Dialogue. CoreWeave Business Breakdown, End of the Year, Podcast Changes

(upbeat music) Oh, welcome to the Synopsis, a business and investing podcast for professional investors. Now you'll notice we switched up the intro because finally in 2026, the Synopsis is broadening our horizons a little bit. It is no longer just a speed well-researched podcast, but we will be incorporating topics from Drew Cohen's, you know, it's weird to say Drew Cohen, you're right over there Drew, hey, say hello to the people. - Hey, there you go, perfect. Perfect transition just as we drew it up. But we'll definitely be expanding the topics here. So it won't just be the in-depth research that Speedwell covers as much as I enjoy talking about R.H. Gary, Florentocore. We're gonna get a little wider talking about more, you know, talking about broader investment topics as well as some of the five minute money series that Drew's just put out. But maybe Drew, you want to expand a little bit on direction change. - Yeah, just before anyone freaks out, we're not taking anything away from the podcast. We're still covering all the Speedwell research names. We're still gonna do the company episodes. We're still gonna do updates. We're still gonna do investor interviews. What we wanted to do though was to have a sort of venue where we could talk more broadly about different businesses, different investing concepts without being just really pigeonholed to, you know, the 20 stocks that we've written up. And so on my YouTube channel, which you could find a link to in the show description or just search Drew Cohen money on YouTube, you'll find it there. I'm talking about a lot of different businesses. I just started doing more like business breakdowns on the YouTube channel. So I'm gonna keep that up. I did core, we've already, I did do Lingo a couple weeks ago. I'll give you a little spoiler alert. SpaceX is coming out soon. So I wanted to look up all these fun companies that I didn't want to spend, you know, a month, six weeks doing an in-depth deep dive on, but still wanted to learn something about 'cause I think even if we're not gonna, you know, ever be investors in them or whether we are is kind of irrelevant, it's just more about kind of understanding the market, what's going on, all these different businesses. And maybe you find, you know, 10 gentle opportunities that are related to that. And maybe you understand a different business like, you know, the hyperscalers, even better by understanding someone like CoreWeave and why they exist at all. Maybe you understand Nvidia better. And so all of these sort of learnings, we wanted a place to put in and made sense to just broaden the podcast out. Once again, we're still gonna be covering all of the speed wall names, speed wall research, the business updates, all that. But now you're also getting this other content from the YouTube channel that's gonna be inserted into a lot of our dialogues. It also allows us to kind of have a little bit more casual of a conversation around all these things without just feeling a need to hit all of these different business update points. And so we hope you enjoy this format. We think it's actually gonna be pretty popular because we've been asked a lot about, you know, different stocks and businesses and kind of more general thoughts. And so I think this is the right venue, the right place to do that. You may have noticed I kind of experimented a little bit with the investor briefings for a little while. That was my attempt to broaden us out a little bit, but I think that was in the wrong direction. I talked a little bit about it in the past on that. But don't want to really be that news focus. Still want to give actual insights and information and kind of do something that, you know, frankly, I don't think a lot of other people are doing really well. A lot of times when you do see, you know, very, they'll call them deep dives, but they just don't get really at the crux of sort of the issues of the business, really understanding them. And yes, I'm gonna miss stuff too when I spend only a day looking at the business, but I do think it'll be better than, you know, 95% of the stuff out there. And hopefully it's still gonna be hitting on the key questions with these stocks. And I think you guys will enjoy it all. Yeah, no, it is funny. I do feel like you can kind of in some more efficient amount of time, kind of come down to boil downward of the crux of the questions. And then to actually answer those intelligently, I do think takes a immense amount of time. But I think actually, you know, in a grid and you had such a very compelling YouTube video about Corrie, which I encourage everyone to check out. But I think it was just a great breakdown of, hey, here are the kind of pivotal, you know, aspects of the business we're gonna talk about. And, you know, whether you want to make a call or notice you probably want to do a little more work than what was done in the YouTube video, but I think it really points people in the right direction there. Yeah, yeah. And we'll link to the YouTube video too and the show notes for Corrie. Not right now, I think Thown has a little over 8K views. So Thown was pretty popular with people. And I think that's because a lot of people get a very high level understanding of the business. Maybe you hear about how much debt they have. You hear about, you know, the Nvidia sort of relationship. You hear about, you know, this Microsoft contract and Microsoft 70% of the business. And all this is kind of just these kind of red flags, honestly, if that's all you kind of heard about. And so it was kind of interesting to actually dive into the business. And we're going to talk more about Corrie in a moment here. The other thing I just kind of want to hit on is just kind of how I'm thinking about everything more broadly at this point. We're going to release this end of the year. Memo, if it's not out already by time, this is released. But just broadly speaking, we're keeping all the speed well research stuff. That is very much still for professional investors. We sell that research through a lot of hedge funds, a lot of asset managers. You know, I'm humbled, honestly, by some of the people that are purchasing the research there and that are subscribers. And really thankful for all that and how that's going. At the same time, I did want to still feel like I can talk about all these other companies without feeling a need to spend 100 hours researching on. And so that's what five minute money is. That's a new newsletter I started. You can go to drucoandmoney.com, sign up for it there. That's going to be once a week talking on different business investing concepts. At some point, I might get into a little bit of personal finance stuff too. You know, different kind of decisions like whether or not you should buy or release a home or a car or sort of the financial impact of that. These are really important decisions. And there's, you know, some people that cover the personal finance stuff and all that. It always feels a really kind of cursory though. And there's a lot of kind of decisions you'll make within that better way more than what stocks you end up buying. And so I feel like at some point, it could be appropriate to touch on all that. So that's part of the reason too. Why I wanted this kind of broader venue to talk about all these different finance concepts. And so for now though, it's pretty business and investing focused still. So I think most of the listeners would really appreciate it. So go to DrewCoinMoney.com. You get the five minute money newsletter there. You could go to YouTube and you could subscribe to the YouTube channel. You're listening to the podcast now. And you got a pretty good content diet going on there. - Yeah, there you go. Yeah, he's definitely keeping the new year filled. But no, I really enjoy the pivot to some personal finance topics because I do think that obviously everything that an investor does is kind of analyze the allocation of capital. And when you're running a business, it's pretty straight forward. You're trying to maximize ROIC, maximize shareholder wealth and things of that nature. But to me, it's funny because in life, it's kind of the same thing, right? You have excess capital and how are you going