A Defense of CoreWeave
Rumors and riots... Anthropic's 5% pledge, Musk's SpaceX giveaway... Alan Gula's defense of CoreWeave... Lenders are betting on Meta, not CoreWeave... Dan Ives on 'Year Three' of the AI build-out...
Rumors, riots, and a new turn in the AI story...
This morning, we read rumors from a Middle Eastern news outlet about a U.S.-Iran "deal" to come later this week. Stop us if you've heard that one before.
Meanwhile, France has faced days of riots, complete with looting and fires in centuries-old cities. Similar financially driven anger has also popped up elsewhere in Europe, including Portugal and Italy.
In Saudi Arabia, political leaders canceled a major investor conference that was set to draw big U.S. bankers like Jamie Dimon, as the Iranian-aligned Houthis keep launching attacks.
And then there's AI, of course, and the unending stream of information about it – from data-center debates to debt worries... and now, a brand-new turn in the story.
In the past 24 hours, Anthropic CEO Dario Amodei and Tesla/SpaceX's Elon Musk – two of the "guardians of the frontier" as we've called them – have pledged to become philanthropic in what look more like choreographed PR announcements.
Amodei said Anthropic will commit 5% of its shares when it goes public to a new "public benefit fund." This fund could allocate money to explore the effects of AI on jobs... donate to nonprofits... or even make "direct distributions to specific populations."
Musk's idea was a little more straightforward. He posted on X that he'll give every child in the U.S. a share of SpaceX next year. Presumably, OpenAI's Sam Altman has his own "good faith" plan in the works, if it hasn't been announced by the time you read this.
And then there's CoreWeave (CRWV)...
This is a company that has been discussed around our offices, and in these pages, for a few years now. It's been all over the financial media lately, because folks are seemingly worried that the AI "neocloud" company could entirely go under.
CoreWeave's senior unsecured bonds have sold off lately, and its stock has been crushed. It's down about 55% from its most recent high in May, and down 67% from its all-time high a couple months after its IPO in 2025.
Last week, we shared an update on CoreWeave, including a warning from our colleague Whitney Tilson about the company. But we also said not everyone agrees. That includes another colleague, Alan Gula, who debated and discussed CoreWeave at our annual conference last month.
Today, Alan has some more thoughts and details on CoreWeave that he'd like to share here, exclusively for Digest readers. In short, he thinks fears about the company are overblown. Alan takes things from here and explains why...
CoreWeave is not facing 'imminent collapse'...
You've probably seen people call CoreWeave the "canary in the coal mine" for the AI bubble, or even a "subprime CDO with GPUs," comparing CoreWeave's debt to the infamous collateralized debt obligations of the 2008 subprime-mortgage crisis.
I (Alan Gula) disagree.
I've followed CoreWeave since 2024, when I invested in it as a private company. My conviction in the business has only grown since. At our annual conference in Las Vegas last month, I told the crowd that CoreWeave is one of the most misunderstood companies I've come across in my career.
CoreWeave operates data centers that provide computing power (or "compute") for AI workloads. It signs multiyear compute deals with clients such as hyperscalers Meta Platforms (META) and Microsoft (MSFT), plus frontier lab OpenAI. It then borrows money to buy Nvidia (NVDA) graphics processing units ("GPUs"), installs them in the data centers it leases, and delivers the compute.
The misconception is that CoreWeave loses money on each of these deals and is gambling on the residual value of the chips.
CoreWeave is small relative to Nvidia and the hyperscalers. Its market cap is around $35 billion. But it's central to the AI boom and is a forerunner in GPU-backed financing. So it's worth understanding why it isn't facing imminent collapse.
The concerns...
It's true that CoreWeave has yet to turn an accounting profit. Its net loss over the last four quarters was roughly $2 billion. But the compute deals themselves are profitable.
CoreWeave loses money because of the lag between when it buys the chips and when deal revenue starts flowing. In that gap, it still pays the data-center lease, the power bill, employee salaries, and interest expense – and depreciation hits the income statement. CoreWeave has also signed progressively larger deals, and the latest are in their loss-making phase.
So on a consolidated basis, the financial statements look like a disaster. What the market is missing, though, is that each deal has attractive returns. With conservative assumptions, I estimate many of them will generate internal rates of return of at least 20%.
CoreWeave keeps taking on more debt to finance these deals because it views them as attractive investments. It now has more than $50 billion in liabilities, and rising. Its free cash flow this year is expected to be negative $27 billion.
That worries the market. Some of its bonds now have a yield to maturity around 15%, and people are panicking.
But the company is not in distress...
It's never a good sign when a company's bonds plunge. But these senior unsecured bonds are "structurally subordinated" to the loans CoreWeave uses to finance most of its build-out. That means if CoreWeave has trouble handling its debts, these bonds wouldn't get repaid first.
CoreWeave finances its larger deals with delayed-draw term loans ("DDTLs") tied to specific data-center projects.
These loans are backed by the chips. But, more importantly, what lenders are really counting on is the "offtake agreement" behind each project – the contract a client signs to buy CoreWeave's computing power for a set number of years at a set price.
