CoreWeave shares fell nearly 25% in the week to October 16. One of the company’s bond yields briefly topped 16%. Together, the market moves put a price on a key question for the AI boom: how costly will it be to keep building data centers?

Reuters linked the pressure to inflation data that raised expectations for higher interest rates, as well as investor doubts about whether AI infrastructure companies can earn enough to support their debt. The pressure also spread to other data-center and cloud stocks.

Building capacity takes cash

In a quarterly filing with the US Securities and Exchange Commission, CoreWeave reported $35.6 billion in total indebtedness as of June 30. The filing also reported $5.5 billion in cash and cash equivalents, plus about $10 billion available through credit facilities. That credit is borrowing capacity, not money already in the bank.

CoreWeave’s spending shows why access to financing matters. In the six months to June 30, the company:

  • Generated $3.7 billion in operating cash flow.
  • Paid $14.1 billion for property and equipment, including infrastructure.

The company said it expects to keep funding investment through a mix of debt, equity, loan facilities, equipment financing and cash.

CoreWeave builds computing capacity and leases it to customers. If lenders demand higher returns, adding capacity may become more expensive, even as the company reports customer demand and continues to invest.

What the market moves show

Reuters reported that the yield on a CoreWeave bond due in October 2031 briefly rose above 16% on October 15. A bond’s yield rises when its market price falls. Reuters also reported that the cost of insuring CoreWeave’s debt against default rose above 1,000 basis points that week.

Those prices do not prove the company cannot meet its obligations. They show that investors are demanding more compensation for risk. In separate Reuters reporting, investors said financing remained available, but terms could tighten for projects farther from completion.

CoreWeave’s figures show the scale of its investment and reliance on financing. The October sell-off suggests that access to money may come at a higher price.

For the wider US AI build-out, the next test is whether demand and customer contracts can support the cost of building more capacity. Higher borrowing costs do not stop that construction by themselves, but they can make expansion more expensive.