The market seems increasingly focused on risks to some of the players in the race for AI computing power. It isn't quite as worried about the data centers that will host that computing.
While stocks are about upside, bonds are about what can go wrong. So when the bond market takes different views on different parts of the AI financing market, it can be a good guide to what investors view as the most salient risks.
"You should not view the AI trade in credit as one big trade," says Lotfi Karoui, multiasset strategist at Pimco. "There is going to be differentiation."
This came clearly into focus this past week when bonds issued by CoreWeave, which sells computing power to AI companies and others, sold off sharply.
CoreWeave is an upstart player in the market for compute power, sometimes termed one of the "neoclouds" that are focused on AI cloud computing, and it competes with the likes of Amazon or Microsoft. And like those giants, CoreWeave has a rapidly growing need for data centers to host that computing.
There has been a surge in issuance of bonds to finance the construction of data centers for AI computing. This year through Oct. 16, data-center issuers have raised more than $100 billion in the U.S.-dollar bond market, or about 25% of AI-related gross issuance in U.S. currency, according to data from Dealogic and Goldman Sachs credit research analysts. That is up from around 16% the prior year.
When the data center is going to be leased to a hyperscaler, the data center's financing bonds are typically viewed as riskier than the bonds of their tenants. Microsoft, for example, is one of the market's few remaining triple-A rated companies. It is a pretty sure bet to pay its leases. But there can be additional risks associated with the data center, such as the possibility it won't be completed. So those bonds still trade at a higher yield.
CoreWeave has a different credit profile. With higher leverage, its bonds are rated "junk" and they trade in the high-yield market. The bonds issued to build data centers that CoreWeave will be leasing actually trade at narrower yield spreads over Treasurys than CoreWeave's own bonds. This spread is how investors price the credit risk of a corporate bond.
Both CoreWeave and associated data-center bonds have sold off this month, but the data-center bonds still trade at narrower spreads. CoreWeave bonds due in 2030 now yield close to 15%, according to MarketAxess data. A bond issued last year to finance Applied Digital's build-out of a North Dakota data center leased to CoreWeave, however, currently yields around 8%.
The recent AI data center bonds typically have structures that are different from a company's bonds. They can be secured by the data center being built, while corporate debt is unsecured. How much security that provides, though, also rests on the completion of the data center, and on the value of the data center to another tenant. Effectively, investors are betting that overall demand for data infrastructure means the data center will be of use to someone, even if the current tenant runs into trouble.
The bonds can also be structured in ways that add other layers of protection. Some of the recent data-center bonds are designed to repay their principal within the life of the lease, so that investors don't have to worry about getting another, future tenant to be fully repaid.
Earlier this year, CoreWeave reworked a lease for an Applied Digital center. The lessee became a separate entity, with its own investment-grade credit rating, which is backed by a CoreWeave computing contract with Meta Platforms. Spreads on those bonds fell sharply when that happened.
Data center construction and usage risks still aren't settled, as recent trading shows. The political tide shifting against data centers in many communities might have added to what investors demand for the risk a center won't be built. But if data centers become harder to build in the future, there might also be more value in one that does get done.
"With a data-center bond, you're underwriting assets, and our view has been that when it gets completed, that can be a meaningful catalyst for spreads to compress," says Andrew Keches, who is tech, media and telecom credit analyst and head of high-yield research at Barclays.
The bigger picture is that as the market for data center financing matures and a wider variety of bonds and business models come to market, investors might be differentiating more between credits.
"Crowding out can happen within the AI financing chain," Karoui says. "Investors now have a lot more variety. Coupled with wider spreads and higher yields, you become a little pickier."
That might not be reassuring to investors right now. But in the long run, a more selective market is healthier than one that just goes through cycles of boom and bust.
Comments