Did the AI Bubble Just Pop?! Anthropic's Leaked Numbers are INSANE

Key Takeaway

Before making an AI-related investment decision, map your exposure: identify whether you own model providers, chipmakers, infrastructure, private-credit vehicles, or broad funds—and decide in advance the maximum percentage of your portfolio you are willing to risk. Keep some exposure to a transformational technology, but avoid betting on a single flawless revenue forecast. Preserve liquidity so you can act if valuations reset rather than being forced to sell during a downturn.

Episode Overview

The episode examines a leaked Reuters report on Anthropic’s purported financials and uses it to question whether the AI capital-spending boom has bubble-like characteristics. Tom Bilyeu argues that AI may be genuinely transformative while still being financed through fragile debt structures, aggressive revenue assumptions, and interconnected guarantees reminiscent of prior credit cycles.

Full Transcript

Transcript of Did the AI Bubble Just Pop?! Anthropic's Leaked Numbers are INSANE from Impact Theory. Auto-generated from episode audio; may contain minor errors.

You're listening to a quick ad. You know what else is quick? Selling your car on Carvana. Just put in your vehicle information and we'll give you an offer. Done. No hassle, no negotiation, no guy from an online listing lowballing you. Which means by the time I'm done explaining this, you could have basically done it. Still haven't started? Hey, better late than never. Sell your car today on Carvana. Pickup fees may apply. Ever notice how life's best stories don't happen in your living room? They happen on the open road, out on the water, or parked under the stars.

At Progressive, they get that you want to focus on the experience, not worry about the what-ifs. That's why they offer quality insurance designed for your ride, whether that's a boat, RV, or motorcycle. Adventure with confidence. Visit Progressive.com and see how easy it is to protect your favorite way to get away. Progressive Casualty Insurance Company and Affiliates. Not available in DC. Prices vary based on how you buy. Let's ask one question. Has the AI bubble popped? Jeff, take it away. This time is different, they said. They say that every time.

But this time was different. This is not going to be the dot-coms because the dot-coms had a bunch of money-losing companies all over the marketplace. And this time, we've got these massive businesses that are entirely profitable. It's not going to be like the dot-coms. Let me tell you why the idea that this time it's going to be different sticks around. Because things change just enough that it will be a slightly different flavor. It's going to come at you from a slightly different angle. But the reason that that always sort of ends up being a joke is that there are physics to the economy.

They're very complicated. It's very hard for us to track all of them because it interfaces with humans and how they feel and how they act individually and as a collective. And so it becomes a very complicated system. But the reality is, it is mechanistic. This is the part that people lose sight of. You lose sight of the mechanisms of the economy at your peril. This is why you can look at something like the eternal accumulation of debt and know that it's just going to be a problem because you can not pay it back.

But there's really only two ways to not pay it back. You can just not pay it back, a hard default. That's going to break the system massively. And then you've got a soft default, which is inflation. They both have very knowable consequences that you can strategically prepare for. So anyway, understanding how we end up with people just not being able to escape the temptation of, I can feel that something is different. And so I can't just use the old playbook. They're right about that. The entire way that we invest is changing.

We're moving into a new regime, I think. And so you do have to go, OK, there's something new here. There's something different. But at the same time, the fail states are mechanistic. And so you've got to watch out Yeah, right. Honestly, given what we know today, I mean, we kind of had an idea. But what we know now today, I really hope that the AI bust happens sooner. I mean, tomorrow, I hope this thing starts tomorrow, because the longer this goes, the more devastating it's going to get down the road.

It's the second person that we're now hearing say this. Is Jeff the biggest voice in finance? No, absolutely not. He's definitely a niche guy. But he's pretty smart, man. He understands the economy. I've had him on the show. I've watched a ton of his videos. So this is not some crackpot in some random corner of the Internet. This is somebody who's looking at the math and they're saying, oh, if you share his base assumptions, I'm going to challenge one or two of them as we go.

But if you share his base assumptions, you can actually build the mental model of why he is so worried about this. AI is hoovering up all the capital in the world, and it's subjecting itself to the mechanisms of debt. And so historically speaking, people that take on this kind of debt, which this really is unprecedented, but people that take on an extraordinary amount of debt for a revolutionary new technology, meaning it's not been propagated through everything yet. History is brutal and you're going to get hammered.

So what he's looking at is, bro, if this just grows and grows and grows, it's just going to get worse because, and this is where Michael Burry agrees with him, the amount of capital from investors that will get destroyed is biblical. AI is going to be a game changing technology, but the amount of money that's going to take to get there and the fact that not everybody's going to win and not all these numbers are going to make sense means that there's going to be a bust at some point.

And the longer this goes, the bigger it gets, the more trouble there will be down the road. I like that take, which is basically, um, what we know about companies is they come and go and sure there are going to be winners and those winners will probably last for decades and decades, but they're, they're going to be losers in this timeframe. We've never not seen that. And so given how much capital is getting pulled in, and we know that some are going to go away, some amount of this capital up to, and including the vast majority of it, remember.com bubble, we draw it down like 90%.

Yeesh. So you're guaranteed. There's going to be some amount of loss. It's just a question of who exactly is going to lose the money. And so given that we know what's going to happen, the sooner, the better that statement in isolation actually is true. Well, Reuters reported last night that Anthropix, one of the stars of the AI bubble, it's S1 was leaked to Reuters. So Reuters basically gave us some insight into one of the biggest names in AI, but previously we had no information. I mean, these are private companies.

And then once they go, once they go public, they, they, uh, undertake an IPO and raise money through the public marketplace. They got to start reporting things. And as they start reporting things, they still, they tend to leak out. And so that's where we are. Anthropix S1 leaked through Reuters. Reuters gave us a bunch of numbers that allow us to, like I said, start to put some numbers and start to put together a case for where this AI thing is going to go. And look, the numbers are just absolutely staggering.

And like I said, I really hope the bust happens sooner rather than later, because the longer this goes, it just gets more and more insane all the time. And this time is not different, including the structure of the AI bubble. And we'll get into all the details there, but let's, let's get into what Reuters reported. Uh, what was it late, like late yesterday, just a whole bunch of, uh, just, you know, just for, I mean, like I said, we kind of knew this was taking place.

