CoreWeave, the Company Riding the AI Boom

0:05In the past few months, spending on AI has skyrocketed. NVIDIA is making a$100 billion investment in OpenAI. Amazon share is jumping today after the company announced a$38 billion partnership with OpenAI. Anthropic is adding another$50 billion to its AI compute bill. Nearly every big tech company has dedicated hundreds of billions of dollars to AI projects, promising a transformational new era. Most of them are names almost everyone knows. Microsoft, Meta, OpenAI. But right at the center of this boom is a company that isn't a household name. It's called CoreWeave. CoreWeave is this company that to a lot of people kind of came out of nowhere, like late last year, earlier this year, but they're actually this kind of major player.

0:52Our colleague Dan Gallagher reports on tech. Even though they're a small company relative to these really big ones, they are a major player in this particular realm of artificial intelligence data center computing. If you've ever used an AI product like ChatGPT, chances are you've interacted with CoreWeave. But you wouldn't know it. That's because the company provides the infrastructure for AI, leasing out the super advanced chips that are needed to power this technology. That means CoreWeave has positioned itself to be a big winner in the AI boom, with its revenue on track to more than double this year.

1:29But investors and industry insiders are watching this company closely. Because if the AI boom is actually a bubble, some expect CoreWeave to be the first domino to fall. So for CoreWeave, it's like really, they need AI to work. That's their entire business. Welcome to The Journal, our show about money, business, and power. I'm Jessica Mendoza. It's Wednesday, November 12th.

2:00Coming up on the show, the company you've probably never heard of at the center of the AI boom.

2:23CoreWeave began in 2017 with the purchase of a single chip. The founders were a group of Wall Street traders based out of New Jersey, and they had a hunch that GPU chips, made by a company called NVIDIA, were going to be a good investment.

2:40Could you just define what a GPU chip is? Sure. If you think about the computer that you're on has a main chip called a CPU, that's kind of the main brain. A GPU is an acronym for a graphics processing chip. And what it does is it helps the CPU process graphics more efficiently. That's the basic function. The way it does the computing, though, is it essentially accelerates the way the main chip works. And so when you put a bunch of GPU chips together, you know, in like a cluster with these other central brain chips, you can have essentially this powerful supercomputer. Initially, CoreWeave put those GPUs to work, mining cryptocurrency.

3:24But it was a volatile market. And when crypto prices crashed in 2018 and 2019, the founders looked at other ways their chips could be used. including artificial intelligence. These guys were coming out of the financial industry. The CEO was a former commodities trader. They've all kind of come up out of that era. And so that kind of fed their approach to this. Obviously, it fed the opportunity they saw. They realized they had this asset that was valuable, that they could pivot into like a profitable business model and went after that really quickly.

3:58Roughly when did demand for these chips begin to increase? Well, if you think about the launch of when ChatGPT became essentially public in terms of public launch, that was in late 22. Right. And really quickly, people flooded onto it, started using it. It got really popular really quickly. And that's what kind of triggered all these companies to start going really fast to build up their services. So with NVIDIA, it was in kind of early to mid-2023 that all of a sudden they started reporting a huge jump in sales of their chips, which just took on this explosive trajectory. And that was these companies like Microsoft, Google, Amazon buying them up, and also CoreWeave.

4:44So, yeah, everybody wanted these chips all of a sudden. CoreWeave had a bunch of them. So this was sort of a case of right place, right time for this company? Oh, very much. Mike, thank you for joining us today. Thank you for having me. Excited to be here. Last week, I got to sit down with CoreWeave's CEO, Michael Entrader, at the Wall Street Journal's Tech Live event in Napa Valley. And it's kind of legend at this point in tech circles, but you pivoted to AI infrastructure from crypto mining. Was there a moment when you realized you needed to change tech? I don't think there was a point at which we said, hey, we need to pivot.

5:20I think what we did is said, hey, we want to make sure that we have the ability to be super flexible in an incredibly volatile environment. And that turned out to be a good call. Today, CoreWeave leases its GPU chip power to some of the biggest tech companies investing in AI, like Microsoft and Meta. And its shareholders include NVIDIA and OpenAI. You know, we want to be the best solution for delivering massive scale. And scale is an important part of this, right? Like I always say, anybody can run a GPU, but can you run a fleet of hundreds of thousands of them so that you can train the most sophisticated models on the planet, so that you can serve inference for the most incredibly demanding consumers of compute?

