US GDP this year will grow about 2%. Half of it is purely AI capex.
Today's guest needs little introduction to anyone who follows financial markets. Best known for calling the 2008 crisis and making one of the most famous bets in modern Wall Street history: Steve Eisman.
Steve Eisman is one of the financial world's most outspoken voices. You may have seen him portrayed by Steve Carell in The Big Short.
One of the lessons of the financial crisis is that incentives trump ethics every time.
He said, "If you can breathe, you can get a subprime mortgage loan."
I literally think anybody could have gotten a mortgage loan back then.
Today he joins us to break down how the US economy has become increasingly dependent on the AI boom.
The only sectors that I can see that are not correlated at all to the AI trade are staples, health care, and real estate.
Where he sees the biggest risks and opportunities across global markets.
At some point of higher interest rates, the stock market corrects.
And whether the warning signs emerging in private credit point to bigger trouble ahead.
Just like private equity has had explosive growth, private credit has had explosive growth. The issue is...
He's here now for our three pressing questions, where we'll get into three hot topics in finance.
My view is, in a way, the entire stock market is one trade because everything is correlated to it, and that trade is AI.
Steve, thanks so much for joining us.
So we were just chatting off camera, and I was so surprised to hear that this is a very new venture for you with the podcast. You only started podcasting last year, is that right?
Beginning of April. Just I think my first interview was literally like days after President Trump announced the tariffs. And I had a great guest on, Dan Clifton from Strategas, who's like a political analyst. But that was complete accident. I had just arranged to interview him because I know him and I like him. And it was like it could not have been more crazy relevant than that moment. That's how we got started.
I was just saying I couldn't believe that because of the huge community that you seem to have built in such a short amount of time. But just to kind of set the stage here, I did a little bit of a deep dive into your background, and I realized we all know you now as this finance guy, right? But I understand this wasn't the case originally. You actually originally studied to practice law. So tell us what was that like for you, and what made you want to make the switch over?
Well, I think I might have been one of the worst lawyers in the history of the law.
Oh, oh, I was. I was a corporate lawyer, and I was utterly miserable. I was miserable. I really just disliked it intensely. And my wife, Valerie, was a tax lawyer, and she was pretty miserable too. And so she arranged for us to go to this institute called the Johnson O'Connor Institute, which doesn't exist anymore. It tests your aptitude, which is kind of your intelligence, but it's more like what are you sort of naturally gifted at? It's a full day of tests. And at the end of it, when they gave my report, they said to me, "You were very, very good at generating ideas, but you are not detail-oriented at all. The worst thing you could be is a corporate lawyer." Which I found very refreshing because I actually now understood why I was miserable. And so I managed to get a job on Wall Street as a junior analyst, and in 1992 I got promoted to be the financial services analyst at Oppenheimer, and that's how my career got started.
You know, obviously, Steve, there's no way we'd sit down with you today without talking about what happened in the years of 2006 to 2008, when you made what some would describe as the call of your career, right? Betting against subprime mortgage bonds here in the US well before the housing crisis became obvious to everyone else. Now, those of us who remember that time or watched The Big Short movie know exactly what happened. But for those who aren't familiar, can you kind of just talk us through a bit of the story, you know, in broad strokes?
Sure. Let's go back in time. In the mid to late '90s, there was, I call it, the first generation of subprime mortgage companies. And I covered them all as a sell-side analyst. And so at one point in 1998, 1999, the entire industry effectively went bankrupt for reasons that are very complicated and we don't need to get into. And then, starting in 2002, the second generation of subprime mortgage companies went public. The funny thing was that most of them were run by the same people who ran the first generation of subprime mortgage companies. They just changed the names. You know, people think that I'm like this genius who foresaw everything. The point was that I had seen the play already, and I was waiting for, you know, Terminator 2 to have the same plot as Terminator 1, because I'd already seen Terminator 1. And so I was just sitting waiting. Well, if you're waiting for something to happen, and then the evidence that you need happens, you're kind of ready. And effectively, what happened was by the summer of 2006, the loans that had been securitized early in 2006 were going bad very, very rapidly. And the reason was that underwriting standards had kept deteriorating. And that's what got me into starting to think about shorting subprime paper.
