Why Citi Thinks Microsoft Beats Neoclouds in AI Debt Squeeze

What Is Happening in the AI Market Right Now?

Rising interest rates are creating a fresh set of financial hurdles for specialized artificial intelligence companies.

CoreWeave, a prominent provider of AI cloud infrastructure, saw its shares tumble by 15% in a single day recently. However, according to Tyler Radke, co-head of US software research at Citi, this drop was not caused by weak customer demand.

Instead, escalating borrowing costs are squeezing the business models of fast-growing specialized cloud providers.

Understanding Neoclouds and Their Borrowing Model

To understand the issue, it helps to look at how neoclouds operate. You can think of a neocloud as a specialized rental service for high-performance computing hardware.

These companies purchase massive arrays of specialized computer chips, such as Nvidia's flagship H100 microchips, and rent them out to software developers who need raw AI processing power.

Because this high-end hardware is extremely expensive, neoclouds frequently borrow large sums of money to finance their purchases.

On the surface, customer demand remains very strong. Radke points out that rental rates for Nvidia H100 chips are sitting near record highs across the industry.

How Rising Rates Sparked the Debt Squeeze

The main issue stems from shifting conditions in the broader bond market. The 10-year US Treasury yield has climbed roughly 30% from its March low point.

This steep increase makes borrowing much more costly, particularly for companies that lack investment-grade credit ratings. It arrives at a time when the market is attempting to digest a growing glut of AI-related corporate debt.

For instance, CoreWeave's bond yields recently surged to nearly 15%, indicating that fixed-income investors now demand significantly higher yields to lend them money.

Venture capitalist Chamath Palihapitiya drew wide attention to this trend when he shared a chart of CoreWeave's bond yields on X (formerly Twitter). CoreWeave's stock plunged 15% that day, as reported by Fortune.

Other tech firms that took on debt to build massive data centers also felt the pressure. Tech giant Oracle, for example, closed 8.8% lower on the same day.

Who Controls the AI Infrastructure Market?

In an interview with CNBC, Tyler Radke emphasized the influential role that bond markets currently play in shaping valuations.

"I think the bond market is really in control of this narrative right now."

Radke argues that higher borrowing costs will eventually be reflected in higher contract prices charged to future neocloud customers.

For its part, CoreWeave told Fortune that its current backlog of customer contracts is large enough to comfortably pay off all of its debt.

Even so, analysts at research firm Gartner offer a cautious long-term industry outlook. Gartner projects that 80% of neocloud providers will lose their independence by 2030, largely due to mergers, acquisitions, or financial stress.

Why Citi Prefers Microsoft Over Smaller Cloud Players

Because of these rising debt costs, Radke ranks Microsoft among his highest-conviction stock ideas.

He highlights the strong profit margins and cash flows of established software leaders like Microsoft and data analytics firm Palantir.

Unlike smaller debt-reliant cloud providers, Microsoft generates positive free cash flow even after spending heavily on annual capital expenditure.

As a result, Microsoft does not need to rely heavily on high-interest debt markets to fund its data center footprint.

Radke notes that owning Microsoft shares allows investors to gain exposure to AI growth without needing to predict how fast hardware depreciates or how smaller firms will refinance debt.

Different Analyst Approaches on Wall Street

Radke clarifies that his view is not a simple either-or scenario. Semiconductor and chip stocks have outperformed software stocks since the AI market expansion began.

Past performance of specific stocks or sectors does not guarantee future performance.

Other market analysts hold contrasting views. For example, analyst Tom Lee issued a buy call following the recent market selloff, identifying both neoclouds and chip stocks as attractive opportunities.

Ultimately, Radke's thesis hinges on whether neocloud customers will accept higher contract prices over time. Until those pricing details become clear, bond market investors hold strong power over how AI infrastructure gets funded.