6:07And that's a very different proposition. I mean, the move is obviously paying off big. So far, so good.

6:18As AI spending has ballooned, CoreWeave continues to reap the rewards. This week, the company revealed that it's secured$55 billion in sales contracts. Here's my colleague Dan again. There is such demand for these chips that every company that's a player here, Microsoft, Google, Amazon, everybody, keeps saying that we have way more demand than we have supply from consumers, from corporate, wherever it's coming from. And so when you're a company like CoreWeave who has these chips in the data center and ready to provide the computing, you've got a lot of business right now because demand is way outstripping supply for it.

6:55All these deals that CoreWeave has secured has made it part of a trend developing in the AI industry. It's called circularity. It works like this. Company A pays money to company B as part of a transaction. And company B turns around and buys company A's product or service, maybe using the money that company A gave them. This is also referred to as a dependency loop. CoreWeave is a company that's very closely entwined with other companies. Now, let me see if I can get this right. They are partly owned by NVIDIA, who also sells them chips. NVIDIA also sells chips to Microsoft, which is CoreWeave's biggest customer, and is also an investor in OpenAI, which is also a CoreWeave customer and a CoreWeave shareholder.

7:39Is this normal? Like, what is this? It feels it's so hard to wrap your mind around it. No, I would not call it normal. It's circular. You put your finger on it. And I think if you press these companies, they're going to say this is kind of how we work together to get this massive infrastructure project stood up because we're going to need all this computing infrastructure to do all the wondrous things AI can do. When I sat down with Mike and Trader, I asked him about the idea of circularity. There is a component of which it is a team sport. Nobody can do it all. And so you're working together to try and deliver a size and scale of infrastructure that the world has never seen before.

8:21It stands to reason that you're going to have companies working together, investing together, driving different parts of this market together. And I don't think it's actually unusual when you have a boom that's taking place in a very compressed period of time to have this level of interaction between the companies. Dan says the system of interconnected companies is OK as long as everyone is succeeding. But yeah, you get in this world where you start to wonder if there was a problem with one of these companies somewhere, does it like kind of teeter the whole system because everybody's now got these super complicated intertwined deals going on?

9:03And if there is a teetering, Dan says CoreWeave's high reward bet might become high risk. That's next.

9:25In March, Corweave made its stock market debut.

9:36Can you talk about how that went? Well, I mean, it's been a pretty volatile stock since it went public. I think it actually went down its first day of trading, and then it kind of had this explosive growth up. And it really swings a lot on things like sentiment about AI, their own reported results. I think last week, the stock lost something recently like about 15 % in a week. And that's not uncommon for them because it is pretty volatile. This week, CoreWeave reported its earnings and announced its revenue was on track to double this year. But the next day, the company's stock fell more than 16%.

10:14Dan says there are a number of reasons for this volatility. One of them is that investors are divided. Can all this spending on the AI infrastructure that Corweave provides continue at this breakneck pace? I mean, it depends on who you talk to. There are some that think, oh, AI is going to, this bubble, it's going to pop and not going to be around. And then on the other side are obviously the proponents, including all the companies involved, saying, no, it's going to reorder the economy and make the GDP double or all these other things. I think the more realistic risk is that open AI is going to be transformational, but it may take a lot longer than its proponents think.

10:53So you kind of have this push-pull with, you know, how fast is this technology really going to be adopted, really going to change things, and is it going to happen fast enough to essentially pay for all these financial commitments? Right. And so if that happens, if the AI bubble does burst or it takes too long for the technology to reap the benefits of this transformation, what would that mean for CoreWeave? It would be bad for them, honestly, because when you're Microsoft or Google, you're generating most of your revenue with businesses you've been in for a long time or very well established and are going to still be there if AI doesn't turn into what people think it will.

11:33Core Weave is like really specifically hinged on AI. So if AI demand really, really stumbled somewhere, theoretically, they could be left. Because if you're Microsoft paying Core Weave like a lot of money for like handling your computing, all of a sudden you just don't need that from them, you know, that's not good for Core Weave. I asked Mike and Trader about the AI bubble. I wanted to know if it's something Core Weave is concerned about. It's very hard for me to worry about a bubble when, you know, as one of the narratives, when you have buyers of infrastructure that are changing the economics of their company.