to maximize your ROI on that capital. But then also thinking about emotional, it's not necessarily making always the smartest financial decision arguably the smartest financial decision could be if you're just trying to maximize for wealth would be live on a shoestring budget and deny yourself how all-worldly pleasure is and save a bunch of money. And when your wife wanted to buy a credenza like Warren Buffett, well, it's not a $3,000 credenza because it's 1970 and 2020. This $3,000 is going to be worth $30 million. So you just spent $30 million kind of credenza logic, which again, some people could be down for. I personally try to find a happy medium between the Buffett. If you're doing well in life and your wife wants to buy a credenza, just buy or the damn thing for the love of God. Don't give her the speed about compounding interest all the time. But enough about that, I'm excited for those topics. So let's get into a little bit about core weave. So we've got a lot of depth on this podcast. Overall, we've been doing the-- we can talk hardwood flooring sales, Gary Friedman breaking into the luxury. And now we're going deep into the AI world. And one thing is funny, which I'm excited about, the SpaceX. But it's-- how did I put it? There's a lot of-- and I don't want to say value investing. I want to say quality investors who are always looking, well, the earnings and this and that. And sometimes you see kind of the AI boom that's happening and we've joked about it. Or you look at a valuation like SpaceX. And it's very easy to just be like, oh, these charlatan idiots. They don't know what they're investing in. And look at this, oh, it's 100 times revenue. How could they see this? And then obviously, that's like, I think, a first level look at. And then you do have to have to give credit because, of course, there's highly intelligent people on the other sides of these investments, I would say. And it's kind of like, well, what are they seeing? Because clearly, I don't think it's not fair to just say everyone doesn't get it, right? Well, what do we not get sometimes? Yeah, I think the problem with all of this is-- and I fall in prey to it, too-- is the loudest voice is usually the stupidest voice. And if you're listening to that as sort of the investment thesis of a company like SpaceX or Core Weaver do Lingo, you get a very warps sort of view of it. And so something I wanted to do was kind of just start from, like, first principles, clean slate looking at all these businesses. Forget about, you know, oh, it's, you know, a growth, a Momo stock, all of that. And just like, what would I actually think about this business from a clean sheet of paper? And it's not like, you know, there's nothing to those perceptions where, you know, yes, Core Weaver does have a lot of debt. Yes, there are a lot of risks that we'll talk about and all that. And we talk about in the video, all of that's there. But is there not something to the business model that maybe you're missing? Is there not something to what they're doing that, you know, maybe would surprise you? And when I looked into Core Weaver, what kind of surprised me was that they actually were more efficient currently than, you know, Amazon AWS was than Google was than Microsoft was in a lot of testing on AI models. And I thought that that was pretty interesting. And then the question is, how is that possible? Given how much bigger these big tech companies are, how much more resources they are, how much longer they've been in the cloud business. And kind of breaking down that answer, we'll touch a little bit on it in this podcast today. But that is also just kind of like getting that next level to really trying to understand something, rather than just like dismiss it at a right, at a hand, which I think is a skill you need to learn as an investor. Of course, you need to, you know, pick sort of your battles of what's actually worth digging into and getting more into the weeds on and where to actually allocate your time. But if you're always just like dismissing every potential, you know, business stock out of hand, then, you know, the universe you're looking at is going to be very small. And maybe that's not the worst problem in the world. But you do want to kind of widen it to as many potential opportunities as possible. And I know the way I'm framing it, you maybe you're thinking, Corwee, is a potential opportunity or not. Like I'm not thinking about this from the perspective of whether or not you should invest in it 'cause I don't give recommendations. It's more about whether or not it's, you know, even worth looking at a stock like this, whether or not it's worth spending that couple hours to dig into it. And I think I've decided a lot of times when there is a stock that does have kind of a lot of attention and maybe it's kind of contentious that I am going to lean into looking into the more, especially on the YouTube channel. 'Cause I think that would be a really interesting topic. And again, 'cause the debates around these, they're very superficial. There's people on the one side who just say, "Yes, you know, this is going to be great." And they're just, you know, extrapolating out high growth rates forever. And they don't really understand the capital structure, the debt stack. They don't understand what's going on with, you know, the convertibles and all sorts of different issues. And then the cynics on the other side just don't even bother looking at it. And so I think there's a place to kind of be in the middle. - Yeah, I agree. I think to have any good understanding of a business, whether you want to take the positive or negative, you definitely, I think need to have an in-depth understanding of the steel man of the other side, right? So I think that you did a really great job of kind of steel manning the bowls of core weave and then also, you know, giving credence to perhaps the bears on it. And again, I think there's intelligent people on both sides of the discussion here. So I want to get into a little bit about, you know, you did a lot of great work in YouTube videos so we don't need to flesh everything out. But I want to get a little bit into core weave. But before we get into that, I would say long story short, it was giving me flashbacks to Oracle here, which was you have essentially a data center company and again, they're kind of unique in what they do and we'll get into that. But it was kind of a big announcement of, with opening eye on Oracle, that they were going to have this huge backlog, you know, all this commitment to them building out more data centers. And same with core weave, right? They have one of the largest backlogs. You know, $55 billion in backlog revenue for building out these GPU AI specific data centers. And initially the market, you know, you looked at Oracle, Larry Allison, was, you know, the wealthiest man in the world for briefly took over you on Musk. And the market said, wow, what an amazing deal. This is incredible. And I remember on a dialogue, you and I were kind of discussing like the actual economics of that. It's like, wow, so he's going to bet the entirety of this company building out the backlog. That open AI, you know, not clear that they even have the money. Like they're kind of projecting that they'll have the money to pay for this. And the market rewarded it handsomely. And then all of a sudden, you know, now a couple of months later, and the market's kind of backtracking a little bit. And then I think a similar narrative took hold with core weave, which was very bullish. Oh my God, they're going to be this great data center company. And then it's like, wait a minute, they've got to put a lot of capex. So I don't know, what do you think about that narrative shift in general? Like I just as an investor, I find it kind of funny that, you know, you could be, something can be known. And then the market just like decides to react to it like a month later, which is why I wouldn't try to play options or anything 'cause it's too tough. - Yeah, I think very much is a case, though, a lot of the narrative following the stock price. 