Is there a scenario in which CoreWeave doesn't end up defaulting? I mean, I guess that is the question that investors are trying to parse out right now, but it does seem that the likelihood that they will not be able to make their interest payments is extremely high. Do you see a world in which they do figure it out?
Welcome to Prof G Markets. I'm Ed Elson. It is October 15th. Let's check in on yesterday's market vitals. The major indices declined on the latest inflation data and continued pain in the chip market. We will cover both of those topics today. The yield on 10-year Treasuries spiked back above 5.3% following that inflation report before retreating by the end of the day. Bank of America stock fell nearly 5% after the company missed on profits and took a $1.2 billion penalty. That charge was for a failure to monitor instances of money laundering. And finally, shares of Bending Spoons soared more than 42% on news that Berkshire Hathaway has built a 4% stake in the tech rollup company. By the way, I recently interviewed the founder and CEO of Bending Spoons. You can go check that out. And one last thing, today is the final day for voting in the Signal Awards. So, if you like our show, please head to vote.signalaward.com. Cast your ballot for us in each of these three categories: best video podcast, best daily podcast, and best money and finance podcast. We'll leave a link in the description to make it easy for you. Thank you so much. Okay, what else is happening?
AI cloud company CoreWeave might be on its last legs. On Tuesday, the yield on one of its bonds touched 15%, reaching a new record high. Those bonds come due in 2032. Then the stock tanked on fears that the company wouldn't be able to pay back its massive amounts of debt. Yesterday it closed down 16%.
CoreWeave brought down other AI infrastructure companies as well, including Oracle, Nebius, Broadcom, Nvidia, TSMC, and Micron, all of which ended the day in the red. The stakes are quite high for CoreWeave. As of the end of June, the company had about $52 billion in debt and lease obligations, up from roughly $15 billion a year earlier. Last quarter, it paid $640 million in interest, more than double what it paid this time last year. In yesterday's episode, our guest Gil Luria of DA Davidson warned us that if the AI trade falters, CoreWeave's equity could go, quote, to zero. So, is that going to happen? Here to break this down, we're speaking with Paul Kudrowski, managing partner at SK Ventures. Paul, great to see you. Thank you for joining us. So, CoreWeave has been taken to the cleaners today along with a lot of other heavily debt-laden AI names. So, Nebius is down 11%, Lambda is down 12%, Oracle is down almost 9%. At least that was the closing price. Let's just start with your general reactions. What is happening here? What are the markets telling us?
So, the funny thing is, this has taken so long to happen. I mean, the situation at CoreWeave hasn't really changed dramatically from eight months ago. It's just that it takes people a long time to notice. And the situation, of course, is that it's among the most indebted of the neoclouds and has, as you pointed out, an incredibly high debt burden. And that debt burden either has to be refinanced or has to be paid for by selling tokens to people who rent space in the neocloud. And all of those get more difficult when you have exactly this kind of debt spiral going on. No different than the US seeing the 10-year spiking and then all of a sudden that has an effect on the deficit. As people look forward with the increasing cost of CoreWeave's debt as the spreads blow out, they look forward and say, 'Wait a minute, that has implications for the refinancing and future raises.' And so as a result that becomes the same sort of debt death spiral that you've seen in some sovereigns and it's just playing out. It took a long time to play out inside of CoreWeave, but it's finally happening, and there's a whole bunch of reasons why. Probably among the largest is that in the same way that AI capex fundraising has squeezed sort of the mid-duration Treasuries, the 10-years for example, which is incredible and has never really happened before, the reverse is also true: that as the 10-year and equivalent duration Treasuries become more attractive, it becomes much more difficult for these highly indebted neoclouds to raise money. So this perverse relationship goes in both directions. And the net effect is both of them have much more consequential debt burdens and then interest rates going forward. And the difficulty for a neocloud like CoreWeave is, unlike the United States, it doesn't have a Treasury Department and can't print money in the basement.
Yeah. The CEO of CoreWeave responded to the price move and he said that the market is acting, quote, 'irrationally.' He said that the problems at other companies are, quote, 'bleeding into the narrative around CoreWeave' and that the market, quote, 'misunderstands the company.' Do you agree with these comments?
No. I mean, these are very typical. I mean, he has to say these things. I don't want to imply that there's any kind of criminal nefariousness in here, but you have to say these kinds of things to support the company and say that this is not company-specific, this is something broader in the AI marketplace. But it's not; it's very specific to the most indebted neoclouds because of these twinned phenomena of higher cost of debt, higher cost of debt servicing. So that's at the one end. And at the other end, Epoch AI had a great report about this, about the continuing staggering deflation in token prices. So, Epoch had a report out, I think it was earlier in the week, pointing out that we've had for the last four years roughly a 13-fold annual decrease in the cost of tokens. So, doing the math, it's pretty straightforward. It's price times volume. And I've got debt that I have to service from that price times volume. That becomes astronomically more difficult given the increasing volumes required. If I'm seeing an annual decline of around 13-fold in terms of the cost of tokens, to put it in context, that means you have to grow to stand still. You have to grow your token sales, if you will. I think about it like copper or oil or anything else. 1,300% a year — that's what 13-fold means. Bringing you back even more, that means you have to double your token sales every 90 days. This is very tough math, made worse by the blowout in the rates and the increasing cost of servicing the debt. So these two feed on each other. So if you've got declining token prices, which essentially means that unless they explode their business, they've got less money coming in. At the same time, you've got rising yields, also rising yields across the board. I mean, you've got rising...
