Amazon plans about $220 billion in capital spending in 2026. In a report published October 19, Business Insider described a timing risk at Amazon Web Services (AWS): the cloud business may take ownership of some server racks before sites have the power and installation capacity to put them to work. That can move cash outlays ahead of productive use and revenue, according to Business Insider’s report.

The $220 billion is a broader capital plan

The $220 billion is planned capital spending, not money already spent. An October 18 analysis hosted by Yahoo Finance and written by Insider Monkey put the plan up from an earlier estimate of roughly $200 billion, partly because memory costs rose. A February 5 discussion by Bloomberg Podcasts described the then-reported 2026 plan as covering satellites, logistics robotics and Whole Foods as well as AI computing, which it called the largest component. The overall figure therefore covers more than AI infrastructure. (Insider Monkey’s analysis; Bloomberg Podcasts)

How Early Rack Receipt shifts the cash timeline

Business Insider reported that AWS’s Early Rack Receipt process transfers ownership of some Nvidia GB300 server racks earlier in the supply chain than the previously described practice of paying when equipment reaches a data center. An internal email from August projected about $5 billion in rack deliveries that month, followed by roughly $3 billion per month for the foreseeable future. Those figures were projections in the email.

The financial wrinkle is the gap between owning equipment and using it. If AWS takes ownership while a site is still waiting for installation or power, the cash commitment comes before the racks can support services and revenue. The reports do not quantify how long that gap might last or what return the equipment will earn.

Power and construction shape deployment

People familiar with the effort told Business Insider that some AWS sites had space for more racks but not enough available power to turn them on; some equipment was staged in intermediary warehouses. The report also describes separate schedules for power permits, cooling, electrical equipment, construction and hardware. Around West Texas projects, skilled-worker shortages added pressure, along with costs for temporary worker housing and server shipping crates.

Amazon’s spokesperson told Business Insider that staging finished hardware near deployment sites is standard business practice, cited AWS’s long-established systems for tracking infrastructure assets, and disputed the broader suggestion that AWS lacks a forecasted installation path. The account concerns some sites and racks, not every AWS facility.

At a September staff meeting, infrastructure executive Prasad Kalyanaraman said AWS expected demand to exceed available capacity in 2027, according to Business Insider. That forecast sits alongside the deployment challenge: building capacity requires power, equipment and construction schedules to line up.

AWS growth does not settle the return question

AWS revenue grew 37% in the second quarter, its fastest rate in 18 quarters, and its backlog reached $496 billion, Business Insider reported. Those figures indicate demand for AWS services, but they do not show how quickly new racks will be installed or what returns this additional capacity will produce.

At the September meeting, Kalyanaraman described Amazon’s AI and machine-learning business as deeply negative in free cash flow and expected Amazon’s overall free cash flow to remain negative for the next couple of years or so. The Yahoo Finance-hosted Insider Monkey analysis also described management assumptions that data centers could remain useful for more than 30 years, while servers and networking equipment could break even in just under three years and remain useful for five to six years. It said most AI capacity was contracted for at least five years. Utilization, customer pricing, newer chips and rising memory and storage costs still affect the economics.

Bloomberg Television reported on July 7 that Amazon was seeking to raise at least $25 billion through a U.S.-dollar bond sale as it increased AI infrastructure investment. The financing question is familiar across money and markets; here, the operational hinge is when racks can begin supporting revenue.