Why Wall Street Is Buying Insurance Against an AI Crash

0:01Anthony Cheung: This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome? That's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required, compatibility and availability varies 18+. Is the AI trade finally unraveling? While stock investors chase the upside, bond and derivative traders are ringing major alarm bells with credit default swaps, CDS's trading volume surging past$9 billion a week and AI debt topping$466 billion.

0:46Anthony Cheung: In this episode of the Market Maker podcast, we're going to try to decode the credit signals, demystify CDSs and examine what happens when hyper growth meets rising debt costs. Quick shout out as well to everyone at Warwick University who joined me late yesterday evening. Many of you stopped by to say hi and that was great. And actually, one of the guys, Zach, said he's from a farm in Dorset and the podcast was his introduction to finance. How cool is that?

1:19Piers Curran: loving it was he in the tractor whilst he was listening he wasn't joking yes he was

1:27Anthony Cheung: awesome so yeah shout out to him and everyone else from yesterday but yeah keen to get into this one because there's there was an ft article which was kind of the inspiration of the of the piece and obviously there's a lot of of noise about this going around about the ai trade and and all these sorts of things. So yeah, wanted to get your take.

1:48Piers Curran: Yeah, there's been a bit of a freak out, hasn't there? Over the last, well, I'd say like, yeah, maybe week, 10 days. And it's definitely all centered around the debt side of the AI trade. Now, obviously, it's normally the equity side of the AI trade that everyone talks about, you know that can be anything from you know anthropic ipoing or it could be you know talking about nvidia's share price or even more broadly where's the nasdaq trading right that's all equity and actually talking about the nasdaq you know that has been it's been sideways for for well a while now let's just say and um you know it's lost a bit of momentum and i'd say you know we've been kind of flatlined a little bit for the last one just looking back at the chart now because where we're trading today actually we were trading back in the middle of may right so we've had five months of sideways action so a top level an index level whilst we haven't gone down it's it's stopped going up so then you kind of look underneath the bonnet of that and we'll dive in but really what caught people's attention i think over the last week or so was something specific it was the credit default swap uh trading volumes in and around uh things like core weaves uh corporate bonds um and so this is not just core weave but also the big guns you know there's been a wave of the hyperscalers issuing and borrowing money, right?

3:33Piers Curran: So huge amounts of debt levels have been dumped on the markets. But very, very specifically, if you look at the volumes, credit default swaps we'll dive into in a second. But for now, it's just insurance on a borrower defaulting, right? And the volume has spiked. So$9 billion worth of average weekly trading volume is now where we're up to and you just go back a couple of months and the average weekly volume was more like 2 billion. So all of a sudden, massive upside in weekly volume. What does that mean? Well, that just means more people are buying insurance. Why would you buy insurance? Well, of course, you're more worried about the risks.

4:14Piers Curran: The risks of what? Well, the risks of the AI world defaulting on this debt. But I've already mentioned the hyperscalers. So this is a nuanced story. I mean, is Google going to default on its debt? I'm here to tell you, folks, that the answer to that is no.

4:34Anthony Cheung: That's a great point, actually, because one of the things, one of the stats that I said at the top of the show is an accurate run, is the level of debt, 466 billion. That's AI-linked corporate debt issued year to date this year and that number's up a lot it's from 101 billion in the same period of 2025 but exactly what you said i mean a lot of that is what amazon alphabet meta like the big boys yeah who've got tons of cash in terms of a lot of the problem i guess the issue

5:12Piers Curran: is, well, we're about to find out this earnings season at the moment, meta are up first. When's meta earnings? That's this week, isn't it? Is it tonight? We're recording this on Wednesday 21st. Maybe you can just check that. But when these hyperscalers report their earnings in the days and weeks ahead, then one key thing is, right, what's their spending plans? What's their CapEx spend for into the end of the year? Plus what's 2027 looking like for them in terms of spending on their data center build out. And to fund the spend, they've dipped into different buckets. Cash flow, because these companies just spew out a ridiculous amount of cash flow, right?

