Investing.com -- Amazon.com Inc (NASDAQ: AMZN) is fundamentally altering its hardware procurement playbook, agreeing to purchase billions of dollars in Nvidia Corp. graphics processors before its data centers are equipped with the power to run them. According to internal August emails obtained by Business Insider, the cloud giant established a new "Early Rack Receipt" program to take delivery of roughly $5 billion in Nvidia GB300 server racks at U.S. facilities ahead of final deployment, with expectations to accept an additional $3 billion in hardware each month for the foreseeable future.

The aggressive staging strategy highlights the extreme measures hyperscalers are taking to secure scarce artificial intelligence chips amid severe power grid constraints. Business Insider reported that finished equipment is currently being staged in intermediary warehouses because several Amazon Web Services sites lack the necessary utility capacity to energize the servers, leaving a portion of the hardware sitting idle as "dark GPUs."

This procurement shift carries substantial financial implications for the world's largest cloud provider, which previously took ownership and recorded expenses for server racks only upon arrival at operational data centers. By taking delivery earlier in the supply chain, Amazon is tying up vast sums of capital in equipment before it can generate cloud rental revenue, placing immediate strain on corporate liquidity.

The capital-intensive strategy comes as Amazon's broader AI infrastructure buildout pushes its financial resources to unprecedented levels. The company projects roughly $220 billion in capital expenditures this year, a figure that surpasses its Big Tech peers, contributing to a swing into negative free cash flow last quarter for the first time since 2023.

Despite the mounting cash flow pressures and tightening credit markets for data center financing, Amazon views early inventory absorption as a necessary trade-off to keep hardware out of competitors' hands. An Amazon spokesperson defended the staging process as a routine "best business practice" designed to enable rapid installation once power capacity comes online, while disputing internal indications that the company lacked a clear path to bring the facilities live.

Original Article