First-order effects

• Nvidia, SK Hynix, TSMC and ASML gain disproportionate influence over AI-system availability because each controls a critical layer of the supply chain, from accelerator design through advanced manufacturing, memory and lithography.
• AI-chip customers and system builders face a supplier stack with few practical substitutes, concentrating procurement, capacity-planning and technical dependency among these four companies.

Second-order effects

• Competing chip designers must do more than match Nvidia’s computing performance: they also need access to the manufacturing and memory ecosystem controlled by key suppliers, raising the difficulty of scaling credible alternatives.
• Capacity competition can spill beyond AI buyers; later coverage that Apple had to compete for TSMC capacity illustrates how AI demand can tighten access for other advanced-chip customers.

Third-order effects

• If this concentration persists, AI infrastructure economics will be set increasingly by bottlenecks across the full hardware stack rather than by the GPU market alone, strengthening the strategic leverage of specialized upstream suppliers.
• The structure leaves the ecosystem vulnerable to disruption at any one layer, creating a durable incentive for customers and rival designers to pursue alternative compute architectures and supply paths, though their ability to reduce dependence remains unproven.