5:53Piers Curran: That's number one. Number two then, more recently, is borrowing money. And so that's 2026's sort of vibe and trend is let's hit the debt markets to finance this they could always turn to equity as in issuing and selling more equity of course maybe that'll be 2027's uh bucket that they'll dip into but they've they've changed along the way um but you know i'd say the kind of the the main thing here is it's not the it's not the hyperscalers of course that we're worried about but if they slow spending yeah even for them interest rates are going up if they don't want to issue more equity and they don't really want to issue more debt at these higher levels and for some of them not all the cash flow is looking weaker then they're kind of snookered and are they then pulling back from spent that that's what we want to hear because any sign of pull back on spend then this whole kind of i won't call it house of cards but but the whole scaffolding of this whole thing you know just you know the leg right at the bottom just gets kicked out right and so that's what we're worried about and then who's vulnerable well then it's the smaller guys in the in the stack in the system if you like that will get crunched and it's people like core weave that are in the crosshairs there um but maybe we can talk about credit default swaps to kind of get things started because it's a bit more of a niche financial product.

7:29Piers Curran: Of course, if you're old enough to have been around during the financial crisis, then you will definitely know what these are. But they're called CDSs, right? So it's a credit default swap. And as I've already said, the best and easiest analogy is just to think of it as an insurance policy on corporate debt. So as companies borrow money to fund whatever, in this case, AI, they're obviously issuing bonds. They're then selling those bonds to investors, which is the vehicle for the investors to lend them money. So it's just a loan. So of course, then the company who's the borrower, well, they have to pay interest for the lifetime of the loan via coupon payments.

8:08Piers Curran: And then of course, they've got to pay the money back at the end of the loan when it hits redemption, right? Just a normal loan. If you're the lender, and indeed, even if you're not the lender, you can trade credit default swaps. So mainly let's go lender first, which makes most sense. The lender is the person at risk. They have credit risk to the borrower. If they're worried the borrower is not going to be able to fulfill their commitments with this loan, they can buy insurance so that if the borrower does default, well then fine, we can get a payout via our insurance. Obviously, insurance though costs money.

8:45Piers Curran: As the lender, we have to pay an insurance premium. So in exactly the same way, credit default swaps, we think about what's the cost of this product to the person that's hedging their credit risk. And this is where we're looking when we talk about basis points in the CDS markets, then we're talking about the cost of borrowing. So here, this is how it works. If your CDS was trading at one basis point, that's a 0.01%. So if you've got debt exposure of$10 million, as an example, and there was one basis point, it was the CDS cost, well then effectively one basis point is 0.01 % of$10 million, which is$1 ,000.

9:31Piers Curran: dollars basically pay one thousand dollars per year as a premium if you like to be insured and fine if the borrower blows up you get your 10 million back from the insurance company okay now obviously no one's trading at one basis point that that's just for a reference now if we look at some of the big guns we've talked about the hyperscalers borrowing money so meta they have a lot of debt so if you're wanting to insure against them defaulting which is clearly an unlikely scenario, they're trading at 108 basis points. So that costs you$108 ,000 per year to insure yourself against$10 million worth of debt.

10:10Piers Curran: Okay. NVIDIA, obviously even more stronger, they're trading at 101 basis points, right? These are the absolute solid, if you like the AAA rated, the safest, okay? 101 NVIDIA. Then you get the others, the rest of the pack, and obviously Google and Amazon and Microsoft, they're going to be in and around those hyperscalar levels, just around above the 100 basis point mark. If you go to others, Oracle, for example, who are also a massive company, but they're fighting this race with less firepower. They've got much lower cash flows, for example, and have had to borrow more. So for them, their credit risk is higher.

10:48Piers Curran: They're trading at 322 basis points, their CDS. So that means you'd have to pay$322 ,000 for every 10 million that you're wanting to insure against, right? Then you go to the core weaves. So that end of these neo clouds, and I'm going to deep dive into them a little bit later on in this podcast, because they've been the ones that have really been leading this wobble in markets. They've been the kind of poster child of this risk in a way. but for them their cds spiked and surged higher and this is what caught the attention of the press they're trading above a thousand basis points now right um so so remember oracle 322 core weave above a thousand and when you get up to that levels you're considered to be so risky that then there's a slight difference in terms of how you insure and you've got to pay up front basically plus then a small amount of quarterly top-up just costs a lot more to ensure and you've got to pay up front and so this is where you get into what's called distressed territory and it is true in one sense core we've cds levels are suggesting and if you do the math and i'll go into a bit more later it's basically a 55 chance they default in the next five years that's what the cds market's saying now is that true like from the fundamentals i would like to go through later

12:22Anthony Cheung: one of the questions there is you said earlier cds is if you're around during the european southern crisis or the global financial crisis some of these other episodes of volatility how behavioral is the cds market as opposed to fundamental particularly in the current context of what we're talking about.

