Why Amazon Wants to Sell Eight Billion Dollars of Nvidia Chips and Lease Them Right Back

Why Amazon Wants to Sell Eight Billion Dollars of Nvidia Chips and Lease Them Right Back

When headlines claim a tech giant wants to offload billions in hardware, panic usually follows. People picture a fire sale. They imagine surplus silicon piling up in a warehouse because demand dried up.

That is not what is happening here.

Amazon is reportedly talking with investors about moving roughly eight billion dollars worth of advanced Nvidia Grace Blackwell chips into a special-purpose vehicle and then leasing those exact same chips right back.

You read that right. They want to sell the hardware to get it off their balance sheet, but they are keeping every single chip plugged into their data centers to run AWS workloads.

Understand why this move makes total financial sense.

The Reality of Hyperscale AI Spending

Look at the numbers behind modern cloud infrastructure. Amazon's total capital expenditures are scaling toward astronomical heights, with projections sitting around two hundred billion dollars. A massive chunk of that cash funds data centers, power supplies, cooling units, and clusters of high-end Nvidia graphics processing units.

CEO Andy Jassy has repeatedly noted that customer demand for AI compute still outstrips supply. AWS is not struggling to find people who want to rent these chips. They are struggling to finance the sheer volume of hardware required to keep up.

When you spend billions upfront on silicon that depreciates fast, your cash flow takes a beating. Bond markets start demanding higher yields for long-dated debt. Free cash flow dips. Wall Street starts asking uncomfortable questions about return on investment.

Enter the creative engineering of corporate finance.

How the Proposed Chip Leaseback Works

A sale-and-leaseback structure is not new, but applying it to cutting-edge AI clusters is a fascinating evolution.

Instead of keeping the hardware directly on Amazon's books as a capital expenditure, the setup works like this:

  • Amazon shifts a massive pool of Nvidia Grace Blackwell chips across multiple US data centers into a separate legal entity, known as a special-purpose vehicle.
  • Outside investors fund this vehicle primarily by buying its debt, while a small equity slice—around ten percent—is offered to participants.
  • Because Amazon holds a strong double-A credit rating, the debt issued by this vehicle can appeal to conservative institutional buyers like pension funds and insurers looking for yield.
  • Amazon signs long-term leases to keep operating the chips, turning a massive upfront equipment cost into predictable operating expenses.

You preserve your liquidity. You keep the compute capacity humming for your cloud customers. You pass a portion of the asset financing risk off to external markets.

What This Tells Us About the AI Infrastructure Race

People love to frame the artificial intelligence boom as a pure tech story. Really, it is a logistics and financing war.

Building out data centers at this scale pushes even the richest companies on earth to look for financial valves. When debt issuance gets expensive, you get creative. You bundle your physical infrastructure into off-balance-sheet structures to free up borrowing capacity for the next wave of expansion.

Nvidia stays happy because the chips are already bought, deployed, and churning through matrix multiplications. Investors get a yield-generating asset backed by Amazon's contractual lease payments. Amazon gets breathing room on its balance sheet without losing a single cycle of processing power.

Keep an eye on how these structured finance deals evolve. As hardware costs climb into the stratosphere, traditional corporate debt won't be enough to fund the future of computing. Expect more tech titans to turn their data centers into financial instruments.

MS

Mia Smith

Mia Smith is passionate about using journalism as a tool for positive change, focusing on stories that matter to communities and society.