Rising spreads, too. It's important to say it's not just the yield, right? I mean, so the spread over Treasuries is blowing out as well. So there's these two things happening at once, which makes it even more complicated. I guess the question is: is there a scenario — if spreads remain this high and if yields remain this high — is there a scenario in which CoreWeave doesn't end up defaulting? I mean, I guess that is the question that investors are trying to parse out right now. But it does seem that the likelihood that they will not be able to make their interest payments is extremely high. Do you see a world in which they do figure it out?
No. So I...
I see instead a world where at best you have fairly radical restructuring of their debt obligations, which is what the markets are telling you, right? When the debt's trading at, say, 80-odd cents on the dollar, when you do the default math, it's saying there's a significant likelihood I will not be repaid in full. And the reason, generally speaking, why you're not repaid in full is not because of an alien invasion. It's because the debt gets restructured. If investors actually thought it was unlikely to be paid at all, obviously it would be trading much lower, because we're talking about debt that matures in 2032. So it's not actually that far away. So when you see it trading at these levels, what the debt market is trying to tell you is: this stuff gets restructured. You're not getting a dollar. Gartner came out with an analysis. They estimated that 80% of neoclouds will be out of business by 2030. And there are a ton of these companies: CoreWeave, Lambda, Nebius, Lambda... I mean, I could go on, all these new names.
Or you can even look at the withdrawn — some of the withdrawn Australian IPO whose name escapes me, was K2-backed, I can't remember who it was.
But there's another example, right? So even the public markets are reluctant to receive these guys. By the way, as an aside, with Gartner, there's a great line from Mike Milken from years ago: you only get maximum points for predicting things before it happens. And so the problem with the Gartner stuff is, okay, yeah, now you're telling me when it's absolutely obvious and these things are distressed.
I'm sorry, I don't put a lot of stake in that. So yes, they're not wrong, it's just not particularly helpful to tell you this as the spreads blow out.
Yeah. I guess the question is, it sounds like you believe that this isn't just a Core problem. This is a systemic problem. This is true of — I mean, does it go beyond the neoclouds? Do you think this is true of Oracle as well? I mean, Nvidia was down a little bit, you know, 3 or 4%. I mean, their balance sheet, I think, is a lot more healthy than a lot of these companies, but...
Yeah. So I guess who do you consider to be in the red zone?
So the neoclouds across the board, and probably Oracle as this kind of, I don't even know, wannabe hyperscaler neocloud, has managed to find itself in the gun sights of two different bullets from two different rifles. So I think that's the group that's most troubled here. And I think the way to think about it is much like what was happening during the GFC, the global financial crisis, where all of these banks were saying, 'Listen, there's nothing wrong with our debts and we're not defaulting on our debt,' but the market instead was saying, 'Here, what you have is a refinancing problem. We see a wall of refinancing ahead.' And there's a lot of data out here to show that between '28 and '32 there's a massive wall of refinancing, and it ain't going to happen at the prices that justify your business as it currently looks.
Yeah. One pushback that I've heard is CoreWeave has this Nvidia relationship. Nvidia invested more than $2 billion in the company. They own around 10% of the company. They have committed to buy billions of dollars worth of compute. Perhaps there's a world in which Nvidia bails them out, or Nvidia saves them, or they make sure that CoreWeave and any of their other little neocloud startups that they have backed, that none of them will fail. Is that something that is, you know, is that a reasonable prediction? Do you think that could actually happen?
No, I don't think so. I think this is like this classic speculative game the market likes to play. It's the old sort of 'if my aunt had wheels, she'd be a bicycle' problem. Yeah, absolutely. If I grant you that, then everything else follows. But I have to grant you something wildly implausible in the first place, which is that Nvidia sees the need to own and restructure a neocloud. And it's really difficult to see why they would do that.
Well, wouldn't the problem be that if CoreWeave, you know, goes out of business, if Core has to default on its debt, is there not a situation where investors' confidence in the entire AI trade will be completely shaken, and that this could be sort of the beginning of the end and the house of cards starts to fall? I mean, might it be in Nvidia's best interest to take their billions of dollars in cash and just say, 'Okay, we'll make sure that you stay propped up, keep you afloat, so that we all stay afloat once the dominoes start falling.' So, once it starts here, how far do you go with this?
It wouldn't surprise me to see someone try to staunch the bleeding as the first neocloud fails. But the problem is then, of course, the cascade will continue because the market will test your commitment to the trade, as it were. And as a result, how many of these companies are you willing to backstop in an effort to make sure that you don't have this rippling cascade? And suddenly you say to yourself, as an Nvidia or someone else, 'Hang on a second, I'm not Hank Paulson, I'm not the Federal Reserve, I'm not in the business of bailing out the system. The system goes down without me.'
So my understanding is you believe that we are going to see a pretty significant correction. How close in your mind are we to that point, given what we saw with Core?
Timing is always impossible. I mean, I've been saying sometime between now and the first quarter of next year, but I saw Paul Tudor Jones on CNBC yesterday saying sort of something similar. You know, so he's much smarter than I am. So I'll go with Paul and say the first half of the year, but I think it could be sooner. The thing is that...
One way to think about what's going on, and I get this question all the time, is the timing question. And they say, 'Well, it can't be soon because market price-earnings multiples are nothing like they were back in the dot-com era. Where's the bubble, Paul? Where's the bubble?' And of course the answer is the valuation of OpenAI alone, combined in the private markets and what we're likely to see as they come public, is already more than the peak Nasdaq valuation when it crashed in 2000. So the problem is people are looking in public markets for a thing that's sitting in plain sight in private ones.
And so we're already there, and the triggers are everywhere.
Yes. But it's your view then that the bubble exists, but it's not pervasive to the entire market. It's in pockets of the market. It's in places like private markets, or the neoclouds, or certain chip manufacturers. There are some stocks which are way over their skis.
