Paying for AI infrastructure has strained Silicon Valley’s deepest pockets, with even Meta reporting negative $2.1 billion in free cash flow last week as its spending on data centers and chips surged. As investors question whether AI spending will generate enough revenue to justify its costs, the world’s richest man offered a grim prediction: Hundreds of smaller companies central to the buildout may not survive.
Elon Musk warned in an X post on Sunday that more than 200 neoclouds, the firms that buy chips and then rent out computing capacity, could face “guppy-level death and consolidation” because lenders are reluctant to finance businesses that depend wholly on AI demand.
“Debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers,” Musk wrote. He said SpaceX is exempt from this potential wipeout because it already closed $40 billion in debt financing and has Starlink to generate revenue that’s not tethered to AI.
Neoclouds borrow heavily to buy expensive Nvidia chips and build data centers, and then lease out that capacity to model developers. While hyperscalers like Google and Meta have other sources of revenue they can lean on to fuel their AI spending, neoclouds’ survival depends on appetite for compute. So if borrowing costs go up or customers cut spending, neoclouds will struggle to pay off their debts.
CoreWeave, one of the biggest neoclouds, illustrates this financial pressure. On one hand, its revenue backlog was $104 billion at the end of the second quarter, with $25 billion in new commitments added early in the third quarter. On the other hand, its latest quarterly filing shows $35.6 billion in outstanding debt at the end of June compared to $5.5 billion in cash and cash equivalents. And last month, CoreWeave took on $4.2 billion more in debt.
Concern about CoreWeave’s ability to manage this debt sent the cost of insuring it against default to a record high last week, with credit default swaps implying a more than 60% chance over the next five years.
This is an industry-wide concern. Analysts at research firm Gartner predicted on Oct. 14 that cost pressure and lack of a differentiating business model will force 80% of neoclouds to stop operating by 2030. Data center owners need to make $8.6 trillion in revenue to cover financing costs and capital spending, but enterprise and consumer AI revenue linked to those data centers is forecast at only $3.6 trillion.
Financial markets may represent another limiting factor. Apollo Global Management chief economist Torsten Slok told CNBC there’s a “shortage of capital” to fund AI infrastructure because the sector is competing against government debt and deficit financing.
Meanwhile, hyperscalers are raising capital by issuing debt and stock, with plans to raise even more in the future. As the biggest companies gobble up the capital that’s available, there are doubts about whether there will be enough left for smaller players like neoclouds.
But CoreWeave said that lenders are not treating every AI cloud provider alike.
“The AI cloud market is not a monolith. Lenders distinguish between speculative capacity and businesses with differentiated technology, creditworthy customers, contracted revenue and a record of delivering,” a CoreWeave spokesperson told Fortune. “CoreWeave is already a market leader operating at scale, with the technology and expertise that the world’s leading AI companies depend on. Those strengths underpin the opportunities we continue to see to finance our growth.”