Why Citi Thinks Microsoft Beats Neoclouds in the AI Debt Squeeze

CoreWeave shares fell 15% in a day last week, but Citis Tyler Radke says rising rates, not weak demand, drove the neocloud selloff. He ranks Microsoft among his highest-conviction ideas. Neoclouds rent Nvidia chips to AI developers, often with borrowed money. Radke, co-head of US software research at Citi, says rental rates for Nvidias H100 chips sit near records. Yet CoreWeaves bonds yielded close to 15%. Can Neoclouds Pass Higher Borrowing Costs to Customers? Radke says the 10-year Treasury yield has climbed about 30% from its March low, hitting borrowers without investment-grade credit ratings. That lands on a market already absorbing an AI debt glut. The stock fell 15% that day, Fortune reported. Meanwhile, Oracle, which has also borrowed to build data centers, closed 8.8% lower. Radke argues that higher borrowing costs will end up in the prices neoclouds charge on future customer contracts. I think the bond market is really in control of this narrative right now. Tyler Radke, co-head of US software research at Citi, told CNBC For its part, CoreWeave told Fortune that its backlog of customer contracts is large enough to cover all of its debt. Research firm Gartner, however, forecasts that 80% of neoclouds will lose their independence by 2030, Fortune also reported. Is Software the Next Winner of the AI Trade? Radke cites the margins and cash flow of Microsoft and Palantir, a data analytics software firm, and singles out Microsoft. Microsoft stays free cash flow positive despite annual capital spending and does not lean on debt markets, he says. Owning it requires no view on how quickly AI chips lose value or how a company funds itself, he adds. Radke does not frame it as either-or. Chip stocks have outperformed software since the AI boom began, he notes. That contrasts with Tom Lees buy call, which named neoclouds and chip stocks as buys after Fridays selloff. Radkes thesis rests on customers accepting higher contract prices. Until then, bond investors may hold the pricing power over how AI infrastructure gets funded.