We knew it wasn't going to be a pretty, um, but if you see the number right there, 518 billion, that was just 518 billion man of dollars that are already promised. And remember they can't just back out of it, no matter how bad the company does, they're still going to have to pay that figure. It's just insane amounts, half a trillion. So Anthropix IPO perspectives, the S1, uh, surging costs and also huge losses. Let's start out with the, from the beginning, uh, revenue grew 12 fold to nearly 4.6 billion.

So that's pretty good. That's last year, 4.6 billion. All right. Given the like scale of the numbers that we're talking about, 4.6 billion probably doesn't sound like a lot, but the reality is it, it is very impressive. That's revenue, man. And given where they were just a year before, that is really a fast growth rate. Um, I've heard people say that it's truly unprecedented, even if it's not truly unprecedented, they are in the absolute upper echelon and it may be the fastest that a company has grown.

It's very, very impressive. When you start looking at like, maybe not by percentage, but by absolute dollars, it's staggering 12 fold increase. It's that's, you know, fits with the description of AI. However, their operating losses, and this is just their operating losses were 8 billion. Okay. So this is where you have to understand operating losses. So we're going to talk about paper losses in a minute. The paper losses can be big without really being something that you have to worry too much about. This is going to be the one knock that I'm going to really make against Jeff's analysis here.

And part of what should temper your thinking, but operating losses is real talk. So that's like, how much did you have to spend on employees on, um, equipment and things like that in order to actually make that money? And the answer is significantly more than what they brought in. So if you have to spend whatever 12 to make five, I think it's a little bit different than that, but that's directionally correct. Uh, you have a bad business. What's known as default dead. So by default, the company is dead.

It is not yet profitable. Certainly not of these numbers at the time of these numbers, this is all 2025. Remember that? So we're deep into 2026. We have a lot of unaudited financials, uh, but we don't have a full year of 2026, but that's going to be a big part of what should temper some of your fear. So it took them nearly 5 billion in revenue, but they had to spend 12 billion in expenses. So 12 billion expenses to create 5 billion in revenue. And that wasn't even, that's just the start of it.

The real big one, as far as losses are concerned, they overall, when you factor in spending in, in, uh, capital expenditures, $42 billion loss in 2025 on $4.6 billion in revenue. And these aren't even the worst numbers that out there, the worst. Okay. So now we've got to talk about this number because this is really where you have to start digging deep and say, okay, for the narrative, you know, 42 billion compared to 4.6, like what a gap, like this is so crazy. But the reality is that a lot of that, in fact, all of the 42 billion is what's called a non-cash paper loss.

Meaning this is not actually dollars and cents that are leaving their bank account. Now, this is about the fact that their, um, shares have gone up so much in value. And a lot of the deals that they've done basically put them on the hook where they might end up having to buy those shares back. And so now you're in a situation where they may end up having to buy back shares at a much, much, much, much, much higher valuation. And so, yes, technically that money they're on the hook for, but depending on what the deal is, that could be years down the road.

And so it was not necessarily a today problem, but they do have to market to market. Uh, so that's where this starts to look a little bit scary on paper, but the real thing that you want to be thinking about right now is just that the 4.6 with an $8 billion loss, that is real. That's like your real cash deficit that's leaving the business. And so that is what we're going to be benchmarking against what's going on in 2026. And I'll talk a little bit more about that in a minute.

Number is 518 billion. That's already been committed, committed as far as spending over the next couple of years. That's not next year. It's not this year. It's over the next couple of years, half a trillion. These, these people have 4.6 billion in revenue last year, and they've already committed half a trillion dollars. This is a big reminder about what people actually think is going on in AI. Uh, listen, it's disruptive for all of us. I think everybody is aware. Nobody's like sleeping through AI, but I don't think people understand the size of the bet that we're making.

And you become numb to the numbers so quickly. This is why earlier I was forgetting, wait, was it less than a hundred million or less than a billion? It's like the difference between those two is massive, but we're getting so used to hearing these huge numbers thrown around trillions of dollars. Dude, these are massive amounts of money. $518 billion have looming over your head is, is absolutely insane. But what the market is saying is guys, this is a, um, a once in a civilization moment where we are transitioning away from humans being the only source of intelligence.

So it takes 18 years and a whole lot of fingers crossed to get something that is really smart. And then you've got to try to convince them to come work at your company. You've got to convince them just to be in your country, let alone work for your specific company. Now we're going into an age where we can spin up thousands, tens of thousands, hundreds of thousands instantly, right? There were tens of thousands of agents that were like self-replicating in the whole hugging face disaster. So we can spin intelligence up now so quickly.

And the only thing that's stopping us is compute. And the only thing stopping compute is our ability to bring on the data centers, the energy, and then raise the capital for it. So right now, from a liquidity, a global liquidity perspective, you've just got this big Hoover. That's like a black hole that is sucking up all the money and everybody's just fucking throwing their money at AI. Like, yes, please, God, let me invest more money, borrow more money, let's go crazy. And everybody is so convinced that this transition is going to happen on a reasonable timeline that they are being really, from where I'm sitting, reckless.

And so, yeah, don't lose sight of what that $518 billion is really saying. It's crazy. So you can see why this is a debt story. Now, Anthropic has already raised a pile of money through equity markets. I think it was $80 some billion to the last equity. And that equity, the last time they went into the equity market, they were valued at around $950 billion, almost a trillion. Let's call it a trillion. So they're a trillion dollar valuation. They're going to do another equity raise, which is why we're talking about Anthropic right now, the S1, which they're hoping to get a $2 trillion market value.

A $2 trillion market value for a company that has $5 billion in revenue, $40 billion in losses, and half a trillion in already committed spending. It's so wild when you say it like that. What I'm saying here, it's not that the AI bubble is... Oh, and by the way, we might lose control of civilization, just to add that. So don't forget, in their S1, it says, hey, as one of the risks that we have to disclose, we could end humanity. That shit is crazy. Priced to perfection.

