Monologue: High Noon for CoreWeave & AI

0:00This is an iHeart Podcast.

0:23Finally, someone let me out of my cage. Welcome to this week's Better Offline Monologue, and I'm your host, Ed Zitron.

0:36We're approaching high noon for regenerative AI, and it's the true test of whether this industry has that dog in it. And I believe it does, but the dog in question is sick, covered in fleas and has swallowed a live grenade. Let me set the scene. A few weeks ago, I reported that Microsoft had pulled at least a gigawatt of planned compute capacity, according to a report by analyst T.D. Cohen, equivalent to the entire capacity of London or Tokyo. To be clear, the actual figure was much, much more than a gigawatt, but that figure was the only one we had solid confirmation for, despite the press sitting on their hands and not really covering this number.

1:10They should have. A follow-up report published on March 26th shed more light on the situation and revealed that Microsoft has actually pulled over two gigawatts of capacity across the US and Europe. That's the equivalent capacity of London and Tokyo combined. Worse still, T.D. Cohen-Orson said, and I quote, The pullback on new capacity leasing by Microsoft was largely driven by the decision not to support incremental OpenAI training workloads. In other words, this is, as I have predicted, Microsoft not only pulling back from generative AI, but walking away from supporting OpenAI's future expansion efforts.

1:44These efforts, I add, were a major driver behind Microsoft's massive multi-billion dollar CapEx spend, which was forecast to surpass$50 billion to$80 billion this year alone. For OpenAI to continue training their future large language models, they'll have to rely on a combination of whatever support Microsoft is still willing to provide, their relationships with Oracle and other cloud providers, and of course the Stargate Data Center project, which is contingent on money coming through from SoftBank, who need to fund OpenAI's$40 billion funding round and their own$18 billion contribution to Stargate, and CoreWave, the subject of my last monologue.

2:20I mean, OpenAI is dependent on them. As a reminder, CoreWeave is a company that sells access to compute for training and running generative AI models, and they're going public on March 28th, which is the day this monologue runs, annoyingly. I can't speak to how the stock will do. I'm not a stock analyst. I will, however, say that CoreWeave is a company burdened by multiple times its revenue and debt, and they're a company that lost$863 million in 2024. And according to the Financial Times, they're also facing nearly$7.5 billion in debt repayments by the end of 2026. And they also don't appear to have any path to profitability.

2:57And this IPO, it's not looking so good. As of the morning of March 27th, 2025, which is when I wrote the script, Corweave planned to cut the size of their initial public offering to around$1.5 billion, according to Bloomberg, which is down from the$2.7 billion or so they plan to raise. When a company goes public, they usually are selling a bunch of stock to get money into the company. Sometimes they do it just because they want to make more money. Sometimes they want to be a public company and take the fame of that. Or in this case, I believe because they're desperate and they need whatever they can get.

3:28Now, this is problematic for a number of reasons, chief being that the information reports that Corweave's cash burn will be$15 billion in 2025, which is up from $6 billion in 2024. None of these fucking companies make any, they just burn money. It's insane. Now, Corweave has just over$6 billion in available liquidity between loans and the$1.3 billion or so in cash they have lying around. While they could borrow more, their largest$7.6 billion loan includes a covenant that states that if the company takes on more debt, it can only be used to pay that loan back. which means that in order to meet Corby's future capex spending goals, they need to raise vastly more than$15 billion, either through new debt or through selling new equity to new investors, or as they're doing via its IPO.

4:13$1.5 billion is a drop in the bucket. Not for me. If you have $1.5 billion and want to give it to me, please email me at easyatbetteroffline.com. I would love that money. Please give me$10. I don't fucking... Anyway, I'll put this really simply. A good, stable company going public does not reduce the size of their initial public offering, nor do they have multiple stories from the Financial Times comparing them to Enron or WeWork. Remember, an IPO is an opportunity to buy equity in a company before it hits the public market, but the price will, at least in theory, continue to rise. It's the next best thing to being an investor in the company.

