Amazon is looking to move roughly $8 billion worth of Nvidia AI chips off its balance sheet, according to a report by Investopedia.

The figure marks one of the largest single GPU inventory positions disclosed by a major cloud operator to date. It signals that Amazon — which operates the AWS cloud division and has committed tens of billions of dollars to AI infrastructure — now wants to convert part of that chip stock into external arrangements rather than carry it as owned equipment.

Why does an $8 billion chip position matter?

Balance-sheet treatment of AI accelerators has become a live issue across the industry. Chips owned outright sit as capital assets: they depreciate, they tie up cash, and they expose the owner to utilization risk if demand for compute shifts.

Offloading hardware through leasing structures, sale-leaseback deals or resale shifts that risk elsewhere and frees capital for new capacity. For a company of Amazon's scale, an $8 billion position represents a deliberate restructuring of how AI infrastructure is financed rather than a routine inventory adjustment.

The reported effort also indicates that supply conditions for Nvidia's data-center GPUs have eased enough that a buyer of this size — a hyperscaler — can contemplate releasing stock, rather than hoarding every available unit as it did during the tightest phases of the AI buildout.

What does this say about the AI hardware market?

An $8 billion divestment attempt by one of Nvidia's largest customers cuts in two directions.

First, it suggests Amazon has procured more capacity than it currently needs to hold on its own books — a data point that hardware suppliers and cloud competitors will read closely. Second, any taker for that volume of chips would acquire serious compute capacity at market terms, which could affect pricing for secondary GPU capacity.

The report does not specify which Nvidia chip models are involved, the structure of the potential transaction, or which counterparties Amazon has approached.

Who else is affected?

Nvidia sits at the center of the story. Its revenue depends heavily on a small group of hyperscale buyers, and any sign that one of them is managing its GPU stock down — rather than up — feeds directly into the debate about how sustainable current AI capex levels are.

Other cloud providers face the same accounting question. As AI infrastructure spending moves through procurement cycles, the distinction between owned, leased and rented compute is becoming a strategic and financial decision, not merely a technical one.

Investopedia's report gives no timeline for a transaction and no indication of whether talks have begun. Amazon has not publicly commented beyond the report.