You had a great line in one of your recent podcast episodes where you were talking about this loosening of underwriting standards from 2002 to 2006, where it was, you know, basically loosened up to such an extent that it was like you said, "If you can breathe, you can get a subprime mortgage loan."
I think that's actually true. It's not an exaggeration. I literally think anybody could have gotten a mortgage loan back then.
And why do you think people weren't paying attention to this, or perhaps weren't doing anything about it at the time? Just for again, those who don't understand.
Sure. I think one of the lessons of the financial crisis is that incentives trump ethics every time. And everybody in the entire subprime chain, from the lenders to Wall Street, which bought the loans, securitized them, and then sold them, everybody was compensated on volume. They were not compensated on quality. So they were all incentivized to keep doing it, and the only way to keep generating more and more volume is to loosen underwriting standards because you increase the pie. That's what happened.
You've said that covering subprime as an analyst early in your career taught you that Wall Street didn't care what it sold. How did that change the way that you approached the markets, and how did you also frankly handle criticism, right? When everyone else was maybe, you know, seeing something different or telling you to look the other way.
I remember I had a meeting with the head of asset-backed research at one of the big firms. So in the meeting was this person, a guy, and it was me, Porter, Vincent, Danny, we're all in the room talking to this guy. And it got a little heated because we're like telling this guy, you know, the world's about to end, and he's telling us everything is fine. And so I asked him, "Well, why do you think everything is fine?" And he said, "Well, since World War II, housing prices in the United States on a national level have never gone down." And I looked at him, and I said, "Is that a law of physics or just an accident of history?" He thought it was like a law of physics. And people, you know, look, when times are very good for very, very long, people can't imagine that it'll ever go bad.
So much that I could go on with about that, but you know, just to kind of turn us back to the timely questions for today, because I know you spend a lot of time every week unpacking what's going on in markets and kind of like decoding, you know, the main questions that your viewers have about finance. I want to get to our three pressing questions today. First up is AI. You know, frankly, this we're going to spend most of our time today because it obviously is one of the biggest talkers right now.
Yeah, especially with the eye-popping valuations, investment boom we've been seeing. You have predicted, Steve, that if this were to stop, the economy will fall into recession almost immediately. Tell us why.
So we had on our podcast the economist Torsten Slok, who's the economist at Apollo, who's quite talented. And he pointed out that US GDP this year will grow about 2%. Half of it is purely AI capex. So I think you need to divide the economy and the market, because the market story, in some ways, is even worse. But let's just focus on the economy first. So the way I look at it is it's like a chain. So Nvidia sells all the chips. Now, people looked at Nvidia's numbers when they reported them, and everybody thought they were incredible, which they were incredible. I mean, the largest company on planet Earth, revenue grew more than 100%. I mean, it's an astonishing statement, and it's unheard of. But buried in the 10-Q is a little footnote that says that 70% of the company's accounts receivable is from five accounts. So yes, it grew over 100%, but it's highly, highly concentrated. Now, who is that? Nobody really knows, but it's either immediately or indirectly the hyperscalers. So Nvidia sells the chips, the hyperscalers buy the chips, and so far that's okay. Where I think things start to get dicey is there have been some reports by Wall Street firms that have been very good, which have said, I'm going very careful how I say this, 70% of hyperscaler AI revenue is just from Anthropic and OpenAI. Okay? Now, that number, whatever it is, is equal to 25 to 35% of their total cloud revenue. Well, that's huge. What that tells you is the reason why the hyperscalers are buying so many chips is because they have all these commitments from OpenAI and Anthropic. So if anything bad ever happened, and I'm not making this prediction yet, but I'm just saying if anything bad happened to OpenAI or Anthropic, the whole chain stops. Because let's say hypothetically OpenAI went bankrupt tomorrow. Think about what that would do to Oracle. Oracle has a $600 billion backlog. Half of that backlog is from OpenAI. So the entire chain of buying all these chips, of Nvidia growing revenue 100%, goes into reverse if one of these two companies gets into trouble. That is the problem. Now, between the two, OpenAI is the weaker company. That is unquestionable at this point. That could change, but it's definitely cheaper. They have less revenue, their growth rate is slower, they lose more money. And both of them are completely dependent. Since they lose money like crazy, I was on this show with another podcaster who said, "You know, these companies lose billions." And I said, "If only they lost billions." So that's the weakness of the AI story. It's highly concentrated. At the end of the day, it's dependent upon two new companies who lose money. And if anything bad were to happen to either one of those two companies, it would be a bad situation. If that were to happen, and the entire AI story reversed, the economy would go into recession immediately.