4:53Facebook had an IPO price of$38, and now at the time of writing, there are over$600. Pretty good, right? Not for call with. Furthermore, if generative AI was this massive growth industry that would usher in a golden age of tech valuations, the company selling the literal fuel, access to the data centers to run generative AI models, would not be the financial equivalent of an HR Geiger painting. It wouldn't be buried in debt with obscene interest rates and restrictive covenants. It wouldn't be burning many times more than its revenue, nor would it be beholden to a very small number of companies for the majority of their revenue, who can and do walk away at any time, as Microsoft has.

5:34Call Weave is not a healthy company. But wait, wait, wait. I'm not a smart guy, alright? Let me just tell you a story, though. Because I learned about this thing. you ever hear of a hedge fund called magnetar capital i'll tell you about them in 2006 magnetar capital got involved in something called a collateralized debt obligation which is a way of packaging up a bunch of debts that you could sell it like a financial asset or even gamble on whether it would succeed a hedge fund wouldn't buy a single mortgage for example but they may buy something with hundreds of thousands of mortgages maybe subprime ones i don't know and it turns out that these thingies were real bad because it made it hard to see the actual risk of the loans that made up the bonds and because they were easy to sell they encouraged mortgage companies to give mortgages to people who couldn't pay them because they knew they could sell them on to investors and wouldn't have to carry on the risk themselves this in turn created something called a let me just look a great financial crisis which apparently was bad for the economy Hey, what happened to Magnetarget?

6:39Wait, what? They're one of the two companies providing loans to CoreWeave? What the fuck? And they're the largest institutional investor? That doesn't sound good at - What? What? Just this fucking industry. Every time I think this industry can't get any fucking stupid. Pardon me. Pardon me. Well, at least they're not involved in any other major deal in general of AI, right? Right? Wrong. Wrong. So fucking wrong. Over in Silicon Valley, OpenAI. is for the third time close to finalizing a$40 billion funding round led by SoftBank. And wouldn't you know it, Magnetar Capital is in there too, according to Bloomberg, for as much as$1 billion, valuing the company at an astonishing$300 billion.

7:23And I recently got told about a term, suicide round. And that term refers to a time when a company gets such a high valuation that any future investment is effectively impossible. because when you're investing at a company at$300 billion, you need to see liquidity at some point. You need to see a chance to sell it. And guess what? Who's going to buy this dogshit merchant company that burns billions? OpenAI is a complete dog. Yet what's weird about this deal, I mean, other than the fact it isn't actually finalized and has taken months and involves Magnitar capital and it involves seemingly every side of the deal taking on debt, is that it also isn't exactly 40 billion dollars yet bloomberg reports that softbank will actually invest 7.5 billion dollars into open ai and provide another 2.5 billion dollars from a syndicate of investors which is just it sounds far fancier than this it's just a group of investors who invest together the remainder the 22.5 billion dollars from softbank and the 7.5 billion dollars from a syndicate will come at some sort of indeterminate time later this year which is weird right it's okay it's not that weird what the funding rounds especially big ones they they're meted out in tranches based on performance i talked to a few vcs before i did this and they say that's that's normal but it is extremely weird that this is considered a 40 billion dollar round despite the fact that only 10 billion dollars appears to exist at this time and i'm not being facetious or joking here softbank literally does not have the money to give open ai right now they have like 30 billion in cash, which is why they're not sending that much.

8:58And they're having to take on loans as well. I'll get to that in a minute. But nevertheless, it's also not obvious how any of this works long term, because the entirety of the generative AI boom appears to rest on OpenAI's ability to continue raising capital. Let me explain. Microsoft neglected to exercise$12 billion of future capacity with CoreWeave, which was quickly picked up by OpenAI, according to Semaphore. Around the same time, the company canceled plans for future data center expansion, referring to Microsoft here, that would have undoubtedly cost tens of billions of dollars. This heavily suggests that Microsoft is pulling back from generative AI writ large, and that a large degree of their spend with CallWeed was to support OpenAI's services, which OpenAI will now foot the bill for.

9:40For OpenAI to pay that bill, they'll need SoftBank to give them the$40 billion of funding, which SoftBank is on the hook for at least$30 billion of according to the information, which includes OpenAI's$18 billion contribution to the Stargate data center project to which SoftBank also owes$18 billion. For SoftBank to give them this money, the information also reports that SoftBank will borrow$16 billion in 2025 and$8 billion in 2026, which still does not add up to the$46 billion or so they promised OpenAI. And I haven't even mentioned the fact that SoftBank has also committed$3 billion a year in revenue to the agent product.

