From Minerals to Models
The AI Supply Chain, Mapped
Every AI system rests on a physical stack — raw minerals dug from the ground, chips etched atom by atom, data centers drawing power like small cities, and the data that feeds it all. Here are the five layers, and the 117 public companies that build them.
The Five Layers · tap to explore
L0
Raw Materials & Minerals The ground floor
The physical foundation of all AI hardware: silicon, copper, lithium, cobalt, rare earths, gallium, germanium. Mined and refined worldwide, heavily concentrated in China. Without these, nothing above exists.
20 companiesCopper · Lithium · Rare earths
→
L1
Chip Manufacturing The hardest step
Turning raw wafers into finished semiconductors. Foundries like TSMC, ASML's EUV lithography monopoly, fab tooling, inspection, photomasks and ultra-pure chemicals. The most capital-intensive layer in the entire economy.
21 companiesFoundries · EUV · Fab equipment
→
L2
Chips & Hardware The engine
Designed silicon and assembled systems for AI compute: GPUs, custom ASICs, networking silicon, HBM memory, optical interconnects, power ICs and complete AI servers. Dominated by Nvidia, Broadcom and AMD. The biggest market by dollars.
30 companiesGPUs · ASICs · HBM · Networking
→
★ In focus
L3
Data Centers & Energy
The physical infrastructure that houses and powers AI: data-center REITs, power distribution and cooling, nuclear and gas generation, grid electrification, construction contractors and fiber. The fastest-growing investment layer of 2025–26 — and the focus of this watchlist, with full company profiles.
29 companiesFull profiles: HQ · CEO · revenue
→
L4
Data The fuel
The information layer that feeds and is produced by AI: training corpora, vector and relational databases, pipelines, annotation, cloud storage and the hyperscaler platforms (AWS, Azure, GCP, CoreWeave) where it all runs.
18 companiesHyperscalers · Platforms · Storage
→