'Cause right now this is $75 stock. It was up to like 180. It had almost a hundred billion dollar valuation. And then when something has a valuation like that and is going up that much that quickly, everyone's always asking why. And so you ask a question, you're going to get an answer. And so the answer people will give is, oh, this is, you know, the next generation, you know, cloud service company. They're going to be able to do stuff that, you know, others aren't. They're going to be a secular winner, blah, blah, blah, blah, blah. And that's just mostly, I think, a by prior to the fact that there was a lot of people making money. And then they have to defend the fact that they weren't just getting lucky by making a lot of money very quickly. 'Cause, you know, from like IPO at, you know, 40 bucks, you're talking about a matter of from March to June, it going from $40 to $180 a share. So that's a big gain. And people don't like to believe that, oh, you know, you just got lucky by going up that much. They like to believe that there's intelligence and their decisions they make. And so I think that that's a factor of why all of a sudden, all these narratives come out as a byproduct of these stock price movements. And then, you know, the stock price drops, which it has, you know, now it's down, cut down about 60% from the peaks. And so now it's a $40 billion, $38 billion market cap company. And now, okay, new narratives coming out, oh, you know, maybe it's not going to actually be that easy. Okay, maybe we do need to be worried about the debt. Maybe the data centers are going to take longer than we expect to build. And so there's all of these different kind of factors, kind of floating around there that kind of show up only after the stock price drops. >> Yeah, and again, this is the topic for another discussion. But when you have companies whose prices are so predicated on future cash flow, future execution, and the business models, and this clear, those perception shifts are everything. You know, you know, the market today believes, okay, AI is going to be, you know, this remarkable thing. And the infrastructure we couldn't build enough, there's no way we're going to need more than anyone can anticipate. And then when that perception goes, oh, wait a minute, maybe we actually are over building our way. Maybe these GPUs don't last as long as we think, or maybe the economics aren't. And that whole perception shifts, it can happen very quickly. And that's why it's always, you know, again, we talk about timing. I look at some of the, you know, housing companies we cover, you know, our age, Florida core, Walker Dunlop, Dreamfinder Homes. And it's kind of, you want to essentially own those companies when you think the housing market's going to turn or rates come down, or there's an increase in housing turnover. But as soon as that leaks into the market, there's even a perception of that. It's probably too late, right? 'Cause now that's going to be baked into the price. So you either need to own it through it, or I don't know, you can get lucky in time. And maybe some people have that ability. I don't think they do, but that's a difficulty. - It is a difficult thing because, you know, you could look at a stock like our age. And it's probably had like, you know, three or four, 50% plus run ups every time there's like a feeling that rates would drop a little bit more. And then it reacts very quickly. And then people, you know, kind of sell the news thereafter. And so these are very hard games to try to play and to try to get right. And that's, you know, very much happening in the AI sector right now too, is there's a lot of people kind of trying to ride it as long as they can ride it, hoping they'll get out before other people. And that's of course not how we would invest. And that's not even what I would really consider investing to be, because investing for me is, you know, the process of figuring out the probability of profits or cash flows. And once you get sort of a probabilistic distribution there, you're making money off of the business, making money, not off of selling it to someone else. And so, yeah, all of that is floating around there for sure. And we're going to get into Cory. But I think these AI narratives have been the most interesting to watch, because I just remember, and again, I think Google's another great example of like Sam Aldman goes on Ben Thompson and I think this is right after Chatchee PT and he goes, I would hate to be a lethargic search giant right now. And then I'm kind of listening like, oh man, I'm a little afraid I would hate to be a lethargic search giant too. And then, you know, you see investor sentiment just turn on Google, the search engine, that's just been pumping cash. Like one of the greatest monopolies ever existed in humanity. Everyone's like, that story's done, you know, not done, but it's a melting ice cube now. It's, you know, they're going to cannibalize their revenue. They're done. And now all of the sudden, it's like just kidding. Like Sam Aldman sending out, you know, red level memos, like red alert, border and trouble. Google's taking us over. And now all of the sudden, everyone's like, yeah, Google. I mean, who could have ever thought they weren't going to figure this out? And it's just, I don't know, it's just funny how really smart people can, I don't know, just flip on a dot, right? Like you have all this data and then all of the sudden, one new data point comes in and they just erase everything they've had. And I get it, it's hard at the moment because I, you know, I was listening to that top set too. And I was like, man, that is kind of a scary sentiment. But it changes. Well, people take the current kind of latest data point and they just extrapolate it out indefinitely. And that's just something we kind of tend to do very automatically because it's very hard for us to kind of see a data point, kind of in a broader time series as potentially being an anomaly. And so instead, what ends up happening is we just continue to extrapolate. And that, by the way, is why a lot of stocks go down so much on earnings is there may be a little fluctuation in the growth rate and then investors start extrapolating that out kind of indefinitely. And that's what happened with Meta, by the way, when their growth started fading, even though they were growing, you know, double digits every year, every quarter until 2022, people just assume once the growth slowed, that was going to be it forever, that those weren't just transient sort of factors because it is a harder thing to kind of take a bet on that and to assume a reflection in growth and all that. So that's kind of my take on that. - Yeah. And, you know, we could go on for hours about this now that we've opened up this synopsis. We could just get, we could lose about any topic. But let's get into what we came here to talk about, which is, and again, not going in depth about CoreWeave, it's more going to be kind of an analysis of what you've already spoken about. But just to give some listeners, maybe they didn't watch the YouTube video yet, why don't you give us 30 seconds about what is CoreWeave doing? How is it kind of positioned against the hyper scalers? And how does it even have a niche here when you have Amazon and Google and Microsoft pumping tens of hundreds of millions of dollars into this space? How do they have a foot to stand on? - Yeah, so if you want more details, definitely check out the YouTube video. But in short, they basically are one of the fastest growing companies ever. In 2022, they had just $15 million in revenue. And now just three years later, they have $4.3 billion. So that's a 286 times increase in revenue. Right now, they have a revenue backlog of $55 billion. And so their customers though, are, you know, you're talking about Microsoft. They're about 70% right now, at least as of the last 10K, they now have deals with OpenAI with Meta. So, you know, these are pretty much like tier one clients. You know, OpenAI may be an exception 'cause of, you know, exactly unclear how they're gonna fund a lot of this. But Microsoft, you know, pretty solid company and that's the vast majority of their revenues. And, you know, Meta too. And so the question is, why is Microsoft who has, you know, Azure, who has a hyper scaler, a cloud service? Why are they using CoreWeave at all? And it comes back to kind of a mix of luck and timing, plus a strategic benefit in video head and kind of a