10:19But without that money or further funding, OpenAI will not be able to pay CoreWeave. So now everything rests on SoftBank, a company with a checkered reputation, to put it mildly, and if they buckle, everything else fails. CoreWeave, OpenAI, all of it. I don't know how any of this ends, but I'll tell you this. Absolutely nothing I'm reading suggests that there is any long-term viability in generative AI. OpenAI loses billions of dollars a year, and will have to pay CoreWeave, which also loses billions of dollars a year and has obligations that are many times in excess of its actual assets or revenue.

10:55Or at least what it will make from the public market to be. That's not going to help that much. If any of this made sense, CoreWeave would be the biggest IPO of the last decade. And while I doubt that will be the case, I promise you that regardless of what happens, you're about to witness something historic. And I want to say something. Some of you have given me feedback, said like, this stuff's really, like, it's a little much. like there's a lot of numbers. It's not. It's just very ominous when you see all these numbers. I didn't get it. I'm not a financial journalist, or I guess I am now, or I'm pretending to be one.

11:30And it all seems like a lot, but it really does come down to addition and subtraction sometimes with the occasional multiplication. I failed a lot of mathematics in my time. And British libel also stopped me really elaborating further at school. But the point is, I was never trained to do any of this stuff. This stuff seems very monolithic, but it really is much simpler. And the way you stop these companies being able to fuck around is learning this stuff. If you ever have any questions, go to the Better Offline Reddit. I will happily help walk you through any of this. It seems dense, but if we want to change things, it starts with stopping this feeling like magic.

12:09It isn't black magic. They want you to believe that they'll work this out, because if you think about any of these numbers for even two seconds, it kind of seems fucking stupid, because it is. Just because they say it's going to work out doesn't mean it well. And I think that we are potentially about to see something really shocking. I think we're about to see the greatest mask off dunce moment in the economy. If AI bursts, if the bubble bursts, you're going to see all these people that said AI was the future kind of walk it back, kind of stop talking about it. Anytime they try, scream at them.

12:46Scream at them, say, weren't you talking about AI like it was the biggest thing in the world like two minutes ago? Why weren't you talking about that? Hey, where's AI? Didn't you say AI was a big deal? Hey, Bob Iger, hey what's happening with the AI shit and they're gonna really try and aggressively walk this back never ever let them forget what they've done here never ever forget the terrible people holding up this spurious specious destructive disgusting bubble it's horrible it will be horrible to watch it collapse I feel bad for our economy but this is tremendous content

13:30Parking shouldn't slow you down. ParkWiz gives every driver a shortcut. Book ahead, save up to 50%, and skip the hassle of circling the block. Park smarter, park faster. ParkWiz. Download the ParkWiz app today and save every time you park. Mint is still$15 a month for premium wireless. And if you haven't made the switch yet, here are 15 reasons why you should. One, it's$15 a month. Two, seriously, it's$15 a month. Three, no big contracts. Four, I use it. Five, my mom uses it. Are you playing me off? That's what's happening, right? Okay. Give it a try at mintmobile.com slash switch. A front payment of$45 for a three-month plan,$15 per month equivalent required.

14:11New customer offer, first three months only. Then full price plan options available. Taxes and fees extra. See mintmobile.com. Stop settling for weak sound. It's time to level up your game and bring the boom. Hit the town with the ultra-durable LGX Boom Portable Speaker and enjoy vibrant sound wherever you go. Elevate your listening experience to new heights because let's be real, your music deserves it. The future of sound is now with LG XBoom. And for a limited time, save 25 % at LG.com with code FALL25. Bring the boom! XBoom. There's a lot going on in Hollywood. How are you supposed to stay on top of it all?

14:49Variety has the solution. Take 20 minutes out of your day and listen to the new Daily Variety podcast for breaking entertainment news and expert perspectives. Where do you see the business actually headed? Featuring the iconic journalist of Variety and hosted by co-editor-in-chief Cynthia Littleton. The only constant in Hollywood is change. Open your free iHeartRadio app, search Daily Variety, and listen now. This is an iHeart podcast.