It is priced to ultra perfection. There's absolutely nothing can go wrong. And the more information we get, the more you just have to shake your head. Like I said, hope the AI bust comes sooner rather than later. Then we can clean out the bad stuff, get the numbers back on track, and get the economy on a road where it can actually get to where everybody's trying to take it with this AI stuff. I get the sentiment, but let's look at the last time that this played out.

So if you go back and look at what happened during the dot-com bust, it ended up being 20 years of pain. So be careful what you wish for. Again, I understand the sentiment. He believes it is going to happen, and given that it is going to happen, let's make it happen sooner rather than later. But I think the reality is that we need to be looking at some way to slowly deflate the bubble rather than just say, all right, let's let this crash and burn. Or like Michael Burry said, let's hope that there is basically a cataclysm, that the markets fail to launch these IPOs so that we don't usher in Skynet and we don't destroy any more investor value than we absolutely have to.

That is a very, very brutal scenario. If we're talking about trillions of dollars lost, holy bejesus, it will set back the economy for a very long time. The AI is going to be a game changer. I keep saying that. It's absolutely the case. It's long run, going to be very productive, very positive. However, getting there is going to be very messy, and the more mess it's gonna get, it's gonna get that way because we didn't stop it soon enough. The question is, what do you mean?

How are we going to stop it? This is just individual investors deciding that they are or are not going to invest. This is individual banks deciding they are or are not going to loan. This is individual lenders deciding that they are or are not going to pull their money into some sort of private vehicle like a blue owl so that these guys can keep getting access to the capital. There's no way to stop this other than to get people to look at the information. But again, there is this, but I can see the opportunity.

I know where this could go. This is gonna be a little bit different than last time, and that's why I think getting the individuals to stop before something happens is very, very low. So let's dig a little bit further into what was reported. You can see Reuters reporting on the losses. It's not really the losses that we need to worry about. Yes, they do. They did have to put in there that, yes, one of the risks of their technology is that maybe it could end humanity.

That's a whole separate topic that we don't need to get into here. Suddenly, the scaremongering among AI. I'm not saying that there aren't tremendous risks there, but there's a political agenda behind all of that. I think you do have to get into it. The reason you have to get into it is to understand the play that's being made right now. There's gonna be a big political collision. You've got, these two are avatars for a bigger movement, but I'll just sum it up with, we're hitting pause for a moment, but there's plenty more ahead, so don't go anywhere.

There's a reason Quince sells cashmere for $60, and it isn't the cashmere. It's real 100% Mongolian cashmere. It's the same stuff luxury labels charge a fortune for. Quince just works directly with ethical factories and cuts out the middleman. So there's no need for a markup and no logo tax. The $60 goes to the actual sweater, not the fancy name on the tag. That's why everything they make runs 50 to 80% less, and it's not just the cashmere. I ordered a few things myself, their fleece joggers, a couple of their tees, which I wear obsessively.

First thing I noticed was the quality. Soft, well-made, built to hold up. They make down jackets and wool outerwear now too. So upgrade your stuff to things you'll actually wear. Find your next fall favorites at Quince. Download the Quince app for app-exclusive offers or go to quince.com. Get free shipping on your order and 365-day returns, now available in Canada and the UK too. And when Quince asks where you heard about them, let them know it was Impact Theory. It's the best way to support the show.

Let's talk about what it means to actually be visible in today's day and age. When you build something public, whether it's a business or an audience, anything on social media, you become a target. This is something that legitimately keeps me up at night because our data isn't just on the obvious broker sites, it's on old news portals, public forums, obscure directories, places you would never think to look. That's where Incogni's exposure scanner comes in. It hunts down your personal information across the wider web, not just the 420-plus broker sites Incogni already removes your data from automatically.

They can't harm you if they can't find you. Go to incogni.com slash impact and use code impact for 60% off an annual plan. Try it risk-free for 30 days. That's incogni, I-N-C-O-G-N-I.com slash impact and use code impact. Let's talk about how much you're paying to stay safe online. Count it up, a VPN, a password manager, something to block ads and trackers, the service to scrub your data off of broker lists, a tool to flag scam emails. That's five, five subscriptions, apps, logins, and most people still have gaps.

That's the problem Surfshark solves. It's not a VPN. It's an all-in-one security suite, one tool that protects your connection, your personal info, and your devices all at the same time. Go to surfshark.com slash Tom B or use code Tom B for four extra months of Surfshark backed by a 30-day money-back guarantee. That's S-U-R-F-S-H-A-R-K.com slash Tom B, surfshark.com slash Tom B. Use code Tom B for four extra months of Surfshark. Thanks for sticking around. Let's get right back into the action. The Anthropic and OpenAI are playing a regulatory capture game that is going to pit the Democrats against the Republicans.

And what you're gonna see both in the midterms and in the lead-up to the 2028 changeover is you're going to see the Democrats say the Republicans are being reckless. They're getting sucked into this money-at-all-cost, we're competing against China, but it's really just about making them rich. That's gonna be their narrative, that we need control, we need an anti-state, and that is very real for the industry. If you're an investor and you're looking at AI and you're thinking, okay, I've got choices. I can invest broadly, I don't know who's gonna win, or I can get very specific, and I can make an infrastructure play or I can make a play on the specific companies.

And the question becomes, who's gonna win all of this? Is it gonna be winner-take-all, Anthropic and OpenAI, precisely because they're able to get the Democrats reelected in 2028 and they start passing legislation that gives them regulatory capture. No other players are gonna be able to make it into the space. And now those guys are driving and in command and they start capturing all the value. Or is it gonna be, nope, they're not able to convince voters. Voters stick with the Republicans. The Republicans in 2028 maintain power and they keep running the playbook right now of we wanna make sure that we're beating China, we wanna make sure that we are capturing all of the upside in the markets, that we're getting AI in the hands of kids.

Remember, Trump accounts are a big part of this, absolutely brilliant move. We wanna get everyday Americans. We wanna get this technology in their hands in the form of shares that they get simply for being born. And now they get to go on this ride. And so why would we ever want to diminish the US in comparison to China by slowing down the innovation, which is exactly what will happen if the government steps in? Why would we ever want to not have open source? We wanna make sure that the cost of this is going down so that Americans benefit from actually engaging with the technology.