change in the business model. And so what happened was they started basically as 2018 as Ethereum miners. And they had a bunch of GPUs. They would go and mine Ethereum or rent out the GPUs. So other people can mine Ethereum. And that was basically the business model. Ethereum crashes 90% and now they're trying to look for these other sort of workloads for their GPUs that they already have. Now, a byproduct to the fact that they were focused on crypto mining means that they were a GPU first data center. That is different than the legacy data centers that are CPU first. One other sort of change in their business model was they would use contracts called take or pay contracts. And so what that basically means is that a customer is gonna be paying them regardless of the usage of the service. It also means that the customer kind of, you could think of it like they get like a rack of servers. They get dedicated compute that no one else can have. And it's gonna be just theirs and they get access to that. And since it's dedicated within a rack within a server, however, whatever word you want to kind of use there, it means that they could use it for whatever they want. And all of the work done on that server is gonna be dedicated to the customer. It turned out that this was actually really important when you were training an AI model. Because in contrast with a hyperscaler, they are trying to optimize for flexibility. They want as much of all of their compute used as often as possible basically because they're charging for it, they're renting it out. But the way that they charge for it is they only charge on a use basis. And so they're not making money unless someone is using the compute so they want people to use all of the compute. Core weave in contrast doesn't care whether or not you use all of the compute because you're gonna be paying for it either way. Now the difference in what this means in the architectures is that you may have people running different workloads on the same server within AWS, within Azure. And then also, you know, the fact that there's more kind of CPU first sort of workloads which aren't as good for AI. And so those two kind of factors means that if you are running or trying to train an AI model on a traditional cloud service, your efficiency is actually lower. They have this term called MFU which is model flops utilization. You don't worry about what it means. Basically the way you calculate this, it's usage divided by capacity. And so the more that you're actually using of compute versus how much you could theoretically use within, you know, we'll call it a server, then that's gonna be a higher MFU, so more efficiency. Core weave actually has about 10 to 20% higher efficiency, higher MFU than these hyperscalers. And that is a byproduct of the fact that their business model's different and they let someone rent out kind of the entire compute and it's dedicated to them and they could run just their workload on it. And contrast again to the hyperscalers, multiple workloads on it, it's less efficient when you're training this AI model. So the observed performance for most models is anywhere from 25 to 45% and then it's 10 to 20% higher on core weave. And that matters when you're trying to, you know, train a model and be the first to have a leading edge AI model. And so all of that is a reason why even Microsoft is contracting with core weave. And my guess is they're probably reselling a good amount if not all of that to open AI or to other customers because they themselves just didn't have that capacity of a GPU first data center that was set up the way that core weave was. - And so that's kind of the thing that core weave basically does that's different. And so they focus on efficiency whereas the hyperscalers are focusing on flexibility to improve their, you know, overall utilization 'cause that's how they charge. - And, you know, one thing, and again, I think we didn't want to get too into the AI's but I did just have a question watching the video. What is actually like the benefit to the AI companies that core weave gives is it just that they can train it in a more cost-effective way, is it faster? Like what does that metric actually lead to as an outcome? - Yeah, both. Like basically you're able to train a better model with higher performance versus something that could be much slower and take much longer because these, you know, performance differences when you're training a model it could take weeks and weeks longer if not even longer than that. - Yeah, so again, it's a funny story. It's kind of one, you know, crypto miner kind of, I don't want to say, I don't want to use the word "lucked in", maybe that's harsh. I'm sure they definitely pivoted. - In some of the efforts, yeah, you use that. I think it's sometimes it's good to be, I think what is it? Opportunity is when luck meets skill, whatever it is. So they took advantage of an opportunity they saw and they did execute on it to an extent that, you know, there's a lot of crypto miners out there who are not as successful at quarried as this. So you get to give them credit there. So you have this large backlog and one of the most interesting things that you touched on is Microsoft is their number one customer is about 70% of their current revenue. I don't know if that includes 70% of their backlog or not, but it's 70% of their existing revenue. - Current revenue. - Current revenue. So again, we don't know exactly if they don't disclose the backlog to an extent of who's man, who that is. - No, I'm sure you could kind of do some sort of math around it 'cause they do announce some of the contract sizes, but I don't know exactly how large the contracts are and then in the future, you know, it could always change because the existing contracts can be amended and all that. - So anyway, so you have Microsoft who obviously is competing with them on some, not in some degree, is competing with them on a cloud-based solution, right? They have a cloud-based solution and you make it very clear that obviously this is almost like a stopgap for Microsoft and how all the hyperscalers Microsoft Amazon Google Meta are going to create a lot of GPU first capacity. So could we to meet like immediately? It sounds like they're on borrowed time, you know, 'cause it's then it's kind of like, well, okay, you're gonna have these hyperscalers that are gonna be able to invest a lot more and then, you know, but again, this is where this kind of very interesting relationship with the video comes where Nvidia kind of wants a horse in this race and is really, you know, seeding them in a very, how do I put it, a very close way. It's almost like they're a subsidiary of the video in a way because I think that's you alluded to in a video and that's a main competitive advantage that Jensen Wong likes them, right? - Yeah, I mean, so Nvidia is an investor in CoreWeave. They're actually a customer of CoreWeave 'cause Nvidia has their own small cloud data center offering and then in addition to that, they're also obviously supplier and they're kind of a preferred customer of Nvidia's because when Nvidia has a new chip, they want to prop up CoreWeave because they want a more diverse customer supply base. But they also don't want to keep strengthening competitors that, you know, particularly you can imagine in Google and Amazon that are designing their own chips that are trying to come after them and, you know, cut them out eventually. And so not that that's a totally different discussion right there in and of itself, but from Nvidia's standpoint, yes, they liked the idea of an agnostic competitor that does not have any ambitions in chip design whatsoever. They also help them diversify out their customer base and so that's why Nvidia's been helping them a lot, selling them a lot of chips, selling them, you know, the latest chips kind of first and that's been a real benefit to CoreWeave and there's no reason really why that would change kind of in the future. Google's still gonna always have their TPUs, Amazon's gonna have their chips Microsoft. Their chips aren't like front and center design but they're probably still, you know, they still want to diversify out from just having a couple kind of large customers and so all of that is, you know, I don't see why that wouldn't