Think of a trucking company that borrows to buy a fleet of trucks for a five-year hauling deal with Walmart. A bank is really lending against Walmart's signature, and the client's payments flow back to cover the loan. So, lenders are mostly sizing up the customer, not the borrower itself.
CoreWeave's customers include the likes of Meta or Microsoft, which have strong balance sheets and creditworthiness. That boosts its odds of paying off these debts. Investors aren't betting on CoreWeave alone, but on its customers.
CoreWeave's DDTL 4.0, for instance, was signed in March. It carries an effective interest rate of about 6%, and it's rated investment-grade thanks to the client: Meta Platforms.
If lenders stop financing DDTL deals, CoreWeave's frenetic growth will slow. That's not a death knell, though. CoreWeave still has plenty of borrowing capacity under many of its recent DDTLs, and none of its senior unsecured bonds matures before 2030.
CoreWeave's stock is a high-risk play on the AI-infrastructure build-out. It's not suitable for most investors. But in my opinion, the company is not facing imminent demise.
And if I'm right, this has broader implications...
The fear around CoreWeave today is yet another hiccup in an AI boom that will continue.
So where does the AI boom go from here?...
Corey here again... As I wrote live from our annual conference last month, AI was everywhere during the week... from the presentations like Alan's... to the cocktail-hour floor, where a humanoid robot danced.
In our coverage, I also told you about one of our featured speakers, Dan Ives, the renowned technology analyst and former Wedbush managing director who recently established Yorkville Ives & Co.
He told conference attendees he thinks we're in "Year Three of what's going to be a 10-to-12-year build-out" of AI infrastructure. He described the technology as the "biggest transformational theme that we've seen in 100 years, with the U.S. leading."
Dan sat down with our Director of Research Matt Weinschenk and MarketBeat's Bridget Bennett at the conference to talk more about his view...
You see, AI has already created trillions of dollars in wealth. Nvidia, Microsoft, Meta, and other tech giants have seen tremendous growth as companies race to build the infrastructure powering the AI revolution.
The boom has drawn comparisons – including in these pages – to the dot-com bubble. But Dan says the biggest opportunities could still be ahead and that we've seen less than 15% of the total AI spending that's coming.
That said, which stocks could benefit the most? And which companies risk getting left behind? In this special collaboration with MarketBeat, recorded live in Las Vegas, Dan explains why the next wave of winners could extend far beyond the Magnificent Seven.
Click here to watch this episode – for free – right now on our YouTube page. Or you can find the episode here, on the members section of StansberryResearch.com under "Media."
New 52-week highs (as of 10/16/26): AbbVie (ABBV), Automatic Data Processing (ADP), American Financial (AFG), Abercrombie & Fitch (ANF), Alpha Architect 1-3 Month Box Fund (BOXX), BP (BP), Berkshire Hathaway (BRK-B), Chevron (CVX), Dorchester Minerals (DMLP), EOG Resources (EOG), Equinor (EQNR), Coca-Cola (KO), Kimbell Royalty Partners (KRP), Altria (MO), Pfizer (PFE), Visa (V), State Street Energy Select Sector SPDR Fund (XLE), and ExxonMobil (XOM).
In today's mailbag, a "tip of the hat" from one subscriber to another... feedback on Saturday's Masters Series essay from Nick Hodge – about what politicians in Washington can't control... and why knowing it can give you an investing edge... Plus, a follow-up on an inflation story from the McDonald's drive-thru we replied to in Friday's mail... Do you have a comment or question? As always, e-mail us at feedback@stansberryresearch.com.
"I'm a cowboy. When we tip our hat to someone, it is like giving two thumbs up. Here's a tip of the hat to Tony G. for his insight on the data center push-back. If data center access becomes 'rationed,' because nobody wants to allow construction of more capacity, the easiest way to ration access is to charge for it. Econ 101, but very insightful, Tony. I am going to steal your point and use it in a current debate in my (mostly retired) community in South Carolina.
"The only real 'affluence' in the county where I now live is retired people who made their money in the great American Capitalist Economy ('ACE'), but now they want to stop construction of a data center that helps fuel the next wave of American greatness and, more to the point, will actually bring NEW money to the county economy and give younger folks a reason to stay here, start a family, build a house, and shop at local stores." – Stansberry Alliance member Darrell W.
"Great summary [in Saturday's essay about how you can make money betting on government 'fixes' not being enough for certain commodities]. So simple, but hard to do." – Subscriber Serge F.
"Mr. McLaughlin, you had a point, I could be mixing things up... It's $1.50 vs. $7.29 vs. $9.99 for the combo. [Sausage McMuffin from the McValue Menu vs. Egg McMuffin vs. Sausage McMuffin with Egg.] Makes the point even starker. Again, thanks for all your excellent work." – Subscriber John H.
Corey McLaughlin comment: Glad we got to the bottom of this.
All the best,
Corey McLaughlin with Alan Gula
Baltimore, Maryland
October 19, 2026