This really is one of the most consequential questions. One of the, like, it's happening in plain sight, but because everybody's lying about what their motives are. But it's very important what's happening in this Doomer debate. And even as an investor, you're gonna be forced to make a decision based on what you think is going to happen. Because the value is going to accrue in very different places whether they get the regulatory capture or not. And if they don't get regulatory capture, and you've got China continuing to run their playbook of using these open weight models to try to chip away at the way that they're driving revenue, which by the way, is already starting to work.

These guys are now shifting away from just like token maxing as people were calling it over to this API model. And so we're already seeing the realities of the business environment beginning to influence the models that these guys are using for how they're actually gonna generate revenue. So all of these things are extremely consequential. And if on the other hand, you get the Republican stay in power and open source remains a thing, you're gonna get the cost or the benefit, excuse me, accruing to the infrastructure layer because the companies will sort of come and go.

You'll eventually get big guys end up staying and just do something better than everybody else. And then they'll have a lot of momentum and scale on their side. But it's a very consequential fork in the road that none of us can see exactly which way we're gonna go. But you're gonna have to, as an investor, have some sort of strategy that takes that into consideration. But as far as the IPO goes, they have to throw that in there because the SEC makes them. They have to talk about every material risk that could potentially harm the stock.

Therefore, the public needs to be informed. The Anthropx technology could end humanity as we know it, at least throwing that in there just out of due caution. But really, it's not just due caution. It's part of their strategy, power as proof of, or danger, excuse me, as proof of power. This time is never different. You have a brand new technology that requires a capital expenditure build out, a sustained CapEx boom, or on the other side, sustained CapEx spending. A scale we've never seen before, it's crazy.

We know that. So Anthropx gave us a half a trillion figure just for this one company where they wanna go. So we know that their spending is gonna be absolutely ridiculous. The reason why they continue to get money and continue to get all this attention and continue to get equity investors salivating at the prospect of investing in their IPO is that this is the other part of it. Again, in every single technology bubble, every single, it doesn't even have to be technology, every asset bubble, you have this open-ended future where spending is a little bit more predictable, but revenue, we can extrapolate revenue to whatever we want.

And yes, all revenue extrapolations are parabolic. And it's not without legitimate reasons for being parabolic because the winners in these technology races tend to see parabolic growth. Yeah, but they do get a little ridiculous. This is something where the historical trend is. People put, like, if I capture 5% of the market, then it's gonna equal this. And it basically never comes to fruition. But it doesn't mean that they won't experience absolutely mind-blowing growth. And this is, when you look at 2026, these guys are on a legitimate tear.

Now, we're gonna need those financials to be audited. But I would imagine if they're off, they're not off by a lot. So they are growing very fast. As we need to spend, spend, spend in the initial years, and eventually the parabolic growth in revenue will get to the point where it outpaces spending. And therefore, as the parabolic revenue continues to rise at an extreme rate, spending left behind, suddenly this technology becomes ultra profitable. They can pay back all of those old loans. We can pay back our equity investors.

We're making enough money to make these extreme valuations worth every penny that people are paying for them. But in order to do that, in order to have any prayer of those two curves crossing or those two lines crossing, they have to invest a ton, a ton of, they gotta get a ton of debt and some equity, like I said. Staggering, unprecedented amount. They already raised some money, Anthropic did, through an IPO. But they need a vast majority of debt. So if they've got half a trillion in commitments already, let's say they get roughly $100 billion in equity, the revenue's gonna advance, but I mean, not nearly fast enough.

So let's say they get, just to be generous, let's say they get 50 billion in revenue, what did I say, 100? That's a wild amount, by the way. And for 50 billion, which is a staggering sum of money, for that not to be anywhere near enough is so dangerous. Billion in IPO equity, that leaves 350 billion for the debt market just to cover the half a trillion that we know about right now. That's likely to grow, by the way. There's no reason to believe that a half trillion is only gonna be a half trillion over the next couple years.

And then we have to have all these other questions. So you can see where these numbers just get absolutely ridiculously insane. And really, the entire issue, again, at every single bubble, this time is not different, is all about that curve. The revenue curve, which is still growing parabolically in the second curve, if it doesn't grow nearly as fast enough, suddenly you've got a bigger problem. You still have parabolic growth, and maybe you get to profitability down the road, but your profitability isn't as profitable as you hope for today.

And in between, if your revenue doesn't get as parabolic as you initially extrapolated, that means you're gonna need more debt, and a ton more debt, because every little bit of delay on that curve adds a couple hundred billion more to your overall debt pile. And it also leads to higher uncertainty. The further out in time that goes where these two lines cross, the more debt investors are gonna say, eh, the probability of you being able to pay me back is looking less and less likely, especially before we get into any kind of shocks or volatility, but the further the extrapolation deviates from its intended path, the further it deviates from the current projections, the more the debt market's gonna say, yeah, no thank you.

And when that happens, when you don't have the level of profitability that you need, and the level of profitability that you need just continues to grow and grow and grow because of the amount of debt that you brought on, the amount of service that you have to do to that debt, now you really run into a problem where it doesn't even take anything catastrophic happening. You can be growing at an unprecedented rate. You can be setting records and still go out of business. If you need debt to cover the shortfall and you don't have sufficient revenues, now you're on the hook for those payments.

Those aren't going anywhere. That's the problem. Even if their growth rate slows, you can still end up being put out of business just by the debt that you have to service. So when you look back through history, what you see over and over and over again is the same thing. Unjust exuberance where, not unjust, poorly timed exuberance and you overdo the debt and then the revenue ends up coming in, but it doesn't come in fast enough. A ton of money gets obliterated because those debts go bad.

And then the inheritance generation that comes in afterwards goes, wow, thanks for building all these data centers. We're now going to make use of them. It's crazy. And so this is the big bet. That curve, as he said, is exactly what matters. How fast does the revenue trend up? And that's not even the only thing. We also have to factor in costs are going to likely go up. The 518 billion is probably the best case scenario here. All of these companies are competing for a limited amount of equipment, which are driving up prices already, which means the 518 billion is the lower bound for what Anthropic has really estimated.