kind of continue and be a benefit for them kind of into the future and then there's also the fact that, you know, once the data, there's also the fact that, you know, capacity is very constrained and so having the capacity at this point in time is very much, you know, a benefit and all of that. And everyone is kind of building a lot, right, you know, all of these big tech companies, you know, have a CapEx budget so basically a hundred billion dollars and most of that is going to new data centers, power, compute, all of that and so there's definitely a fear like that at some point this is gonna be too much and there's gonna be oversupply and if demand doesn't meet that supply, then that's gonna be a very interesting question as to what happens with pricing and what happens with all of these contracts too and what kind of position core weaves in because if you are a Google, yes, you're gonna use your own compute for something else, medic and find another purpose. Microsoft can find another purpose. They could resell it to whereas core weaves, not less clear, less clear and so I agree with you, it does seem like it is kind of this stopgap measure but you know how long that lasts and also whether or not they may always be necessary if Nvidia keeps kind of capping how much GPUs are giving these other players and keeps propping up for weaves. That's kind of a key question. - Yeah and again, it makes so much intuitive sense for Nvidia to kind of have this champion who isn't competing with him directly, right? Because of course Amazon and Google, I mean it's a no secret Nvidia is extracting an immense amount of value. You look at their gross margins continue to increase and their chip prices continue to increase and so they're extracting a lot of consumer surplus you would say from the big scalers, the big hyperscalers as the AI boom is happening right now and so you could see how Google and Amazon and again Microsoft, I mean their efforts are a little less pronounced but I'm sure that they're not resting on their laurels and just going to be beholden the Nvidia for the rest of their lives because that's a lot of supplier risk on their ends. So you definitely can see Nvidia's incentive to keep a third person who is highly dependent on them and to the extent that Nvidia can outpace Amazon and Google and their GPU development giving this kind of AI ecosystem another hyperscaler who perhaps continues or kind of a third cloud-based solution who is fully reliant and fully kind of all in on the Nvidia stack is a pretty attractive place to be. Now one thing I want to kind of say that's interesting before we talk about the capital structure and not that we usually get to into this because most of the companies that we talk about other than R.H. I wouldn't say they're bankruptcy risk but yet Gary Friedman took a more depth than we would have like but most of the companies we talk about capital structure is not as relevant because there are nowhere near that being kind of an investment consideration but you know Corrie that's definitely a consideration that I want to get to but I want to talk a little bit about just like the ecosystem that's happening right now because you have all of these backwards dealing so it's kind of like open AI has committed and again a lot of the Microsoft spend that is kind of going to Corrie is directly related to their service agreement with open AI and open AI directly has you know I think I just looked up about $12 billion of backlog with Corrie so you have open AI you know Nvidia is investing in an open AI which is creating you know contracts with Corrie which is a Nvidia subsidiary and it's just it's kind of like there's like churning capital through the same system but of course all of the spend is kind of predicated on the fact that you know open AI and all of these individuals who are investing heavily into the AI ecosystem are ultimately gonna sell us the consumer and make money somehow right far into the future because open AI clearly has already you know 100 billion plus of commitments without the revenue so what do you think about that ecosystem to me like that like just talking about Corrie is embedded in this I mean it seems very like I don't know all dependent on one another and I mean how can that go wrong? I mean it sounds like inevitably it could you know it seems very intuitively could go wrong but I mean what do you think about that? - Yeah I mean I think that open AI is probably the linchpin in all of this because they're the ones with the most kind of contracts all over the place and the least sort of current business cash flows that can support all that I think their revenues right now around 15 billion and that's revenue and you know their commitments run in the hundreds of billions right and so where that shortfall is gonna come from is you know very very questionable and what it sort of is the knockoff effect of all that if they are not able to pay some provider then I don't know what exactly happened. - I know it's great like I don't even think open AI has any like I'm a chat GPT subscriber but if they raise it from 20 to 40 I would go to Gemini like immediately like I don't see enough of a distinction - I actually X AI in chat GPT like to pay a premium service I don't you know like that's pretty weird. - I thought you know chat GPT was pretty good versus Gemini and actually as of late I have been using Gemini more and I just canceled chat GPT premium and so this was also the first time when I went to cancel it where they offered me a month for you to stay on so I'll take it in a month and then cancel but that's kind of I don't know I think that kind of signals a little bit of what's happening and how quick this is I still like the UI of chat GPT much better than Gemini but you know they're always just one model away from all this changing so this is again a very precarious position for all of them and that's why Ben Thompson was always hitting on why open AI needed an advertising product because if they had one then they would be able to monetize their free users better which would be important to helping them you know support their build out of data centers which would allow everyone to get kind of the faster better models but would also kind of support their growth whereas right now you know they're pretty small pair penetration and they just don't have enough revenues to really you know support the AI for you know hundreds of millions of people using it the way Google could kind of subsidize it in the interim but then Google also has advertising they could insert in when they're ready to and so yeah you're asking what you know how this all falls apart I don't know how this all falls apart there's some path where you know for somehow it just all kind of works out I don't know how high of a probability that is but it does seem you know a very precarious position for open AI in particular just given their commitments versus their kind of current revenues and then also you know the product isn't that dominant you know it may be number one with a lot of consumers but it's being threatened for some - Well listen you know Michael Berry you know trying to recruit did you read all this he's finally shorting the video again you know he's the one that's gonna work you know - Yeah he keeps going back to that well huh - Yeah he's going back to that well he's like there's another big short you know man with a hammer syndrome but we'll see he's very public about trying to you know it's funny I was just reading how Michael Berry and again for those who don't know he was you know kind of the famed investor and too big to fail who kind of saw the housing market and made I think what a billion dollars off the short of now more no one knew that - Oh wow yeah I know we're near that all right whatever - Didn't make that much money but he made a lot of money - Sorry quick red herring I have a little bone to pick with Michael Lewis in his book The Big Short who profiled all of these investors that shorted housing market but didn't include John Paulson who was the one that made by far more money than all of them put together so that's I needed someplace - Maybe that's what I was thinking - I think Paulson I mean he hit he hit - Yeah Paulson made I don't know a couple billion dollars - Yeah shorting anyway he made a lot of money let's say that I don't know what the exact figure is but he did - Paulson lost it all in the substance point he had a good call in gold and then he was just calling also for you know another crisis thereafter for a long time and then he got into like Puerto Rican hotels or something - There you go I digress I just thought that was a funny thing and oh one thing I was going to say about Michael Berry is I just read that he has I think a sub stack and his latest numbers he has like 100,000 subscribers paying him like 20 bucks a month or something so it's like he's now I think making more money on the sub stack than as like a hedge fund guy I would imagine because he closed his hedge fund so he's now a sub stack writer not to discredit sub stack writers you know I've discredit them you've got that if you want but no leave it in let the people know where you stand yeah yeah yeah that's discredit but you know anyway so what my point about all this was you know what was your point but I'm going to get there it's a little circuitous but I'm going to get there but this this kind of behavior and it's actually in John Malone who was you know famously you know ran TCI cable which was one of the original cable operators back and then you know I think started as or late as the 70s and really ran all the way up until they inevitably sold it to AT&T but and he has a great book born to be wired in which he talks about his comparison to what the cable companies were doing and what kind of the AI companies were doing because the book came out about a year ago and this was even before kind of some of the latest deals with some of those big opening ideals were trickling out and he had some commentary about it which was back in the 90s 2000s that you know this was a similar behavior which was the cable companies would take a stake in one of their major suppliers and prior to announcing you know a big order and right four or five cable companies would announce you know a multi-billion dollar order and before they announced that they would you know invest in that company get convertibles or get an equity stake because inevitably when the market saw that you know the supplier had such a big backlog there was a jump in the stock price because there was you know kind of a lot of vision into the you know increasing cash of that supplier and essentially the cable companies took their equity investment and the differential between what they invested in and what the company and everly popped to helped offset the supply cost and so this has been happening right so and that was a very similar situation which this was happening quite a bit and it was very messy and intertwined and again that more or less worked out I mean the cable companies you know did pretty well and these suppliers didn't go bankrupt or anything of that nature and they made these large deals and they increased manufacturing still not anywhere to the scale of this but John Malone's major point was the demand for cable and the usage of bandwidth through these wires was increasing right and they got that bet right which was more people watching more tv they wanted more channels the internet came along this infrastructure was highly utilized it was utilized effectively and it was utilized for decades and so that's kind of I think the similar pet to AI here and then you have on the other hand all of the telecom and internet build out during the tech bubble and they built way too much capacity and that resulted in a lot of the lines being dormant for a long time there's a lot of kind of private equity money invest in a lot of those that kind of went bankrupt and so that's kind of on the the other end of it yes we eventually used all of that capacity but there was you know a multi-year period where we weren't in the investors and in those businesses got burned and so you get both yeah that sure and again I don't want to like you know there's a more nuanced history there which is yes some of the cable companies did well and obviously very famously a lot of telecom companies did not do well the dot com bubble and that was a bus so I'm not going to say it all worked out but in those instances there were successful maneuvering in terms of you know kind of the end user investing in the supplier and then making money off of that investment when they announced the backlog so this isn't it's kind of a repeat of history but inevitably what worked out which the end usage was there and so the big question for everybody in AI right now is again it's obviously amazing technology and it would be hard for anyone to sit here and say this isn't going to be revolutionary and it isn't going to be a big thing the question is is how much money is there to be returned right like there's so much money flowing out of this there's so much money opening it has to make and is core we're going to be able to inevitably you know it's kind of like just a to me a side player of this broader bet on AI and I think that if this bet that all of these hyperscalers work out the core weave is going to kind of be on that wave I don't know to what extent core weave survives to meet like if this thing doesn't work out core weave is like one of the first to go and I think that's because they're capital structure yeah I think that's right and I think you know it's just kind of like capital returns you know the point Edward Chancellor or I guess he's the writer of it and it's the marathon letters but they talk just a lot about following kind of supply because demand is much harder to estimate and if you overshoot supply then it really doesn't matter what demand as prices are going to still end up falling anyway and I think that dynamic very much exists with data centers if you have a data center built that cost billions of dollars you're not going to let it sit dormit especially if the chips are only going to be good for five years either way and so you're going to rather reduce your pricing to get some sort of utilization on it rather than let it sit dormit and that's going to result in pricing pressure and then you're going to have someone like core weave who may have contractual pricing but I don't know you know maybe you find out that the you know they try to weasel out of some of those contracts even someone like Microsoft might if there's other opportunities there and they don't want to blow you know 40 billion dollars on compute that they can no longer really properly utilize or could get cheaper elsewhere and so I think it's also very precarious when you kind of are reliant on funds to come in in order to just support your existing business you know they have a billion dollars of interest expense against 150 million dollars of ebit currently and let's dive I mean we're getting the capital structure so let's dive into that well yes let's dive into that but the the point I was making there is that you could have these people kind of just not pay them for a little while to dispute the contracts and then you're in an awful negotiating position at that point because you're going to go bankrupt basically and so if the funding environment falls out at that point you're really beholden to your customers and I feel like they could just negotiate over them at that point you know they may not be exactly how Microsoft has always behavior historically but you know no one wants to lose you know tens of billions of dollars if that's the situation they're pushed in and you know they could maybe feel they are right about some of the the points on the contracts we don't really know what's in there and if they take too long to build some of the status centers maybe they do lose the customers either way okay fine sorry I was trying to interject on my capital structure you shut me down it's all right listen you know what is the host therefore except they're actually the conversation it's fine I get it the word you know I might feel you know the host also is it supposed to like machine gun like mid-sentence into like a different topic supposed to have a little bit of poise this is my hosting style if I want to just jump right in there I'm going to jump in you know this is my yeah this is this is like too much TikTok going on or something like I'm bored swipe next yeah yeah all