Amazon does not break out capex by segment with the granularity that would make attribution easy. But the investment is flowing to three primary destinations. AWS infrastructure, both for general cloud capacity and AI-specific compute, is the largest piece. Data center construction, chip procurement, and networking equipment for the Trainium and Inferentia AI chip programs are all included here.
Logistics and fulfillment infrastructure accounts for a significant second tranche. Amazon's delivery network, which now competes directly with UPS and FedEx in terms of volume, requires ongoing investment in sortation centers, last-mile delivery infrastructure, and automation. The long-term goal is to reduce per-unit delivery costs to a level where the logistics operation itself becomes a profit center rather than a cost of the retail business.
Kuiper represents a smaller but growing capital commitment. Satellite internet is a multi-year project with significant launch and ground infrastructure costs that will not generate material revenue until the constellation reaches operational scale.
The critical analytical question is whether the AWS portion of this capex generates returns comparable to the returns AWS has historically produced. AWS operating margins are among the highest in any technology segment. If the AI compute buildout fills with high-margin AI services, the return on investment will be excellent. If AI workloads commoditize or if hyperscaler competition intensifies, the return profile deteriorates.
Q4 2024 and Q3 2024 earnings surprises were exceptional: 26.6% and 26.3% above consensus. Execution has been strong when it can be measured.