Amazon Raises Chip-Rental Prices, Moves Nvidia Chips Off Books

Amazon is raising the prices it charges customers to rent high-performance microchips, the company announced, as it reportedly prepares to sell its Nvidia processors and lease them back.

The two moves, taken together, mark one of the clearest signals yet that the cost of financing the AI infrastructure buildout has begun to reshape how the largest cloud providers run their businesses. Amazon, one of the handful of hyperscale cloud operators bankrolling the AI revolution, is now adjusting both what it charges for compute and how it accounts for the chips that compute depends on.

What Amazon announced

The company said it will increase the rates customers pay to rent high-performance chips through its cloud business. For Amazon's cloud customers, that means the effective price of accessing the silicon that powers AI training and inference work is going up, effective with the announced change.

The price increase lands on the rental side of Amazon's chip business — the service through which customers pay to use advanced processors rather than buy them outright. That rental model has become the backbone of AI computing for companies that cannot or will not build their own data centers, and it is the segment where Amazon's pricing decisions ripple directly through the budgets of AI developers.

The sale-and-leaseback report

Alongside the pricing move, Amazon is reportedly planning to sell its Nvidia chips and then lease them back, according to the report. Under a sale-leaseback structure, a company sells an asset to a buyer — often a financial institution or leasing firm — and immediately rents it back, converting an owned asset into a leased one.

For Amazon, the reported transaction would move Nvidia processors off its balance sheet. The chips would keep running in Amazon's data centers, still serving customers, but Amazon would no longer carry them as owned assets. Instead, it would pay to lease the same hardware it previously owned.

The mechanics matter for how the company's finances read. Owning billions of dollars in cutting-edge processors ties up capital and shows up as assets that must be depreciated over time. Leasing the same chips converts that ownership into recurring payments, and the chips themselves sit elsewhere on the financial statements.

Why hyperscalers are feeling the squeeze

The report frames both moves as a sign of the financing crunch now facing the hyperscalers that are financing the AI revolution. Building out AI compute at scale requires enormous upfront spending on advanced semiconductors, and Nvidia's processors are the industry's most sought-after and most expensive hardware.

Amazon's reported pivot to sale-and-leaseback financing follows a well-worn corporate playbook for capital-intensive businesses: when the cost or the balance-sheet weight of owning core assets becomes a burden, sell the assets and rent them back. Airlines have done it with aircraft. Utilities have done it with power plants. The reported Nvidia chip transaction would apply the same logic to AI hardware.

The price increase on chip rentals points at the same pressure from the other direction. When the capital cost of the hardware rises or the financing around it tightens, the rental rates charged to end customers are one of the first levers a cloud operator can pull. Amazon has now pulled it.

What it means for customers and competitors

For companies renting compute from Amazon, the announced price increase raises the cost of building and running AI workloads on the company's cloud. Any customer whose AI budget depends on rental rates for high-performance chips will absorb the change directly.

For the broader market, Amazon's dual move — charging more for chip rentals while shifting reported ownership of Nvidia processors — offers a concrete data point on how the industry's biggest spenders are managing the financial weight of the AI buildout. If the financing crunch the report describes continues, sale-leaseback structures and rental price hikes may spread from Amazon to the other hyperscalers carrying similar hardware burdens.

Amazon has announced the rental price increase publicly; the Nvidia sale-and-leaseback plan remains, for now, a reported intention rather than a completed transaction. How quickly the deal executes, and whether other cloud giants follow the same off-balance-sheet route to fund their Nvidia fleets, will say much about the financial architecture of the next phase of the AI buildout.