The Research · Dossier 02 · Updated June 2026
The machines coming for your job run on silicon. That silicon has a supply chain you can fit in one country. Every AI accelerator on Earth is designed in a handful of offices, fabbed on a handful of lines, packaged in a handful of plants, and gated by a single Dutch machine. Trace it stage by stage — and find the points where the whole thing stops.
The Pipeline
A GPU is born fabless, etched in Taiwan, stacked onto memory, and shipped to a data center. Tap a stage to isolate it below. The red chips are single points of failure.
Single Points of Failure
The replacement infrastructure looks vast. It is not resilient. Five constraints gate the entire chain — and most of them sit on one island, ~100 miles from a military adversary.
Who Controls What
Every player in the chain, by stage. Tap a card to open the file. Filter by where they sit in the pipeline.
Filter by stage
Where The Silicon Lives
Advanced-node capacity by location. Red = operational. Iron = under construction. Bone = announced. The leading edge — 3nm and 2nm — does not leave Taiwan.
Cascading Failure
The chain has no slack. Tap a scenario to trace the blast radius. There is no second supplier waiting in the wings — that is the whole point.