12:41Piers Curran: Well, I mentioned you can trade CDS if you're the lender and you actually have exposure, but it's just a financial asset. Anyone can trade it. I don't need to own core weave debt in order to trade core weave CDS, right? So it just becomes a vehicle for speculators. And this is where you get moves in markets that can't just be explained by fundamentals. this is where you get momentum you'll get speculation volume coming in you'll get you know behavioral you know FOMO issues there's so many different things that influence price right and so it definitely can be the case that you get a market trading at a price where you know that suggests they're going to most likely going to default but do the fundamentals say that and often there's a gap and that's the difficulty of operating in markets right that kind of behavioral gap between what fundamental theory says should be happening and actually what is

13:44Anthony Cheung: happening so let's say an equity manager a multi-asset fund manager they're buying these cds's as a kind of tail risk hedge against the ai bubble bursting what other options do they have though is there not like an options play here or is yeah how do they choose this vehicle to express

14:06Piers Curran: that hedge right so i mean look there's clever people out there in the world there's lots of different products a put option would be another good example where you could trade a put option on core weave's shares share price basically and so here a put option is i won't go into the finer of details, but it's a derivative that essentially you pay a premium, so your cost is fixed, but then you only get paid out if the share price drops, right? And the further the share price drops, well, the more you'll pay out. And so again, it's a different way of doing it. However, the big problem with the options market, when you're trying to hedge against this kind of stuff is volatility.

14:53Piers Curran: So the heightened volatility of these share prices just leads to a prohibitively expensive options pricing. Basically, the options pricing, one of the components is implied volatility. And when that's mega high, it just makes the whole equation not quite work out. So credit default swaps are essentially the... It's like equity managers and multi-asset funds are buying this stuff because it's a cheap tail risk hedge as they might say um and so if this ai trade starts to derail well then that cds product is is one of the cheapest ways of you know getting a nice little hedge against that situation one of the things that steven and i

15:38Anthony Cheung: often discuss is this kind of ai value chain because i just think it's a really neat way of thinking about the ecosystem and then the idea of like where are the the parts of fragility in the system it's almost like i feel like it's like an orbit if you think visually like space in that way there's planets so here can you explain to me this kind of multi i've heard it termed the multi-tier risk spectrum and you've mentioned a couple names already here so companies like you mentioned hyperscalers so what are those tiers and how how do you articulate that yeah and i think

16:14Piers Curran: that when i was talking about different level of basis points on cds's basically i was answering your question so tier one when you're talking about multi-tier risk spectrum tier one is your mega hyperscalers who are all over this race and are spending i just i lose count i mean is it like 700 billion dollars this year i think just the hyperscalers are spending on data center build out and just generally the ai race more broadly so meta alphabet microsoft amazon okay big notable absent member there from that list is apple who have chosen to go an entirely different path. They have decided to not compete in this AI race.

17:05Piers Curran: And whether they'll be right or wrong in that, obviously, time will be the judge. But those big guns, Meta, Alphabet, Microsoft, Amazon, right? Even them, as I mentioned, even them, their yields are climbing, their spreads. And we talk about spread. We talked about this on another episode, I think. When you think about corporate debt and what's the yield on corporate debt, and we talk about a spread, what's the difference between a corporate bond yield and the government bond yield. And so those spreads are widening. So even though US government bond yields are creeping up, it's the corporate debt yields that are going up as well, but at a faster rate.