So costs go up. Revenue maybe doesn't grow as fast as projected. And suddenly there's a huge amount of uncertainty where the point of profitability comes in, which means the debt market stands back and says, we don't know if we're going to get paid back. Therefore, we're going to start slowing down what we're doing. It raises a number of potential frictions for a system, the AI bubble system, the AI capital expenditure, but whatever you want to call it, it raises a number of frictions that makes it more difficult for that to get to where it needs to be.

In order to make all of this work. And we've heard a number of. Skeptics from the very beginning, IBM CEO, I think it was last December, said there's no way these numbers work. This is what they're talking about. What they're talking about is these overly optimistic extrapolations where the parabolic growth easily outpaces spending and therefore outpaces predictable spending. That's what they sell you on. Parabolic growth in revenue, predictable spending, eventually everybody's happy because the green line goes much faster than the red line, profitability for everybody.

That is the case, but not likely in the near term and not likely in the short run, especially with- I'll point everybody back to the CEO of Cisco after the dot-com bubble. The numbers that he was dealing with were way, way, way lower than this. His was like 10x or something. He was like, you guys made decisions as investors that were absolutely insane. Think about how right my business had to go. It had to go perfectly for 10 years. It had to be like I wasn't paying anybody any money, that everything was just every dollar that we made went straight to shareholder value.

What were you guys thinking? There was no way that was ever going to come to pass. The numbers that we're now talking about with AI are just ridiculously higher than that. When he says that everything is priced to perfection, I think a far better way to say it is it's not realistically priced. Just because something isn't realistic doesn't mean that it can't happen. My own company, Quest Nutrition, grew by 57,000% in our first three years. In manufacturing, nobody in their right mind would ever say that was going to happen.

I even remember living through it, being absolutely terrified every month because you have some things you have to buy so far in advance that you're like, God, are we really going to- If I have a six-month or even four-months, it was a really big one. If I have a four-month lead time where I've got to spend the money today, but it's not going to arrive for four months, and I've been doubling every month, am I really going to double four more times before that stuff arrives?

I've got to buy month one, month two, month three, month four, all assuming that this kept up. Can you really bear doing it? Because if you stall out at any point in that growth, now you get yourself into some real trouble. The numbers that we were looking at just were not on the scale. Going back to 2008, the thing that killed Lehman Brothers wasn't some acute event. What killed Lehman Brothers was the euro-dollar market was suddenly very skeptical. They just were like, we're not going to loan you guys overnight money anymore.

That was it. That's all it took. Put them out of business. When you get a debt market where no one person thinks that they're the one making the decision to kill the AI bubble, they just go, eh, we're not willing to do it in that amount. Or we're not willing to do it at that rate. Or whatever it is that they decide that they're not willing to do. And now all of a sudden, literally, there's nowhere to get the cash from. Then you go out of business.

The credit market, the credit cycle turning in the wrong direction. You can also appreciate with a higher uncertainty exactly why this thing is priced to absolute perfection. Everything needs to go right. Spending needs to be controlled and predictable. Has to go on a predictable path. The revenue has to come as projected and as extrapolated, which is another problem because nobody has any frigging clue what revenue is going to look like. And there's a number of risks to it. For Anthropic specifically, what about the lower-tier models, the lower-cost models?

Maybe more of the business sector decides we don't need to pay for front-tier models. Let's use some lower-cost models and control our own budgets. We don't know what happens with a macroeconomic downturn. That could push the revenue curve further out into the future, less parabolic growth. There's a whole number of factors. And like I said before, spending is a huge one. And just think about what's happening right now with the interest rates on the 10-year. Up, up, up, up, up, up, up. And this is a big part of it.

These guys are just competing for the same pool of cash. And so now they're creating a problem for each other. As the cost of the debt, so the actual bonds themselves, as that goes up, everything is getting more expensive. Then consumers are pulling back. Is AI going to be one of the things that they pull back on? Do we get this battle between the Fed and the entire AI industry where they go, bro, we've got to beat China? You don't understand. This is a winner-take-all. We're just going to keep going, keep going, keep going.

So now you've got the Fed trying to break the back of inflation by telling people, stop borrowing so much money. You've got the entire AI movement saying, nope, we're going to keep borrowing this money, which then means that the treasury is basically competing with everybody, trying to get the people to come and buy the US debt. And the more that they have to raise rates in order to be enticing, to get people to come buy that debt, then the more you get this inflationary arms race between the treasury and the AI industry that's being made worse by the Fed, who's trying to do the very thing, which is break the back of inflation, that they're creating and making worse.

And so, boy, we've got, these are questions, none of this is guaranteed, but hot damn, we have a series of questions that have to be answered, and they're going to be answered in the most economically violent way you can imagine, because somebody borrowing the money coming in so fast and furious, somebody's going to lose. 518 billion today could be 618 billion next month, and it could be a trillion dollars by the time we get to the middle of next year. I wouldn't be surprised. All of these things need to go exactly right in order for all of these numbers to have even a small chance of working, and the more that it seems unlikely, the more credit investors are going to say, sorry, I don't think so, and the more credit investors pull back, well, we'll get to that.

So, like I said, the further on we get in the boom, the worse it's going to get down the road, because these numbers just get bigger and bigger and bigger, and they get harder and harder and harder to justify. But a lot of this has to do with the specific structure of the AI boom here, and it's not really all that different from the dot-com era, at least much of the dot-com era. I've said this, too. You hear people all the time say, this is a debt story.

Well, the dot-coms was also a debt story to a significant extent as well, but this one is much, much bigger debt, much, much bigger reliance on the credit system as a whole, and all parts of the credit system, including the stuff that we used to bring out back in 2006, 2007, and into 2008. Not making the same comparison, not saying we're going to repeat the 2008 crisis, it's just that all of that stuff, the SPVs, the special purpose vehicles, are central to this case, just like they were the mortgage bubble a generation ago.