If you're a founder like me, a lot of your mornings tend to start the same way: gathering intel from all over before you make your first decision. Checking analytics, taking your emails, pulling up spreadsheets, and opening about 47 different tabs. That's what caught my eye about Perplexity AI, the computer, the sponsor of today's episode. Instead of opening every dashboard by yourself, Perplexity AI computer connects to the tools you already use and builds a command center that gives you a complete picture of your business. You can have it pull together your revenue, ad spend, inventory, email performance, and more, and flag what's changed, and even send you a daily report on Slack. The best part is you can actually put it to work. So let's say you want to build an ad campaign or work through a CRM task. It can actually do this for you. It's a smart idea. Instead of being locked into one AI model, computer sends your project to the right one for the job. It essentially unifies all the different AI capabilities you normally use across these tasks into a single system. I just tell it what I need done, and I let it do its thing. If you're interested, for a limited time only, our listeners can try Perplexity AI computer for free. Head to pplx.ai/btb. That's pplx.ai/btb to try computer today. Most AI tools are still answering questions, while Perplexity AI computer is completing full projects. Okay, here's something my younger self could have really used: being more intentional about where I keep my money and how much I'm paying in fees. Right now, I'm saving for my honeymoon safari in Africa, so I've been thinking a lot more about this. That's why we're glad to have Chime sponsoring today's episode. Chime is changing the way people bank, and they're not like traditional banks that gatekeep the best perks unless you have some massive balance. Chime offers the most rewarding fee-free banking with no overdraft fees, no monthly fees, and no minimum balance fees. And there are a bunch of other perks too. You can get 5% cash back on your Chime card in a category of your choice, say gas or groceries. You can also build credit through everyday spending without a credit check. And if you're ever in a pinch, they've got this feature called SpotMe, which lets you overdraft up to $200 fee-free. For me, the best part: you can grow your money faster with a savings rate that's nine times the national average. Join the millions who are already banking fee-free with America's number one choice for banking. Head to chime.com/btb. That's chime.com/btb. Sign up now, it only takes a few minutes. Chime is a fintech, not a bank. Banking services and Chime card provided by Chime's bank partners. Qualifying direct deposits required. Terms and limits apply. Go to chime.com/disclosures for details.
When you first said that, I was just struck by that, right? And I thought, wait, are we being too alarmist here? And then obviously then you backed it up by saying that AI capex alone accounts for about half of US GDP growth this year. That is astounding, right? Because hyperscalers, I think you've said as well, are projected to spend, and it's been shown, about $700 billion this year.
Yeah. So you've pointed out much of the risk here is partly because the boom relies heavily on just two AI companies, OpenAI and Anthropic. Can we talk a bit more about, you know, the underlying risk? Because again, like I know that you are always very careful by couching everything you say.
I'm not making this prediction yet. I'm just saying what if, and I'm very nervous about it.
So the what if here, right? Like, what is the potential risk here? Because for instance, I think that, you know, a lot of analysts have noted that if cheaper open source, or let's say Chinese models are, you know, extremely competitive in this space,
trigger a price war and hurt these two companies, then the massive capex that we're seeing by big tech obviously could face a severe downturn. Is that your view?
I think there is a risk of a price war. Yeah, there was earlier this week at the Goldman tech conference in San Francisco, the CFO of OpenAI said that they had just cut price on their most recent model, that it increased compute their business by like tenfold. I have no idea what that means, but let's take it at their word, but they cut price. Does that mean Anthropic's going to cut price? The other issue, which is just as important, is that a lot of this revenue growth that Anthropic and OpenAI have experienced is because of what's called token maxing, which is basically I work at Uber, let's say. I'm the tech guy, and I am under orders, literally orders by my superior, to use AI 24/7. Forget the cost. That lasted through the summer or much of the summer, and that's over. People are much more cost-conscious now.