17:43Piers Curran: So everything's going up, but also the spread's widening, right? So even for the hyperscalers, this is happening, even though obviously their balance sheets are absolutely rock solid. But massive capex spend is eroding that cash flow visibility. And so, you know, it does introduce a complexity premium, you might call it, where it's just a bit more risk. How sustainable is this? Fine, they're doing it for now and it's fine for now. But what about in 2027? And that's why this earnings season is so key. What's the profit growth what's the revenue growth from ai and is it still ramping massively higher meaning they're justified in steadily increasing their spend to say in the race it's all you know it definitely we're in the part of the cycle here where they we need tangible revenue and profit generation from ai in order for the markets to be happy for entities to carry on with these mega spending outlays.

18:50Piers Curran: So this earnings season is really key. So that's tier one, right? Rock solid, balance sheets, but they're the big mega spenders. All right, then you've got tier two. That's more enterprise infrastructure. Also, well, do you call it private tech? It's not private tech anymore when you're about SpaceX, but Oracle and SpaceX would be kind of tier two, right? I've already mentioned Oracle. They're trading at 322 basis points. SpaceX, well, they borrowed a ton of money. They tapped the corporate bond market literally the week after their IPO. I think they raised 25 billion, right? The CDS on that is trading at 226 basis points.

19:29So higher than the mega

19:32Piers Curran: cat hyperscalers but lower than oracle okay so they're in the mix there i just wanted to throw

19:39Anthony Cheung: in you you mentioned spacex there we had spacex earnings uh last night and uh you know one of the things there was pretty spectacular numbers actually um spacex posted 4.4 billion dollars in net income for the quarter ended september expectations were 1.88 billion and then check guess out spacex is ai business so as to your point how is ai performing spacex is ai business which elon's doubled down obviously into now going by the name super intelligence or i think it's si he calls it that went to 7.56 billion dollars in the quarter that was nine times what the company generated a year ago yeah so is that is that kind of just outlaying some fears then

20:29Piers Curran: short term in the last 24 hours i'm going to dive into this when we talk about core weave if you're in the business where you have compute power that is what you have as an asset you are absolutely laughing at the moment because the cost of compute is just ramping up and up and up and up and we'll come back to that point when we think about core weave and are they you know are they fit you know is it more than half is it more than 50 likely they're going to default in five years we'll come back to that point um but yeah spacex and did you see obviously there's now tesla spacex merger noises i don't know if you've been seeing on social media tesla's uh you know with rocket jets and jumping over cars and doing somersaults and all this crazy stuff so So obviously, this is all part of the kind of circus and the show, you know, further down the inevitable path of obviously SpaceX and Tesla coming together to become one entity.

21:34Piers Curran: But yeah, it looks like we're getting closer and closer to that point. But look, SpaceX are tier two. You've got tier threes then, neoclouds, right? They're right in amongst this. So CoreWeave, for example. Now here, they've got high debt levels. and I'll go into more detail in a second, but they've got very high, well, actually, why in a second? Let's go into it right now. But essentially, they've got high debt levels, but they've got a lot of collateral. Collateral is assets of tangible value that underpin these loans. GPUs, for example, you might've heard of NVIDIA GPUs. Well, cool, we've got a bunch of those and that's a really great asset to be having in your back pocket.

22:20Piers Curran: But they've got rapid expansion plans because the demand is just off the scale and they want to try and tap and take advantage and monetize that demand. But it's just a little tricky. So let's maybe dive in on the CoreWeave case study. And we can start, let me just get to my notes here on that. So yeah, other than CoreWeave, If there's IREN and Nebius, okay, these are some of those other NeoClouds. And these are cloud providers that are, you know, you've got your AWSs. I think we covered this on another pod. You've got your AWSs, your Azures, and your Google Cloud, right? That's general purpose cloud compute.

23:04Piers Curran: Now, then you've got the specialist ones. They've built out their cloud specifically from the ground up with AI in mind, okay? so that's core weaver kind of at the front of that race by the way their share their share prices dropped 20 percent um over the last um couple of weeks and you know as their cds has blown out and as i said it's over a thousand basis points this is now feeding into their share price you even had elon musk you know tweeting uh yikes was musk's uh thoughts on the movement and the pace and the speed of the share price and the cds price movement here but um did see

23:47Anthony Cheung: did see another one as well uh elise did you see that one okay yeah elise was a soul based so korean is one of the bigger korean ipos of late and they their shares fell on their opening day nearly 40 they finished down 32 it's the worst ever opening day kind of loss if you like for a a bigger IPO in Korea.

