSMC fabs (~92% advanced node), Samsung, Intel, SMIC alternatives
4 disruption scenarios with distinct propagation signatures across 17,900+ real FactSet supply chain edges. Substitution analysis quantifies the 85-95% capacity gap at alternative foundries. HHI: 8,500 for advanced logic.
What co-located DC capacity goes dark if Saudi grid infrastructure is hit?
Exposure
Drone/cruise missile strike on grid infrastructure
Region
Saudi Arabia (Eastern Province)
Key Facilities
Google Dammam, Aramco Dhahran, AWS Riyadh, STC, NEOM
2019 Abqaiq attack: 0% intercept rate from undefended azimuth. Google Dammam shares Eastern Province grid with Abqaiq, Ras Tanura, Jubail petroleum facilities. Grid co-location is the threat model, not direct strike.
Seismic risk to DRAM/NAND/HBM concentration in Japan and South Korea.
Exposure
Nankai Trough megathrust (M8.0-9.1)
Region
Japan, South Korea
Key Facilities
SK Hynix Icheon, Samsung Pyeongtaek, Micron Hiroshima, Kioxia Yokkaichi
HBM market (Q3 2025): SK Hynix ~56%, Samsung ~25%, Micron ~19%. A seismic event at any of three locations disrupts the entire AI GPU memory supply chain. Historical price spikes from 2011 Tohoku and 2016 Kumamoto quantify the magnitude.
Submarine cable chokepoints connecting Asia-Europe compute.
Exposure
Cable severing (anchor drag, sabotage, earthquake)
Region
SCS, Red Sea, Strait of Malacca, Persian Gulf
Key Facilities
Fujairah cable hub, Singapore, Hong Kong, Gulf DC clusters
Cable-dependency-to-data-center mapping based on TeleGeography route data. 8+ major cable systems, 4 historical incidents (Tonga 2022, Red Sea 2024, Taiwan 2006, Egypt 2008). Rerouting latency penalty modeling through alternative paths.