And the anthropic is no different. Over the last couple months, we've heard these massive private credit deals have been put together with the anthropic at the center of it, and I'll diagram and get through it in just a second here. Another key point here, which shows this time is not different, is everybody's involved. One of the reasons why you have these credit cycles, the upswing and then the downswing, the downswing part of it is because everybody's tangled up with everybody else. It's not like you have discrete parts of it that can be separated and compartmentalized as the cycle turns to the downturn.

You can protect yourself over there because everybody over here has nothing to- no. It is absolutely entangled. Everybody's entangled up with everybody else. There are guarantees everywhere, there are collateral deals, there's distributed debts. We talked about this a while ago. You've even got Nvidia getting in on it saying, hey, look, our chips are going to be worth a lot more going into the future, and so we're going to backstop people. I think it was up to 25%. If our chips go down in value, we'll cover that gap up to 25%.

When he talks about all of this stuff is interconnected, that is one of many, many, many examples that people have started mapping out all the circular financing. Ed Zitron is somebody that's doing a really good job of showing how incestuous all of this funding is, that these guys are basically lending money. If you've got, let's say, Nvidia lending money to Anthropic, knowing that Anthropic is going to take the money that they just lent them and buy their chips, now you just have Nvidia gave money to themselves.

It's like, wait, hold on. Was there actually new wealth that was just created? Not sure. Yeah. This stuff is so tangled up. I don't even think people know who owes what, exactly when, where. You need AI literally just to track all of the connections. I'll get to it in just a second. This deal that we're talking about here and a bunch of other deals that are related to this are no different. The SPVs are back. Under the specific one that Anthropic put together a couple months ago, I believe it was in early June or July, Broadcom is backstopping some of the payments with their residual supply or residual value support.

Residual value support's the term that everybody's been using because Wall Street's great about putting together slogans. It's basically nothing different than a backstop, like I said here, backstopping payments on some of the debts. The debt that was specifically put together through private credit, you can see we've got Broadcom, which is the guarantee. Morgan Stanley arranged it. Apollo's involved. Blackstone's involved. Basically everybody gets involved in these deals because Wall Street loves these types of things on the way up. Well, they also have to, man. There's just so much debt to be bought.

Residual value support, which means that if Anthropic doesn't make its payments, because let's face it, it's a company with $5 billion in revenue and $500 billion in committed expenses. If somehow they fail to make up their payments, the Broadcom is guaranteeing part of the loan, which has the effect of allowing Broadcom to use its reputation and credit profile to cheapen the debt that ultimately benefits Broadcom, yeah, but Anthropic is basically the customer. This stuff only goes so far. Look at what's happening to Oracle. Yikes. Oracle right now has been downgraded, is one step above junk grading because they're so far out over their skis on debt.

Go through some diagrams. But first, so you can really see what I'm talking about. Let's go back to the .com era so you can see why it's really not different. I brought this up about a month ago when we went through NVIDIA. It's really the same kind of thing. But back then, it's some of the biggest technology companies that were very profitable. These are firms that had a long business profile. The idea that .coms are all a bunch of small money losing companies. In one sense, that was true.

But you had Nortel and Lucent in particular, and many, many more who are doing basically the same thing we're talking about here, but in a different way. They provided essentially direct loans to their customers. All right. So one of the big arguments right now is, all right, listen, .com, totally different. These guys were basically fake. They weren't making any money. Everybody was just so hyped about the internet, you put .com on something and it's going to be big. But AI, AI is different. There's a ton of money that's being made.

Yes, it's true, but it doesn't matter that there's money being made. The only thing that matters is that graph. You've got debt. Debt is a ticking clock, and the clock is counting down. You need revenue by that certain point. If you don't have more revenue coming in than you have need to service the debt, you go out of business. So yeah, it ain't just that. Their customers had no money. In this scenario, the customers are like Anthropic, and Nortel and Lucent are like Broadcom and NVIDIA.

Broadcom and NVIDIA established businesses, massive growth. They got profitability. Their numbers look good, longstanding profile. Anthropic, just like the .com names of a generation ago, basically no track record, no revenue. We hope that their parabolic growth actually comes to fruition, but we don't really know the case. So in order for the smaller .com businesses back then or Anthropic these days to be able to buy the equipment from Nortel and Lucent, they're going to need to borrow the money. So Anthropic is borrowing the money from Wall Street, but also borrowing the money in different ways.

But back then, we didn't have the SPVs to the same extent. They were there, but they didn't have them to the same extent back in the late 90s and early 2000s that we do to these days. So Nortel and Lucent did the same thing that we just talked about with Broadcom. In order to make the loans more economical so that Nortel and Lucent's customers could buy their products, they went into the bond market and borrowed the money themselves and then redirected those loans to their customers who then used those proceeds to buy the company's products.

Here's the big problem that I have with all of this, and this is my big problem with a lot of things, politics, world affairs, certainly the economy, is if you want to take huge risks that anybody in their right mind would stop you from taking, all you need to do is come up with a very compelling story that people can hold onto, they can explain to their friends, and then you make the reality so complicated and so hard to track that the vast majority of people just aren't going to track it.

And that is where we are with all of this financing. Even spending time on this stuff, it's hard to keep up with because the next deal is going to come out and it's going to have new weird things, somebody's guaranteeing it, backstopping it. It just gets so wild so fast that there's no way to regulate this and stop it from happening. It's so big and so complex that this is how this stuff becomes a risk of systemic failure. Very simple, in theory it's very elegant, but you can already see the problems of where this could all certainly fall apart.

If the money stops coming into Nortel because the numbers don't look as rosy as Nortel told Wall Street or Bank of New York Mellon, which is one of the biggest investors in Nortel, or Lucent, the same thing, the house of cards comes crumbling down, which is essentially what happened. Because all of those extrapolations back in the dot-com era were way, way, way too optimistic. The spending was also overly optimistic, but the revenue growth just didn't happen nearly the way that it was projected through Wall Street and through everybody else in the financial media and everybody who invested in it.

So once the money starts to dry up, customers start to dry up, but Nortel is now on the hook for their customers. All those equipment deals that were done, not only do they have less revenue coming in because their customers are all bust, they also have the back loans when they were funding their customer purchases, which is why Nortel and Lucent aren't around anymore. They went bankrupt throughout the decade of the 2000s, it got kind of messy there. But remember, we're already starting to see distress in the private credit market.

