Loading page…
Internal emails describe billions of dollars in planned early deliveries. Amazon calls staging routine and disputes claims that it lacks a clear deployment path.
Listen to this story
Amazon is reportedly accepting Nvidia GB300 racks before some AWS sites have the utility power to run them, leaving completed systems in intermediate warehouses instead of earning cloud revenue. That shifts ownership and expense recognition ahead of deployment, tying up capital before capacity is usable. The reported receipts are plans, not installed compute; Amazon calls early staging routine and disputes that it lacks a clear path to bring the facilities online.
August internal emails describe an “Early Rack Receipt” program that lets Amazon accept racks before final deployment at U.S. facilities.
The reported plan included about $5 billion in racks and another $3 billion in hardware receipts each month for the foreseeable future.
Amazon’s roughly $220 billion 2026 capital-expenditure projection is companywide, not a measure of warehouse inventory or this program’s cost.
Some of Amazon’s Nvidia AI servers are reportedly waiting in warehouses rather than serving cloud customers because their intended sites lack enough power. Business Insider reported the staging arrangement, citing internal emails, according to an October 20 Investing.com article published by Yahoo Finance. Amazon defended early staging as routine and disputed claims about its readiness to open the facilities.
The internal August emails describe an “Early Rack Receipt” program for Nvidia GB300 server racks at U.S. facilities. The program allows Amazon to accept the equipment before final deployment. Business Insider reported that finished hardware was being held in intermediate warehouses because several Amazon Web Services sites lacked the utility capacity needed to run it.
The reported plan called for roughly $5 billion in Nvidia server racks, with expectations of accepting another $3 billion in hardware each month for the foreseeable future. Those are planned receipts described in the report, rather than a measure of how much computing capacity Amazon has brought online.
The financial change is when Amazon takes ownership. Previously, it took ownership and recorded expenses for server racks when they arrived at operational data centers, the article says. Accepting them earlier moves that step ahead of deployment, committing capital to equipment before Amazon can earn cloud rental revenue from it.
An Amazon spokesperson called early staging a routine “best business practice” intended to allow rapid installation once power capacity becomes available.
Amazon also disputed internal indications, as described in the report, that it lacked a clear path to bring the facilities online.
Loading discussion...
Join the conversation
Explain when readiness would justify tying up the money.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.