What's also interesting, as you know, we're taping here at the New York Stock Exchange today, and both of these companies we're talking about, OpenAI and Anthropic, are obviously set to go public soon. How much does that even change the equation in your view? Is this something we should really be paying attention to? Does it change anything basically for investors, the fact that if they become public suddenly, or what's your view?
I can't wait for the S-1. I mean, I just can't wait.
Special episode coming up. Yeah.
I mean, we know actually so little about the financials of these two companies, but when you go public, you got to open the kimono, so to speak. They have to show the numbers. So we'll have a much better understanding of what their numbers are really all about when they go public.
I think it's so important that we're talking about this because, like, candidly, and I'm sure you've seen just anecdotally, I have so many friends right now who are just so excited and all in on AI. So many of my friends across different sectors have been like, "I want to quit my job and try to get a job at Anthropic, you know, they're paying so well," and such. I'm sure you've seen that craze, right? And so it's just so important to contextualize what's really going on here, right? Because there's so much hype and excitement, but at the same time, like, people then don't stop to really think about what's going on. You know, there's been a term...
Can I interrupt you just for one second? Because I think this is something we have to explore as well, because what we've just spoken about is the impact of all this on the economy. But I think in some ways the impact on the market is bigger.
The stock market. And the reason is that when you look at the economy, 70% of the US economy is consumer. So that's why it's only half of the growth in GDP, even though it's a very, very big number. But when you look at the market, you know, if Infotech is 38% of the S&P, and then if you add in all the companies that are really tech companies but are not technically in Infotech, like Google and Amazon, etc., you get over 50%.
Yeah, I was going to say, what is the like ratio of the Mag 7's like total value?
About 38, 40%. There are 11 sectors in the S&P. The only sectors that I can see that are not correlated at all to the AI trade are staples, health care, and real estate. That's it. Everything else is correlated to one degree or another. Like utilities, which are obviously building power plants for AI, to banks and investment banks, which supply the financing, to industrials, which are building the stuff. So much of the market... My view is, in a way, the entire stock market's one trade. It's one trade, because everything is correlated to it. And that trade is AI. So as long as AI works, the market will go higher.
But if something were to happen to AI, it's not just going to be a recession, it's going to be a massive correction in the stock market.
What are the main bottlenecks of this growth, right? You've talked about like power availability as one of the biggest constraints, for instance, and we've seen that, right? Obviously, like they're all rushing to kind of build out and whatnot. You know, can you kind of talk through some of the main risks that could happen here, you know, to kind of constrain the growth that everyone's so excited about?
Look, I think the big risks are the end of token maxing and what that does to revenue growth. I actually think it's possible that Anthropic is rushing to go public because they want to go public before third quarter numbers, because the growth is explosive through the second quarter and may have flattened out starting in the third quarter because of the end of token maxing. So that's one thing. The second thing is a price war, and the third thing, which we can't really handicap because it's so political, is how difficult it is to build these data centers. I mean, we've never built data centers this big.
Or this many. Right. So look, I'm no engineer, but what I'm hearing is that things are getting difficult. That doesn't mean it won't get done, but it's hard.
There's a term increasingly coming up right in this whole world, which is circular financing. You've talked a lot about this. For those newer to the subject, can you kind of briefly walk us through what's going on with that and the risks involved?
So a lot of the circular financing is happening via Nvidia, but others as well. So just for your viewers, a circular financing from Nvidia would be: "I'm Nvidia. I sell chips. So I either make an investment in company X, or I lend money to company X, or some version where I give money to company X, and then company X buys more of my chips." That's circular financing. Now, I have concerns about it, but I don't think if Anthropic and OpenAI go on to glory, it won't matter. So at the end of the day, you know, people write a lot about circular financing. It's intellectually interesting to me just to see what Nvidia is actually doing. But...
Because have we seen something like this before? Like, forgive my ignorance here, but have we seen something like this before in regard to circular financing, or is it kind of a newer trend?
There've been circular financing in the past. I mean, not to this massive degree, but like I said, I don't see... I mean, if tomorrow OpenAI goes bankrupt, you know, Nvidia's going to have problems with all this circular financing that it's done. But as long as those companies are okay, that's not the bottleneck.
Yeah. What metric are you watching most closely to kind of gauge whether this enormous spend we're seeing on AI infrastructure is actually generating adequate returns? What are like the numbers that you're watching the most?