24:10Piers Curran: Yeah. Well, there's nuanced cases. There was an Australian data center builder, I forget the name now, that was going to IPO. They've had to back off it. And they've pulled out of doing the IPO entirely. That's more of a regulatory thing because basically the government in Australia are blocking the building of data centers. And if you're in the business of building data centers and the government won't let you obviously that's a problem so you've got regulatory risk is right in amongst this but um like think about core weave i want i want to explain this it's really complicated again you've got to be really really clever to build complex financial arrangements in order to try and take advantage of what is the biggest demand spike we've ever seen ever in human history for anything so if you're in the business of trying to monetize that the time is now and the time and the opportunity is a monstrous one it's just how can you get it right so well you you take risk and you double down and you borrow a ton of money to make it sound like a salesperson from goldman

25:25Anthony Cheung: trying to sell an investment vehicle to call we for it to get you another 100 100 billion

25:32Piers Curran: well let me i'm going to explain it in simple terms right and actually the way i'm going to explain it is how the call we founder and ceo explained it

25:41Anthony Cheung: this episode is brought to you by google chrome you think you know a browser but gemini and chrome that's new it can help you with practically anything on the web like restoring a vintage motorcycle from a 50-page restoration block or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required compatibility and availability varies 18 plus.

26:07Piers Curran: So he's in the business of winning compute contracts from clients. Microsoft, for example. i'll come to core weaves risks one of their key risks is their revenue concentration 75 percent of their revenue comes from three clients um which are three of the big hyperscalers of course one which is microsoft so look microsoft need compute power okay and core we've core we've got it right or got the ability to supply it so they win a contract and this is a long-term contract Microsoft might buy five years worth of compute off them. Today, they'll fix a price. And that fixed price, again, is a problem for CoreWeave that I'll come back to.

26:53Piers Curran: It's a fixed price. I buy X amount of compute off you over the next five years. Okay, fixed price. Right, that's a contract that has now value. From CoreWeave's perspective, that's an asset. You have one of the biggest companies on the planet guaranteeing you a certain amount of revenue every year for five years. Awesome. Now, Compute, sorry, CoreWeave need to then buy some NVIDIA GPUs in order to essentially create this compute power to then service to Microsoft. So they'll then go, they'll use that contract essentially to buy NVIDIA GPUs. Okay. Then they'll sign a data center lease and they'll put all that to work.

27:39Piers Curran: And he describes this as being a box, this financing arrangement. The CEO says, look, there's a box. Into the box goes Microsoft's contract for five years. Into the box goes NVIDIA GPUs that I've now bought. Into the box goes a data center lease contract. Obviously, we need to finance all of this. So also into the box goes borrowing. so this is where CoreWeave will borrow money they'll tap the bond markets and they'll borrow money and and that goes into the box as well okay CoreWeave then builds and delivers the compute to Microsoft Microsoft pays as we're going along they don't pay that five-year contract up front obviously so they pay in stages the cash then waterfalls down in this order firstly the data center gets paid then you pay the power bill then you pay the interest and the principal on the debt and whatever's left goes to core weave.

28:36Piers Curran: Each box, it's an amazing model because each box pays itself off entirely within two and a half years. This is for a five-year deal. The way it's all working, they're clean and out of debt in each box in two and a half years. And the more this was working, and each box is ring-fenced, so there's no contagion risk one box to another, okay? And as the model was keeping on working, lenders were offering progressively cheaper financing because we're, hey, this is obviously a good thing and everyone's a winner, right? So that's how the kind of system works. Now, what happens, and by the way, Call Weaver borrowed$35 billion in order to finance these boxes.

