Scenarios
Five stress cases over the AI-compute supply chain.
Five stress cases over the AI-compute supply chain.
Disruption to Japanese wafer and equipment suppliers could affect semiconductor production beyond the damaged sites. These cases select locations for assumed disruption and compare the relevant suppliers’ market shares. The Korean finding examines how shortages of Japanese inputs could reach memory production as inventories deplete; that propagation is reasoned by hand because the graph lacks the input-dependency path to SK Hynix.
The active M9 case selects 4 locations for assumed disruption. The supplier shares below show how concentrated the relevant markets are; they do not give the share of output produced at each site. Historical company reports can describe recovery in more detail: Shin-Etsu recorded partial operations at Shirakawa from 20 April 2011 and reported pre-earthquake capacity restored on 1 July.
Supplier shares and HHI for the selected markets, followed by the locations assigned to the active case.
Authored cases that assume simultaneous disruption in Japan and Taiwan or South Korea.
Manufacturer reports of factory shutdowns and restarts, with a product-specific price observation from TrendForce.
Chosen magnitude, radius, severity and site list, with separate probability-source context.
Stored scenario settings, geographic assignments, market-share sources and the limits of the historical comparisons.
The engine runs the Japan-to-Korea coupling over the whole supplier–customer graph: it seeds the Japanese semiconductor-input cluster (Shin-Etsu, SUMCO, Tokyo Electron, Advantest, and the rest) and propagates the shock forward through recorded edges, reaching Samsung and the Korean supply chain. It does not reach SK Hynix by the input-dependency path this page’s fab-level model reasons about, because that dependency is not a recorded edge; nor does it hold the seismic-probability and HBM-qualification detail above.