Every AI query travels through a sequence of monopolies and oligopolies stretching from rare-earth mines to foundation models. This is how the most concentrated supply chain in modern history works.
What makes this supply chain extraordinary is that the bottlenecks are serially dependent. A single Nvidia AI GPU requires ASML's machines (monopoly) to pattern circuits at TSMC (near-monopoly) using SK Hynix's memory (oligopoly), built with materials China controls (near-monopoly), deployed in hyperscaler data centers (oligopoly) connected by cables those same hyperscalers own.
The concentration extends beyond hardware. Energy and water have become binding constraints, driving hyperscalers into power generation itself. Extreme capital requirements, learning curves, and ecosystem lock-in create barriers that compound over time. And for the Global South, complete dependence on foreign providers raises fundamental questions about digital sovereignty in an AI-driven world.
A disruption at any point — a natural disaster, an export restriction, a production delay — cascades through the entire AI economy. The US, China, and allied nations are investing hundreds of billions to build alternatives, but these chokepoints will persist for years. Whether this chain of concentration represents efficient allocation or a profound vulnerability is one of the defining strategic questions of our time.
Twelve layers, a handful of companies, and a geography that leaves most of the world dependent on decisions made elsewhere. The AI supply chain is not just concentrated — it is concentrated in sequence, with no alternative path from minerals to models.