right point made let's get into the exciting stuff okay I know because to me it's and again I'll preface this with we're not you know distressed that investors we usually don't invest in companies that have you know an immense amount of leverage but this is an interesting one which is you know again you you talk about 180 million dollars and an EBIT against a billion dollar interest payment 150 an EBIT 150 an EBIT okay so they have to have to borrow four times there what is that what's the math on that five times I'm gonna be quiet we're gonna let the people see your skills doing live math whatever so you know they're gonna they're gonna have to borrow they have to they have to borrow cut that not hunting that so they have to borrow multiples of their EBIT just to break even on their interest payment let alone the additional capital expenditures they have to make the service this 55 billion dollars in backlog so can we just take us through one like how much debt do they currently have like are they do they need to raise more are they like almost that a runway like what's going on here yes yes yes so yes they have a lot of debt 22 billion dollars of debt and if you're including you know these operating leases the way they're financing a good portion of this in in addition to just long-term debt is that it's actually doing financing on the the GPUs themselves and so that's one of the way there there's kind of financing a portion of that I think that the thing to keep in mind those that and this is the part where it does get a little interesting they have a 55 billion dollar you know backlog and so even those you know saying yeah you could weasel out of them you know these contracts if they really want to or at least try to stall payment on them and even if they get in trouble for doing that core we've could so that's like a bearish take if everything's falling apart if all of the music keeps going and playing they're you know can go to a lender and say we have a 55 billion dollar you know revenue backlog our customers are meta open AI and Microsoft can we you know use this to help finance a data center actually then build out the product that will allow us to earn that revenue and so it is a very different position to be in than not having you know actual revenues that are contracted and so before they're actually laying out the catbacks they know that they're going to have a return on it at least theoretically and so it is kind of an interesting position to be in that it's not the same as just being like they have to raise money they have no idea who their next customers are going to be for for they have no idea where they're going to earn this money from that's not the position they're in they just really have to execute on it that's kind of the the big risk right now is you're building out a lot of data centers to capture a lot of revenue data centers are physical infrastructure you know there's power you're dealing with regulation you're dealing with infrastructure you're dealing with water you're dealing with you know putting this facility near people who may object to it all of that means that this could be delayed and if it's delayed then you're not collecting your revenue and then you do still have the financing portion of that continuing to bleed out money and so that to me is kind of one of the bigger sort of things to think about there is that dynamic and how it all plays out especially as you're growing very quickly you have to manage a lot of these projects all over the place I think they have you know like 30-ish 35 something like that data centers right now and they're going to need a lot a lot more to capture this revenue opportunity and so that's something we're thinking about and also when they are doing some of these financing a lot of it is you know convertible debt and stuff and so even if you are an equity holder a good portion of equity if this works out it's going to be converted away to the debt holders and so that's just a couple comments there but right now as it stands yeah the balance sheet it's atrocious to look at the balance sheet is atrocious to look at that's what everyone wants to hear well we're only doing muted takes on this you know very balanced yeah very balanced to say it's kind of one of those things where for me looking at this company and you know obviously I think financial modeling has its limits but to me I would want a really detailed like all right here's the interest payments what's the kind of I don't know roll over on this debt when do they have to raise more debt is the capital market's going to be open during that a lot of questions that again are really just tied up to the the fervor and the optimism around the AI boom which again can last for a long time in my you know everyone might be right in my last perpetuity or you know it contracts and they're kind of stuck I mean what do you think the company and I think you highlight a great risk in terms of you have to go out and build all these which isn't an easy thing to do right and again that revenue is contingent on them having the data centers to be able to do that and so that's that's like an operational risk which I which I think is a great point but I mean to what extent can this company whether any type of downturn I mean you did talk about before they kind of were building this out through their existing through their existing data centers that they were earning you know a high teens operating margins so theoretically if everything went to you know hell in a hand basket could they theoretically shut down use their existing capacity maybe survive for a little bit I mean or the debt payments is a little too significant at this point yeah it's a tricky thing to answer I would imagine that they need to keep it all going and they need to actually build these out to earn those revenues because if they're trying to rationalize and like conquer down with what they currently have just the amount of debt that they have it it's not just to support the current business it's for a lot of growth opportunities that need a materialize and so that would be kind of my first take there whether or not you know it's possible for them to like sell some of the GPUs or something like that or sell some of the other assets they've been purchasing and I don't know we didn't do that kind of analysis I just have a lot of I just have a lot of anxiety for the CEO I just feel I feel the pain he must go through it I don't know maybe or maybe he's fine but to me I would just be like oh man we got to make this work I mean you're living on borrowed time I mean everything's got to go pretty well but again we're doing we're doing the bear take I'll explain a little bit how this could go well because theoretically they if they can execute on this and again the bulls would say hey they have this structure they have the unique positioning in the market then the video wants them to exist I do think a lot of AI companies open AI for example I think you made a great point you know why we're going to keep giving Google money why we're going to keep giving you know Amazon money who is probably going to be competing with us or you know keep giving I guess Amazon's a less of a threat to open AI but I could see how AI companies want a more you know even if you're an AI yeah even if you're an AI company you just don't want to be beholden to anyone that may get into having their own AI models and all that and Amazon may be behind but they still are doing that research right and so I mean to the extent core we've I mean what what do you think of the upside here is that they are literally that you know the fourth hyperscaler I mean yes that's that right yeah for I know you know how big does that look like I so hard to tell so hard to know I mean you know people put a trillion dollar sometimes plus valuation on AWS maybe a one and a half trillion dollar valuation and that took them you know 15-ish years to get there so I don't know that's just throwing a number out there that's not a very good analysis so and you look at the other hyperscalers I mean they were in a very unique position to fund this entire build out on their own dime on their own volition you know they could weather any downturns I mean they all had very high cash flow generative