Shares of CoreWeave are under pressure after the AI cloud-computing firm issued full-year guidance below Wall Street expectations. The company’s struggles to expand its data centre capacity have raised doubts about its ability to sustain rapid growth.
BNN Bloomberg spoke with Gil Luria, head of technology research at D.A. Davidson, who maintains a sell rating and the lowest price target on the stock. Luria says the company’s heavy debt load and operational delays could destroy shareholder value as competition intensifies in the AI infrastructure market.
Key Takeaways
- CoreWeave shares fell after issuing full-year guidance that missed analyst expectations.
- The company delayed data centre expansion due to construction and supply constraints.
- A planned US$9-billion acquisition of Core Scientific failed after shareholder opposition.
- Analyst Gil Luria says high borrowing costs and weak returns are eroding value.
- Luria maintains a sell rating and US$36 target, warning the firm could face bankruptcy.
Read the full transcript below:
ROGER: Shares of CoreWeave are under pressure today after the AI cloud-computing company provided guidance for the full year that fell below analyst consensus. My next guest has the lowest price target on Wall Street and believes the business is not worth scaling. Let’s welcome Gil Luria, head of technology research at D.A. Davidson. Gil, thanks very much for joining us.
GIL: Thanks for having me.
ROGER: We saw revenues come in better than expected, but it’s the guidance that has you worried. Talk a little about the revenue, though. What did you like or not like about it?
GIL: Well, the company is in ramp-up mode. They’re providing compute capacity for Microsoft and, in the future, they’ll be providing capacity for OpenAI — and that’s what got pushed out. They were supposed to invest US$20 to US$23 billion of capital expenditure this year to create that capacity, and now they’re saying it will only be US$12 to US$14 billion, which is obviously considerably less. That puts them behind, and they actually had to go to OpenAI to renegotiate their deal so they could continue to get paid, even though they’re that late. That’s the disappointing part. CoreWeave billed itself as the company that can get compute capacity out quickly, so the fact that they had to push out this much of the data centre expansion into next year is the disappointment today.
ROGER: So what went wrong then?
GIL: Data centres are hard. CoreWeave is mostly a financial company that exists to set up loans, match them with chips and customers, and they’re stumbling on the fact that building data centres has always been difficult. It’s especially hard now, since there are so many bottlenecks — it’s hard to get the land, the power, and even electricians to come in and install HVAC systems. They had a deal with Core Scientific, which they tried and failed to buy, to help with that capacity. Now Core Scientific isn’t delivering on its promises.
ROGER: Were they too ambitious, or was it something else?
GIL: We don’t know all the details, but again, these are challenges that many companies building out data centres are facing. It’s just that most of the other companies — including Nebbias, a competitor that reported this morning — are still able to get the capacity out. CoreWeave has needed to push that into next year, based on some combination of those difficulties. A lot of things need to happen for you to be able to turn on a data centre and provide that compute capacity to a customer.
ROGER: The backlog, I think, is at US$55.6 billion. It includes 2.9 gigawatts in contracted power. Are they going to be able to come close to that? Are they looking at lawsuits or any kind of challenges if this doesn’t come through?
GIL: I’d ask another question — should they roll out that capacity? So far, CoreWeave has around 30 data centres and is providing a return on assets of about four per cent. It costs them nine per cent to borrow the capital to invest in those data centres. That’s very much like selling $10 bills for $5 — it’s like borrowing on margin to buy treasuries. Even if they do deploy that capital and complete the backlog, they’re destroying value. CoreWeave is the poster child for what’s going wrong with the great data centre buildout. Companies like Microsoft, Amazon, Google and Meta are using cash on hand and customer funds to build data centres and get good returns, while these incremental players are borrowing very expensive debt to build something they don’t have experience doing. They’re destroying value as they go. So it’s less a matter of whether they can convert the backlog — it’s whether they should.
ROGER: Is any of that tempered by the fact they announced their sixth contract from a prominent hyperscaler, though they haven’t said who it is?
GIL: Those hyperscalers are taking advantage of the situation. If someone else is willing to build data centres for them and provide capacity, that means less capital expenditure for the hyperscalers — less drag on margins, less borrowing they need to do. So they have leverage there. They’ll continue to buy inexpensive capacity from CoreWeave as long as someone is willing to lend CoreWeave the money to build those data centres. That’s a good deal for the mega caps and hyperscalers — not a good deal for CoreWeave’s shareholders.
ROGER: Right, and you have a sell rating on the stock, with the lowest price target set at US$36?
GIL: That’s right. We think the debt holders really own all of CoreWeave’s assets. Equity holders will get almost nothing by the time it’s all said and done.
ROGER: Can they move it then? Is anyone out there interested in it?
GIL: Once they go bankrupt, Microsoft or Meta will probably buy the assets for pennies on the dollar.
ROGER: You really think that’s where it’s heading?
GIL: Within the next few years, yes.
ROGER: Wow. We appreciate you joining us with this, Gil. Thanks very much.
GIL: Thank you.
ROGER: Gil Luria is the head of technology research at D.A. Davidson.
---
This BNN Bloomberg summary and transcript of the Nov. 11, 2025 interview with Gil Luria are published with the assistance of AI. Original research, interview questions and added context was created by BNN Bloomberg journalists. An editor also reviewed this material before it was published to ensure its accuracy and adherence with BNN Bloomberg editorial policies and standards.