29:22Piers Curran: The problem they've got is they need to borrow another$35 billion in order to deliver on contracts they've already got from the Microsofts and so on. So they've got$35 billion more to borrow. That's the big problem. It's not the debt they've got, because actually they've got about $75 billion worth of assets. So that's not a bad cover ratio,$75 billion of assets, they've got $35 billion debt, that's fine. It's just the next$35. And how much will it cost them to borrow? Because hey, yields are rising, right? So you can see that. That's why the CDS market is starting to go well hang on a second is this is this sustainable um and so what happens from there is basically you've also got you know why are the cds markets spiking yes um obviously we're worried about the whole spend sustainability of this whole thing but there's also macro factors so you've got rising treasury yields that we've spoken about this at length inflation's on the up all right we're worried about debt sustainability at a government level this means treasury yields are rising when treasury yields rise everything rises right all yields um so core weaves hyperscaler contracts by the way that are sold at fixed prices for five years they sold these when funding was cheap but they're funding it as they go along they're borrowing money as they go along and that the longer you go down the path at this rate the borrowing is getting more expensive okay so um the average they've bought borrowed at so far for that 35 billion is eight and a half percent the problem is where's the next 35 billion where's that what where's that going to land in terms of costs and it could be anywhere between 10 to 15 percent at the moment and the problem we have is that's not sustainable so that's why your cds markets are kind of freaking out given the

31:25Anthony Cheung: the tiering so the interdependency the circularity of the funding all these sorts of things which people talk about with with the ai ecosystem the systemic risk then a single link in revenue generation at the app user layer does that threaten to then reverberate backward through the entire credit chain and given that we keep getting this delayed anthropic uh ipo and some of the noises we were getting a week or two ago were incredible growth but it's at that level is this also kind of inputs that have resulted in some of the uh the kind of pickup here in volatility that we've had um i i think this is like people obviously have been saying

32:17Piers Curran: oh it's a bubble it's dot-com bubble you know 2.0 there's such a big difference between now and then And actually, the key deal, if you go through this fundamentally, well, demand. Is there still demand? What's the demand levels like for all of this stuff? And we know for sure demand is just off the scale. And I mean demand right across the spectrum. I think it was the Amazon CEO, Jassy, he was saying about a few weeks back, that if you want to get hold of nvidia gpus basically they're sold out until the end of 2028 that's more than two you can't get any for more than two years so and that's just at the gpu level right but then if you think about demand from a consumer and a corporate level i mean that's only going up right i mean for sure we're really probably just at the beginning of that demand cycle.

33:24Piers Curran: So I think from a demand side, there's this thing about the dot-com bubble, the demand, basically the build-out of the internet and all the infrastructure, just got way ahead of itself compared to the demand. Demand was restricted because, well, you needed an internet connection, you needed a computer, you needed... There weren't such things as smartphones. Back then when the internet started, if you can believe this, to get onto a browser, as in a Yahoo browser, you had to pay Yahoo, a subscription fee, to use the browser. Obviously, Google came in and tore up the business model and captured the entire market share.

34:06Piers Curran: But pre-Google, you had to pay for this stuff. People didn't have connection. Demand just wasn't there. It was such a brand new thing infrastructure build out got way ahead of itself and the supply therefore was way way more than demand and and actually there's a story about one of the neo clouds back then if you like um who called loud cloud um back then it was bandwidth cost right we talk about compute cost now back then it was bandwidth cost and when supply was super short and the internet was blowing, was kind of expanding. It got out to the bandwidth cost$1 ,500 per megabit in 1999. By 2021, oh sorry, by 2001, it dropped and crashed to just$50 per megabit.

34:58Piers Curran: So the whole supply demand dynamic just collapsed. And so everything collapsed. Here, there's no supply and demand imbalance. There is demand that's off the scale and growing, and there is just a chronic shortage of supply. And so these drivers mean that we are in a whole different world than we were back in 1999, 2000, 2001, in my opinion, right? And then the bull case, if we're thinking about CoreWeave, the bull case would be, as I mentioned, currently they have 75 billion in assets against 35 billion of debt. Right now, they're solid. But of course, it's more about the future. They've got to borrow more.

35:41Piers Curran: But they've got GPUs. That's their asset. And there aren't any GPUs. So that's an asset whose price is going up. Plus, you look at the older GPUs, the lifetime of value you can get out of these assets has also destroyed expectation. You've got H100s. that are still being rented at full 100 % premium prices. Go back a couple of years ago and people were saying, well, you're going to have to write those off in three years because they'll be obsolete. No, they're still being sold at absolute premium price because you can't get any GPUs, right? So this asset in their back pocket is really, really powerful.

