- CoreWeave’s stock slide puts neocloud debt under the microscope
- AI compute demand is booming, but the big question is whether cash flow can keep up
- Investors are watching when the bills come due
CoreWeave’s stock plunged early this week as investor concerns about the company’s debt collided with a fresh Gartner report that predicted up to 80% of neoclouds could cease independent operations by 2030.
As of Tuesday, its stock price was hovering in the $62 range, down 47% from six months ago.
In its report, Gartner argued current data center capital expenditure trends are unsustainable given enterprise and consumer demand won’t match spending until the early 2030s. The report specifically pointed to neoclouds as the most exposed if revenue falls short of expectations.
Fierce reached out to CoreWeave following the report’s release with questions about the company’s long-term future. The company responded by sharing a financial fact sheet.
“CoreWeave secures take-or-pay, long-term commitments from some of the world’s largest and most creditworthy companies, then uses those commitments to finance the infrastructure required to serve those customers,” it wrote.
CoreWeave added the majority of its debt facilities are matched against customer contracts, with the revenue from those earmarked to pay down the associated debt and interest without requiring a contract renewal.
CoreWeave had a revenue backlog of $104 billion as of the end of Q2 2026, with another $25 billion in new commitments in the first part of Q3. All told, it said it has $35.6 billion in debt principal. While payments on this debt are expected to remain in low single-digit territory through 2030, CoreWeave revealed it expects its principal repayments to spike to $14.9 billion in 2031 and beyond.
For comparison, Gartner Senior Director Analyst Will Sommer told Fierce it doesn’t expect AI revenue to catch up to data center capex until 2032 or 2033 at the earliest.
Rose colored glasses are off
CoreWeave and fellow neoclouds – including Nebius, Lambda, Crusoe and a long tail of smaller players – have enjoyed darling status in the early innings of the AI boom, as the industry scrambled to bring more compute capacity online. But cracks in investor sentiment began to show up over the summer, and it now seems investors’ rose colored glasses have finally been taken off.
Why? Because while demand for compute is high, it’s not entirely clear how much of that demand is actually coming from enterprises and consumers. And to the extent that it is coming from them, it’s not clear how much they’re willing to pay for it.
“I think the AI investment boom is increasingly facing a financing test,” Varun Datta, founder and CEO of venture capital firm Truth Ventures, told Fierce. “Demand for computing capacity can keep growing while some of the companies building it struggle to generate enough cash to service their debt.”
Gartner followed up its report last week with a fresh press release noting that only 13% of CIOs think their organization has seen significant value from AI so far. As Gartner VP Analyst Alicia Mullery put it: “Volume is not value.” And CIOs are increasingly facing pressure around AI governance and cost accountability.
The investor perspective
So, how are investors thinking about data center investments these days?
Datta said “A large order book wouldn’t give me much comfort if fulfilling it requires further borrowing or depends on a handful of customers continuing to spend aggressively.” He added he looks closely at how secure contracted revenues actually are, how concentrated the customer base is and – importantly – whether revenue will come in before debt repayments are due.
Courtney Olujobi, principal at Moon Pursuit Capital, noted there are also different stages of data center financing that carry a different kind of return for different kinds of risk. These include development (associated with securing real estate and power), construction, equipment (for all the hardware inside) and permanent financing.
“They suit different investors with different objectives, and the right question is always whether the return matches the risk at that specific stage. The places where the structure gets tested are the transitions between stages,” he said. “Investors who understand these handoffs can ask better questions of any deal: what has to happen after this stage for it to pay, and would it still pay if demand grew more slowly than expected. That framing is more useful than asking whether AI infrastructure is a good investment in the abstract.”
Read more about data centers on Fierce Network:
Hyperscalers face growing investor doubts over AI data center spending spree
About 700 data centers are being built in U.S. as public outcry increases