Nvidia’s (NVDA) AI chips have fueled a new debt market, and famed short seller Jim Chanos is flagging its risks.
Emerging AI cloud computing companies looking to rival Big Tech in the market for artificial intelligence have secured large loans backed by the $5 trillion chipmaker’s graphics processing units (GPUs), using that capital to buy more AI chips to scale up their operations. The firms lease space in data centers, fill them with AI hardware, and rent out computing power from that hardware for tech companies to train and run artificial intelligence models.
Four firms in the growing sector — three of which count Nvidia as an investor — have more than $20 billion in debt using Nvidia’s AI chips, or GPUs (graphics processing units) as collateral, the Information reported in July.
Chanos, who is famous for predicting the fall of Enron during the dot-com bust, sees red flags in the rise of GPU-backed debt because the neoclouds securing the loans have no clear path to profitability, making the debt hard to pay back.
“Business models like the neoclouds, a lot of the AI companies themselves … are just loss-making enterprises right now,” Chanos said in an interview with Yahoo Finance last week. “You’ve gotta hope that changes, because if it doesn’t, there’s going to be debt defaults on these things.”
Read more: How to protect your portfolio from an AI bubble
The financing trend, pioneered by CoreWeave (CRWV), is a form of asset-based lending: Money-losing companies can use their assets as collateral to secure capital they otherwise wouldn’t be able to access, typically at higher interest rates.
Because the so-called neocloud firms are so much smaller than the tech giants they look to rival, GPU-backed financing provides a helpful path in scaling up quickly. Nvidia has invested heavily in the neocloud space as it looks to grow its customer base beyond Big Tech.
Nvidia-backed CoreWeave, which went public earlier this year, and rival cloud provider Fluidstack each have roughly $10 billion in debt using their store of GPUs (graphics processing units) as collateral, according to the Information. Lambda and Crusoe, which are also backed by Nvidia, have $500 million and $425 million in GPU-backed debt, respectively.
CoreWeave reported a loss of roughly $65 million in 2024, while Fluidstack’s was less than $1 million during that year. Lambda’s and Crusoe’s financials aren’t publicly available.
The other big risk in AI chip-backed debt is a topic of major debate in the tech investment world: Nvidia’s GPUs may depreciate, or lose value, more quickly than companies are accounting for.