businesses that were able to put the upfront cost into these data centers which is kind of the whole point which is right startups struggle to make these upfront costs where the big hyperscalers didn't have that that was the whole point and the other thing that's really kind of worth thinking about too is that as we were talking about the hyperscalers are optimized for efficiency and so when you're optimizing for efficiency that means you're utilizing more of your capacity that means you're charging for more of it in contrast that's not how core we've set up they're optimizing for efficiency and they're hoping they could kind of get that you know same or higher premium pricing by optimizing for efficiency I don't know though whether or not that's gonna carry that same sort of margin it's kind of a volume sort of pricing trade off there and I don't know that exactly that the the less volume they're going to be able to kind of get from that is always going to be made up for pricing especially if you're talking about an environment with excess capacity I think pricing in an environment with excess capacity everyone is getting hurt and what people are building a lot of more now are these GPU data centers and so you know it's just really tough because you know what kind of going back to the capital's returns thing what ends up happening in capital cycles is a lot of money floods into an industry the industry builds out of a lot of supply and maybe for a while the the demand is still higher than the supplier meeting the supply but at some point there's enough money flooding into the industry that the supply exceeds the demand and it's at that point that pricing comes down and when pricing comes down a lot of these people that were the builders of the supply start going bankrupt or at least they're very severely pressured and who has the weakest capital structure in the link of all of these cloud providers well maybe there's others but you know the person with 22 billion in debt isn't in that strong of a position well and again one thing we talk about the infrastructure which were you know again in the kind of in the broadband telecom boom right those undersea cables I mean they had quite a bit of a useful life to them rather than you know we talked about these GPUs and we've hit on this point before what is the useful life so three years as a four years I mean there's a data if a data center is built in 1990 you know rest assured there's no utilization that anyone can get out of that at that point because the compute is so diminished that I mean no one can run any loads on it so yeah to what extent you know do they have to reinvent these you know is there investment done after they make this no they got to I don't know how much frequently they're going to have to turn over all the GPUs and that's a big question mark on on all of this AI spend that neither drew an eye or equipped that to answer and I think a lot of experts are questioning that as well one final point kind of on on core weave you made a comparison and we don't really talk about it's only YouTube video but in the letter you did which is actually arrow cap which is an airplane air cap air cap air cap all right by air cap which is like an airplane leasing company I have no idea how it relates to AI drew take us through that I don't know why do you think there's similar businesses or what are we learning from that so air cap is basically a lesser of planes and so if you think about what core weave is basically in the business of doing they're basically in the business of buying GPUs and renting them out and they're financing a lot of that with that and when you think about air cap it's kind of the same business like they're a buyer of all these planes and they finance a lot of the purchases of them and then they lease them back to the customers because the customers themselves don't want to either get into that business or they don't want that on their capital structure and so if you think about it you know the aircraft manufacturer you know basically wants to sell the planes and the airlines don't want to pay up front pricing for it and so air cap stands in between which is very similar to sort of a specialized bank and so in video they want to sell GPUs but they don't want to be the financier of the GPUs to the customers and the customers don't necessarily or at least the very end customers the AI models themselves don't want to be paying up front prices for all this they they want to be able to sort of you know lease them basically as they use them and so that's basically the business that core weave is in core weave stands in between them very similar to sort of a specialized bank and so a lot of that too comes down to the fact that their capital structure is what enables this sort of business to survive and I bring this up because in the long run if they don't have their own kind of competitive advantages in the way they're running the infrastructure which it's not clear that they will because AWS Google Microsoft they'll have you know very talented teams they could set up data centers as good or better than than core weave and anything core weave doing at least in my opinion is copyable by them and so once that competitive differential goes away differentiation goes away all that's left is sort of this Nvidia relationship which if you're no longer in a capacity constrained world that also becomes more relevant and then you're basically just kind of a financier of compute and so what does a financier of you know airplanes sort of get at least the market right now puts a ten times multiple on that business and so I've kind of put that in there just to help people think a little differently maybe be a little provocative I'm not trying to say core we should trade you guys know opinions on this company let me tell you he's got not he's very I'm on bias I have no opinions and I'm on bias but that's just something that might be worth thinking about for some people something about it's the worst balance you've ever seen and similar businesses right at that time certainly it's your decision all I could flip it though and I could say you know if you want to throw 25% mature margin on them you know then they're roughly trading at 65 times you know mature no pat right now and they're growing a hundred percent plus you only need several more years of that for that multiple to come down a lot and then if they could continue to grow their afterwards because the AI you know demand is still there then you could be looking at you know stock that's you know continuing to grow earnings very fast thereafter yeah you know so listen there there's a there's a world that's all we're going to say there's a world where it works out and again like we said at the beginning of the discussion which is there's a lot of smart people a lot of people on both sides to debate we think that there's a a fighting chance for core week to really succeed and become a major player in the space or you know again the six to seven times you know operating income interest expense might be too much for them to did you have to do that on a calculator because you weren't able to do that normally it was six four six seven times that's a hard I can't I don't have that math like that I was like the six and seven I didn't break it evenly all right I don't need this yeah numbers man you're in the wrong business I'm not going to stand for this much longer drew I swear to god I will walk I don't get paid enough for this level of abuse so enough about that so anyway I think great YouTube video on core we've this is a kind of a nice supplementary analysis to it and go to you know Drew Cohen money.com he calls it five minute monies the YouTube videos 30 minutes so I don't know why you're misleading people it's a five minute read maybe it should be a five minute read this one was probably like seven or eight but I'm very idealistic and you know it's a way to bring them in they're like five minutes like I got that five minute read on a 30 minute video whatever you want but now I think it's I think it's a great breakdown of entertaining it's in depth enough you're going to have a bit you know good understanding of core we've and it's a fun watch so take a look at that and until next time until next time