So you've got Blue Owl, BlackRock, these guys are starting to have oopsies on their books because they're already finding that some of the loans that they made have gone bad. So it's not like we don't see those early cracks. Neither one of those companies survived. And that's the basic framework that we see here. When I say this time is not different, this is what I mean. The only real true difference, which is in this case, from a macroscopic perspective isn't a meaningful difference is that these companies, the modern version of Lucid and Nortel, in this case NVIDIA or Broadcom, you can substitute Broadcom in there, they're not making these loans directly.

They're making guarantees, which make it similar, but they're not making the loans directly. Instead, we've got the shadow banks, we've got the private credit providers, at least the asset managers. The asset managers, they help set up these special purpose vehicles, the SPVs that we just talked about. They get funds from equity investors, wealthy individual institutions, they sell bonds into the marketplace, they borrow money, they create leverage. They basically use all of their resources to gather as much debt and equity investments as they possibly can. And keep in mind, a lot of this stuff came about because after 2008, regulators came in and said, all right, you guys are unhinged, man.

We cannot do this anymore. We're going to regulate you guys. We've got to get you back in line. Yes, sure, but just like with anything, what that does is it does not make the demand go away. If you can't make the demand go away, the demand is going to manifest somewhere else. And so what we ended up doing is creating all this market for private credit where people can go in, aggregate their funds, and say, yeah, we're willing to take this bigger risk. But then more and more people pour in.

It seems like the opportunity to take advantage of this hot new thing, AI, and so all the traditional lending isn't touching this stuff because it's a little bit bigger of a question mark. And so it just finds its way elsewhere. And then we start financializing everything, packaging it up, and selling it into places that it really shouldn't be sold into. But it's all complicated enough that people don't really understand what's happening. And you don't realize just how exposed you become to all of this stuff. This is why, man, I truly am empathetic to people that have the Jeff Snyder, Michael Burry approach of like, hey, can we just make this fail now so that whatever damage is done is confined to this?

Because I can see that you guys are too fucking stupid to realize you need to pump the brakes a long time ago. So let's just smash a few people in the face. And then we can start approaching this in a far more sensible fashion. That is where I'm like, everybody needs to understand what they're exposed to. Like when you guys start thinking about, how exposed am I to AI? You really want to start thinking about how to break this stuff out. We talked about this in one of our recent Deep Dive Reacts, where I'm saying, listen, there's really two different ways that you can be approaching this stuff.

You can say, OK, I completely believe in this industry, and I want to go all in. And now you're very exposed. And if it hits, you're going to make a ton of money. And if it doesn't, you're really in trouble. Or you can understand what are the mechanisms of why somebody like Trump, who we were looking at, why is somebody like Trump moving away from some of the hotter tech stocks and into things like Visa and MasterCard, which are basically like told chargers? And no matter how much inflation is, those guys are going to get their percentage.

And so they're going to grow as inflation grows. The same thing you should be thinking about now. It's not like you necessarily want to have zero exposure to AI just because it is growing so fast. But boy, oh boy, do you want to limit your exposure and not just be like, there's no way I can fail. There are a thousand ways this thing can fail. And then they're the ones that are making the loans to the customers, which are in this case, the anthropics of the AI cycle.

So they're making loans to the AI customers who can then use those proceeds to buy the products from Nvidia or buy the products from Google. All the chips are Broadcom. What is it called? The Google TRU landscape. They have all these fancy terms for it. But it's basically the same thing. The same thing is we're borrowing money to fund customer sales, customer capital expenditures, so that they can then buy the products from these companies who are providing support because they want to make sure that these deals get done.

Again, these small businesses, just like in the dot-com era, they have no track record. They have no real revenue to speak of. They're a tremendous risk. If they went into the debt marketplace themselves, they would be charged a huge interest rate. So instead, to cheapen these loans, to make these extrapolations have any prayer of coming true, they have to make the interest rate as low as possible. So Nvidia or Broadcom, they have to get involved with these guarantees because these companies have absolutely no track record.

They have no collateral other than what they can pledge, which is the stuff that they're actually buying, in this case, the chip. So basically, Wall Street came up with a way, this residual value support, to create collateral out of the products that all of these technology companies are buying from the other companies that are supplying them and supplying, in one way, the debt. So we have a company like Anthropic that buys a bunch of chips from Broadcom. They pledge those chips. The debt's a little bit, the structure's different, but we don't really need to talk about that there.

Talk about that here. The chips that they buy from Broadcom, they post that as collateral to the SPV so that the SPV makes a cheaper loan. And then Broadcom or Nvidia, whatever the company happens to be, stands by with this residual value support, which is essentially recognizing that the equipment that is posted as collateral has a very uncertain value to it. And a lot of these chips depreciate very quickly. And you don't, as a lender, the last thing you want is to have an uncertain value of collateral, uncertain collateral value, because the whole point of collateral is to be protected.

And the whole point of having collateral to be protected is because you're offering a lower rate than you otherwise would for the amount of risk that's actually being undertaken. So basically Broadcom and Nvidia have said, we'll guarantee the value of those chips because they're thinking if the customer doesn't live up to the obligation, they can't pay back the loan, we'll take the equipment back, the chips or whatever else the equipment is, servers and a bunch of everything else, maybe even data centers themselves, we'll take them back and we'll either sell them in the marketplace or more likely we'll repurpose them with somebody else who can pay.

So they're thinking that there's no loss to them because they can just take their equipment and resell it or reuse it. The SPV is made whole because they have the value of the equipment that these companies are buying, the smaller companies are buying, and they also have the residual support, the residual value support from these bigger established firms who say, we're going to make sure that you get made whole. And it all seems to work really, really well so long as those two lines actually live up to the projections.

The green line grows parabolically, the red line grows in a predictable fashion. They cross early enough that it leads to a huge enough profitability opportunity that's not completely uncertain and completely discounted down to nothing. And so we have a bunch of debt that's being raised and funding basically the same circular structure that we saw on the dot coms. All right. Shout out to Jeff Snyder, man. This guy is really fantastic. We've got all the major pieces that I want to make sure that you guys are aware of right now.

