Capital expenditure is money spent on building durable “things” rather than day-to-day operations. In the context of cloud infrastructure, capex is the money that a cloud provider has to pay up-front to build data centers and buy servers (this stuff is collectively called capital assets).
The goal is to rent out those capital assets out over their useful life so that the money generated by those capital expenditures exceeds the capex spend that went into buying them up front.
Capex for AI
In third quarter CY2024, the following hyperscalers spent the following in capex:1
- Amazon: $22.7 billion
- Microsoft: $14.9 billion
- Google: $13.9 billion
- Meta: $8.3 billion
In addition, only half of Microsoft’s capex spend was on customer demand; the remaining was on long-lead buildout of data centers to meet anticipated future demand.1