Nvidia Has Put More Than $170 Billion on the Line for Neoclouds

Nvidia has become something of a central bank for companies expanding the market for its chips beyond tech giants, backstopping customers and demand. But as Nvidia-backed cloud firms come under pressure, the question is how much further it will go.

The chipmaker has already disclosed as much as $164.5 billion in guarantees and other commitments to support new data centers, including agreements to pay landlords if tenants default on leases and promises to buy capacity, filings show, as of the company’s July quarter. That’s on top of more than $7 billion it has invested or agreed to invest directly in AI cloud providers.

Those investments are part of Nvidia’s broader portfolio of equity stakes, valued at nearly $100 billion as of July, which also includes investments in data center developers like SB Energy and big AI labs including OpenAI and Anthropic.

When it comes to AI cloud companies, Nvidia’s investments help seed their growth, providing capital to go toward project costs or purchases of graphics processing units that lenders typically won’t finance in full. Nvidia’s lease commitments and guarantees, meanwhile, can make it easier to finance new data centers or the Nvidia chips inside them.

But equity markets are souring on AI cloud firms and their ambitious growth plans, with shares in CoreWeave, Nebius and Iren tumbling last week. Nvidia-backed companies including Firmus and Lambda have pulled or postponed initial public offering efforts, while SB Energy and Nscale made their IPO paperwork public last month but haven’t moved ahead. Meanwhile, financiers of some data center projects are getting jumpier.

That all raises the prospect that Nvidia may have to go further to shore up confidence in the sector, despite its substantial existing commitments and growing demand among investors for protection on its own debt. Nvidia declined to comment.

Nvidia says the financial risks it faces from supporting AI companies are manageable, with Chief Financial Officer Colette Kress saying on an August earnings call that its “risk is limited” in backing AI labs because it can redeploy its chips to other customers. And Nvidia’s commitments wouldn’t necessarily all come due at once if major cloud customers were to default.

Still, stepping in if neoclouds struggle to find customers or secure financing could raise bigger questions about Nvidia’s growth prospects, especially as Google, Amazon and Microsoft invest more heavily in alternatives to Nvidia GPUs. Nvidia’s nonhyperscaler customers, including neoclouds and large enterprises, accounted for 45% of its data center revenue, or $40 billion, in its July quarter, more than double the number a year earlier.