It is a huge bet. Everybody needs to understand the architecture of the world is changing. This is going to be a timing question. And given how much uncertainty there is in terms of any one person's ability to map where all of this debt is and how many question marks stand over the entire industry, we don't know how all of this stuff is going to be played out. So there's one last thing that I want to talk about. And if you can pull this up, I believe it's in the doc.

When you look back at what ended up being the disaster of the 2008 great financial crisis, it was just people lost faith. And once you lose faith, you don't make the debt available. When the debt's not available, you can't meet your obligations. And that was debt just overnight. Literally, I'm just not going to give you the debt overnight. And so you can imagine when somebody has debt that has to be paid back over a very long period of time, they have these massive outstanding obligations that right now is all ready for one company, half a trillion dollars.

You can imagine that this starts getting uglier and uglier where people just go, you know what? I'm just not sure how this is going to wrap up. And I want to play this clip because I want to remind everybody how much uncertainty there is. We're all convinced it's going to be big, but it's very difficult to say exactly how it's going to be big. It's very difficult to say how it's not going to be a problem in terms of damaging the job market. And just because we've got a bunch of machines that are doing cool things, are people going to be out of a job?

And if they're out of a job, what's that going to do to the broader economy? And if the broader economy is taking a hit because we're getting productivity, but it's out of bots and we don't have any mechanism to sort of figure that out from a UBI, like, what is this? There's just a lot of question marks. All right. So let's hear it from our lead minds. Sorry, the question was about age of abundance. What's the job of the future? Because people looking at this will be worried, particularly laptop jobs.

What are they going to do in the future? Well, I think it is worth highlighting the positive benefits of AI. That's the most political thing I've ever heard in my life. So she said, no, no, no, for real. Like, what jobs are people going to actually have? People that are using white collar jobs, they're very afraid. He, I don't know whether he couldn't hear it or what, but he basically sort of preps her. Was that a question about the age of abundance? No, motherfucker. It wasn't a question about the age of abundance.

It was a question about what am I going to do? I've got a job. It's me on a keyboard. What am I going to do with my life? And yeah, he switches the frame, but keep going. An age of abundance where we don't have just a universal basic income. We have universal high income. How exactly? People speak about medical benefits. I think the, but maybe it's worth a bit more about that. When you have the question robots and you have super intelligence, it means that everyone in the world can have better medical care than anyone here, including me.

Okay. So I think you guys get the idea. So you've even got now somebody like Elon Musk, who normally is just going to give you an engineering answer, uh, is like. Let's talk about the age of abundance. And honestly, let's talk about the age of abundance. But if you've got people that are asked to give more and more and more debt over time, and they're not seeing that the projections are being hit, you're going to lose that confidence. And so the, what we all have to be paying attention to is given the projections that they're putting out there, given the fight over liquidity that you've got going on between a government that's running $2 trillion a year in debt has $40 trillion behind the scenes of fed that is raising rates because they think inflation is being caused by something else that isn't being caused by.

If you, if Bill Ackman is right and you have inelastic demand and that they're just going to keep going crazy, taking on this debt, building out the data centers, all of that, uh, the question marks will begin to stack up. So this is one of those times where I'm not expecting the industry to get itself under control. I'm just talking to you guys as individuals, um, make sure that you think through, how do you have some exposure to what is clearly the greatest transformation of our time, but understanding that historically speaking, this is going to look something like the.com, uh, boom and bust where it is going to be the real thing.

It is going to change things and people are going to get extraordinarily wealthy, but the people that won the most were the people that had dry powder when everything crashed, still believed in the thesis, put money in and then wrote things up and given just the historical precedent for that, you could be wrong, which is why you don't want zero exposure, but man, history has a brutal way of reminding people that these are mechanisms. And if one of the mechanisms is, oh, people just have to be scared enough not to loan you money.

Doesn't take a lot to get there. And then that breaks things. And given all the other stuff that we've talked around that are also influencing this, who buddy, um, yeah, I'd be very thoughtful about, uh, your confidence in going all in on something. All right. If intelligence does get commoditized though, and models are kind of becoming interchangeable, do you think we're like, where do you think the durable profit, um, accrues? Is it more in like the electricity companies, cloud providers? Like, where do you think that ends up being true?

And you've got, uh, open weight models, no regulatory capture. Um, then it will be who's got the best product at the best price. That's going to be a huge part of what makes it quote unquote, the best product. And if these guys aren't able to drive their own costs down, uh, disrupt themselves, then the durable value is going to end up at the infrastructure layer. I mean, it's a tale as old as time. So the infrastructure certainly feels like it is likely to be a safer bet.

You may miss out on some of the upside, but the odds that something happens where we don't need any infrastructure is very low. But the thing that worries me about the infrastructure play is I wonder how much of this is going to go to space. And so if it ends up going to space or the same infrastructure company is going to be the ones that went in space. And so it's sort of geography independent, or is that going to be a fundamental shift? And it isn't like, oh, my money was in debt on a very specific New Mexico data center.

That didn't happen. I'm in trouble. Um, or are people being more broadly distributed than that? Um, so that goes back to that series of question marks, but AI is not going anywhere. It's really a question of where is that curve? Um, but that curve is everything right now. Let's talk about a pattern that is guaranteed to be killing your progress. You know what you need to do. You need consistent nutrition. We all do. You need vitamins, probiotics, greens. We all know that we should be doing more of it.

When your morning gets chaotic, you skip it. When you travel, you skip it. When your routine breaks, everything tends to break. And that inconsistency compounds against you every single day. AG1 is designed to solve the execution problem. One scoop, eight ounces of water, and you're done. You're getting 75 plus ingredients, vitamins and minerals, pre and probiotics, nutrient dense, superfoods, everything that used to require six, seven different supplements and perfect planning now happens in one drink that takes about 30 seconds to make. Right now, AG1 is giving you $87 worth of free gifts with your first subscription.

You get a welcome kit, travel packs, vitamin D3 plus K2 and flavor samples. Click the link in the show notes or visit drinkag1.com impact to claim this offer.

Go deeper

• All Impact Theory episode guides