12:10They're building the future. And one of the narratives that has come up, too, is this idea of, even in the long term, companies that come up as winners and losers in this space. Are there ways that you can insulate CoreWeave from being the one left holding the bag, if anything does happen in that sense? So, look, from a risk management perspective, you know, one of the narratives around CoreWeave at the, you know, as we went public was, oh, you know, we're so dependent upon Microsoft. And now within the last quarter, you've seen us announce more deals with Meta, you know, a deal that's at least$14 billion.

12:48You've seen us announce more deals with NVIDIA. You've seen us like we're diversifying. We're doing all the things that a normal standard company should be doing to insulate themselves from the exposure and making the company more resilient, more scaled and better able to withstand any type of slowdown that could potentially occur. You know, it's not going to be a straight line. We know that. You're going to hit air pockets. We structure our deals out over term so that we're insulated from that. We don't have to absorb that. There are better balance sheets in the world to hang that on than ours.

13:29We're really thoughtful about that. We spend a lot of time kind of making sure that we're well positioned. And in Mike's view, one way to build fast is to leverage debt, which Corweave carries a lot of on its balance sheet. The way that we built our business and were able to scale our business to such enormous scale so quickly is that we used debt because debt is the correct way to do this. I realize I'm in Silicon Valley and maybe that doesn't get as much traction as it does in New York, but debt is the right way to do this. This strategy of using debt to finance a business is common on Wall Street, but not so much in Silicon Valley.

14:09Here's Dan again. Tech is usually a very cash-rich business. So for years and years, I've been on earnings calls where tech companies have bragged about the fact that they have no debt or almost no debt. So to see a company come into essentially the tech space and be open about the fact that they're building up with debt kind of contrasts with companies that can write these massive checks out of their bank accounts. That's where it stands out. What are the pros and cons of this strategy? Well, obviously, you know, debt carries a risk because if you need to generate enough cash, free cash flow to service the debt, you know, pay the interest.

14:46And in the case of Corweave, they've got a lot of debt they've got to incur in the future because they have to keep buying chips, keep expanding their networks, keep expanding data centers to service the contracts they're getting. So there's not just paying off the debt they have. There's going to be, well, they have the growing cash flow to pay for the growing amount of debt that they're going to have. But so why do it this way? Is it the only way you can do it? If you're their size and you're coming into this market and competing with the companies they're competing with, that's how you have to do it.

15:19During our conversation, Mike and Trader made it clear that he's confident that AI growth is going to continue. But he also talked about a part of his personality, which is key to understanding why he does business this way. his appetite for risk. I mean, ultimately, you know, for the first 20 years, 25 years of my career, I was a risk manager. That's what I did, right? And, you know, my job is to sit over, as I continue to move up through the ranks, over desks and then ultimately over full trading floors and over full companies. And so when I think about how to participate in this market, it's really through the lens of risk management as you're building at this unprecedented scale globally, as you're basically building a layer of infrastructure that, for all intents and purposes, truly did not exist five years ago and certainly will not exist in a way that we understand it in five years from now.

16:15and trader told me that the demand for this technology is just going to increase this is the line that we're hearing from all of these ai uh ceos and cfos and uh all the all the people who are very bullish on the growth of ai everybody seems so certain of it is that fair like are they seeing something that maybe the average person me isn't because when i think of the amount of money that they're throwing at this. It's just mind-boggling. Well, I mean, obviously, if you're in the AI business, that's the way you see the world. That's why you're in the business. You think it's going to be huge. They have a reason to see it that way.

16:50I don't know that anybody has a secret crystal ball to know the future. What they're seeing is the exponential demand that things like ChatGPT have seen. They see the potential for what AI can do for lots of different industries and think, okay, that's a definite thing that's going to be huge, but nobody can tell the future. You know, they might be right, they might be wrong.

17:19That's all for today, Wednesday, November 12th. The Journal is a co-production of Spotify and The Wall Street Journal. Additional reporting in this episode from Ben Cohen and Jonathan Weil.

17:33Thanks for listening. See you tomorrow. Thank you.