36:27Piers Curran: So I think from a bull case perspective, the fundamentals are solid it just doesn't it doesn't mean they're not immune though because i mentioned customer concentration 70 75 of revenue from free customers and if those three start to go you know what we're going to spend a bit less then you then the negative feedback loop kicks in.

36:58Anthony Cheung: Well, Meta are reporting tonight and I don't think there's any risk of Zuck not spending more money given the noises he's been making in the last two, three weeks. I agree. So I guess as to your point then, maybe to conclude, it almost feels like this opportunity in the modern history of man is so great that the ingenuity and therefore the complexity around how these financial arrangements are made are going to get more complex, which means there are going to be casualties along the way because there's such an opportunity to monetize the demand, if you like. But I guess what you're saying is that whilst the difference between that and the dot-com is it can withstand as long as that demand pattern remains.

37:48Anthony Cheung: And I guess this earnings season... It's kind of come at a good or a bad time. It's kind of come at a good time that we now get to see and answer those questions pretty quickly in literally the next seven days probably and we get all the big hyperscalers come out.

38:05Piers Curran: Yeah, exactly. And I think ultimately for CoreWeave, they've got a break-even point, right? Because remember those fixed contracts they entered into with Microsoft? You know, how much is it costing them to finance the delivery of that compute well as the cost of finance creeps up and up and up and up that is eating their margin that they had when they locked in that fixed price contract and it's going up and up and up and right now rough maths we reckon they're about five billion dollars away from the break even the tipping point if you like so if yields continue to climb and they're forced to borrow more at these higher levels, there is a tipping point where suddenly this whole thing becomes unsustainable.

38:57Piers Curran: Then your CDS market rockets higher, which feeds into yields going up more, which feeds into borrowing costs going up more, which means their profit margin gets blown up. And that negative feedback loop can break the system, even though the the fundamentals of the whole thing are still solid as long as the hyperscalers continue to

39:21Anthony Cheung: spend do you know what i can't help but think while i've been listening to you through this episode of that scene from the big short of ryan gosling with the jenga bricks and i've just got you you're so animated when you're talking about it but it's almost the opposite you're not trying to sell uh kind of this whole house of cards you're just pumping it up you're like come on load up some more debt i've got a couple more boxes for you i can fill up we can roll that over give me another

39:48Piers Curran: i'll give you another 35 i get you another 235 that is one of the great scenes in uh movie history but well certainly finance related movies maybe i could get the ai i'll take your face and then

40:06Anthony Cheung: i'll get the transcript and then i'll take that scene and see if we can see basically talking

40:11Piers Curran: about upgrading ryan gosling's face to mine is that what you're saying well we'll let the audience

40:19Anthony Cheung: decide on that one but i'll also drop a poll actually on core weave and their future okay

40:25Piers Curran: cool i thought you're gonna say who's who's better looking i'm not sure not sure we should do that

40:31Anthony Cheung: poll um but i i guess is it too is it too bold to say will core weave still exist at the end of 2027 yes or no or is it better to do a poll of cds like or default percentage and i'll have some

40:47Piers Curran: brackets i think make it much shorter term will the cds price of core weave be higher or lower than it is right now by the end of this earnings season so in let's say two weeks time are we

41:06Anthony Cheung: higher or lower right hit us in the comments as well of what you think all the poll if you're on spotify and thanks always for listening if you've made it this far you're not a subscriber remember to do so love to have you in the community and we love engaging with comments as well so feel free to drop one thanks very much here's and shout out to elon and the starlink team not that they sponsor the show but wherever peers goes the starlink team have got you covered they should

41:35Piers Curran: sponsor me can i get a free couple of free dishes that'll be good all right cheers piz okay mr

41:42Anthony Cheung: monopoly here miss me monopoly is back at mcdonald's with millions of prizes a 2027 cheap charity a carnival cruise vacation gift cards from doordash and best buy he'll then play in the mcdonald's app get your bag no purchase necessary 50 us in dc age 18 plus see play

42:02Piers Curran: Play at mcd.com for rules and play at mcd.com slash AMOE for info and playing without purchase by mail by November 2nd. Chance card game ends November 8th. Game ends November 22nd, 2026. See rules for prize details. Monopoly is a registered trademark of Hasbro. See you later.