The Machine Behind AI
Mapping the supply chain from raw quartz to running GPU clusters
The world's largest technology companies plan to spend $725 billion building AI infrastructure in 2026. This report traces where the money goes and where the supply chain breaks.
Five findings
The real fragility is in companies nobody watches
Carl Zeiss SMT and Trumpf (both private, both German) scored highest risk in our analysis. Higher than TSMC or NVIDIA. They make components inside ASML's EUV machines that nobody else can make. Three points of failure within 200 km of each other, controlling all leading-edge chip manufacturing on Earth.
Apparent redundancy is illusory
Three companies make HBM memory. Sounds diversified. But all three depend on the same TSMC packaging, the same Ajinomoto substrate film, the same ASML lithography. A disruption at the packaging layer disables all three simultaneously. The surface suggests resilience. The deep structure reveals fragility.
Bottlenecks migrate over time
In 2025-2026 the binding constraint is HBM memory and CoWoS packaging. By 2027-2028 those ease. But transformer lead times (128-144 weeks) and grid interconnection queues (4-10 years) persist. The constraint migrates from semiconductor scarcity to power scarcity.
The capex cycle is self-reinforcing until it hits a wall
Cheaper AI compute opens new use cases, which drive more demand, which justifies more spending. This cycle breaks at one of three walls: financing ($1.5T projected debt), permitting (200+ opposition groups), or revenue (if AI fails to deliver returns justifying $725B/year).
Edge inference amplifies the same chokepoints
Smartphones, autonomous vehicles, and defense systems all require chips from the same TSMC fabs, the same ASML lithography, the same materials. Edge AI does not have its own supply chain. It shares this one. 40% of advanced node demand comes from non-data-center applications.
The supply chain in four pillars
Three enabling supply chains converge at the data center
Semiconductors
Sand becomes a GPU
Networking
Linking thousands of GPUs
Power & Cooling
The hardest wall
Data Centers
Where the three supply chains converge