Three Layers of the AI Buildout
NVIDIA revenue, five-company hyperscaler capex, and U.S. data-center construction spending · 2022-Q4 = 100. The supplier layer moved first and fastest. It distinguishes supplier revenue, buyer investment, and physical construction as separate stages of the AI capital cycle.
What does it show?
The supplier layer moved first and fastest. It distinguishes supplier revenue, buyer investment, and physical construction as separate stages of the AI capital cycle.
Methodology
Sum quarterly capex reported by Amazon, Microsoft, Alphabet, Meta, and Oracle; compare it with NVIDIA quarterly revenue and the quarterly average of the Census monthly seasonally adjusted annual rate for private data-center construction. Rebase each series to 2022-Q4 = 100. The three measures cover different scopes and fiscal calendars: NVIDIA revenue is not entirely data-center revenue, capex is not entirely AI spending, and construction excludes equipment. Quarter labels are accounting-period labels rather than perfectly aligned calendar windows. This shows scale and sequencing, not a supply-chain flow identity. What this isn't: The three measures cover different scopes and fiscal calendars; this shows scale and sequencing, not a supply-chain flow identity.