You need to focus on Anthropic and OpenAI. If anything bad happens to one of those two companies within the next year, everybody's in trouble. You know, if it's five years from now, the business will be much more diversified. But within the next year, given the concentration levels, you know, again, something like 70% of AI hyperscaler revenue comes purely from Anthropic and OpenAI. That's incredible concentration risk. So if one of these two companies gets into trouble, everybody's in trouble.
Welcome to Prof G Markets. This week marked the end of the third quarter and there is a lot to unpack. We had AI researchers warning that the technology could pose an existential threat to humanity. The Fed raised rates for the first time in three years. Treasury yields hit multi-decade highs and we started to get a sense of what we're going to see in the Anthropic IPO. Those are just a few of the highlights. So, we wanted to take a step back and break down what happened this quarter, what the biggest takeaways are, and what we can expect heading into the final quarter of the year. So, to help us make sense of it all, we are joined by a guest who has been helping us navigate this year's headlines and separate the signal from the noise. Here is our conversation with Steve Eisman, investment analyst, portfolio manager, and Big Short legend. Steve, great to have you back on the show. I want to start with a clip that you went very viral for. This was from your recent interview on CNBC where you all were discussing all of these doomers' threats from these AI companies saying that, you know, there's a 10% chance that humanity will go extinct because of AI or even higher probability ratings. You had a very interesting take on the matter. A lot of people heard it. I'm going to play it and we'll get your reaction.
The idea that this whole Terminator thing is nonsense. So I think something completely different is going on. And what I think is happening is that token maxing is over. The open-weight models are taking big market share. I think these companies are very nervous. They realize that there are no moats around their business whatsoever. And they're trying to manufacture a crisis that will create regulation and that they think they can then manipulate to create the moats, to create the duopoly that they want.
Talk more about what you mean by all of that.
I didn't know that went viral. Thanks for letting me know.
You know, I grew up reading a tremendous amount of science fiction. Like an insane amount. And I'm just amused by the fact that I think all these people who live in Silicon Valley basically read almost as much science fiction as I did. They just take it very seriously. And that's funny, but I think also true. I mean, the idea that AI, first of all, there's absolutely no evidence whatsoever that AI is anywhere close to AGI. None. Absolutely none. I mean, what AI is, is like a next-word retrieval model. That's what it is. It doesn't think. And there's no evidence at all that it's ever going to think. Now, maybe one day it will, but even so, the idea that somehow that is going to literally create Terminator, I think is insane. But, by the way, I rewatched Terminator 1 and Terminator 2. Still good. Okay, I enjoyed both movies enormously, even though the sci-fi effects obviously weren't so great anymore. You know, when somebody says that the world's about to end, I just don't take it seriously. So, I have to ask myself, okay, if they really don't believe that the world's going to end, I mean, maybe some of them do, but I don't think any of the senior people really do. What's really going on? And you know what has changed in the last six months or so is, you know, if we were going back six months, token maxing was going crazy. I mean, employees at firms were basically told, use AI till you're ill and then use it some more.
And we're going to track you and we're going to put you on a leaderboard. I'm going to track you to make sure that you are using AI 24/7.
And what happened was they blew through their budgets within months. So sometime this summer, I think token maxing probably ended. I think everybody who uses AI, which is a lot of enterprises, have gotten a lot more cost-conscious. And at the same time, the open-weight models have really come into their own and, as far as I can tell, are starting to take very, very large market share. So if I was Anthropic and OpenAI, I'd be nervous because I had everything to myself and now I don't. So what can I do about that? So if I can manufacture a crisis where there's regulation and then the regulators come in and I can manipulate them so that we don't want the open-weight models, all of a sudden I've got a duopoly and that's what I think this is really all about.
Well, this is very relevant to the news that we just got this week, which is we now have some insight into how Anthropic is doing as a business. This was the reporting from Reuters. Who knows, by the time this episode comes out, maybe the S-1 will actually be fully out. But what we know about Anthropic as a company: last year they generated $4.6 billion in revenue, up more than 1,000% from 2024. Extraordinary revenue growth, but their operating losses came out to more than $8 billion. And their net loss, which has a giant caveat, which is that a big portion of this number was a non-cash charge tied to revaluing these financing instruments. But still, the net loss was almost $42 billion.
Eight billion on an operating basis.
So, we should probably pay most of our attention to the $8 billion operating loss.
Eight billion is bad enough. We don't need to pursue the 40.
Yes, we'll stick with the eight because there are no asterisks there. No caveats. That is a real number. That is how much they're losing or they lost last year just from day-to-day operations. What do you make of those numbers? What does it say about the AI business? And does it confirm your suspicions related to this crisis manufacturing?
I don't think the 2025 numbers matter.
I think what I want to see is, I actually have another conspiracy theory, which is the reason why Anthropic is going public now. And I'm not 100% sure I'm right here, but if I am right, it's a good one, which is that I think Anthropic might be going public now because the first half of the year looks really good because of token maxing and the lack of open-weight models. And maybe when they report their third-quarter numbers, which they won't have to report when they do the IPO because it's before the third quarter, the third-quarter numbers might show something of a slowdown because of the end of token maxing. I actually think by the fourth quarter we'll definitely see a slowdown, and so I think they needed to go public now. I'm more interested in comparing the first half of 2026 versus the second half.
Do you take anything from the 2025 numbers, or do you think that we purely need to understand what's going on right now? Because to be clear, we don't know anything about how they've done in 2026. We've heard some rumors in some of the reporting. There was, of course, the reporting that they might have achieved adjusted operating profitability on a quarterly basis. But then we kind of dig into what does the adjusted operating profitability actually mean? And there are some big questions there. I mean, what are you looking for right now, and is there anything we can glean from the current information that we have to understand how this business is actually doing?
I don't think the information that came out in Reuters about 2024 and 2025 is going to be all that relevant to people. People are going to want to look at 2026 and they're going to want to compare what the company's saying about the second half versus the first half of 2026. I think that's the most relevant.
This crisis manufacturing. Do you think that these companies are actually in trouble, or do you think that they think that they're in trouble?
I really don't know. I think they're nervous. You know, how in trouble are they at this point? I don't know. I think they see that they're losing market share. I think that, you know, you're starting to see signs of a price war breaking out by all the various players, which is very bad. I think what I said on CNBC that this business has no moats, I think is really true. You know, Google with its search had a moat that was insurmountable for decades. None of these companies have any moats. You know, one day one model is up, next day another model is up. One day Grok is up, the next day somebody else will be up. You know, there's nothing protecting what you're doing.
Does that indicate then the possibility of a bubble? I mean, you are famous for predicting the bubble in 2008. Where are you in your perspective on the possibility of the existence of an AI bubble at this point?
What worries me is the concentration risk because, I mean, if you look at Nvidia, for example, and you look at, you know, Nvidia had revenue growth of over 100%. But 70% of their accounts receivable were from five accounts. You know, if you look at the hyperscalers, 70% of their AI revenue is from OpenAI and Anthropic. The whole chain basically flows to Anthropic and OpenAI. If those two companies succeed, you know, we'll be back in a year from now and saying, 'Wow, AI is really, really triumphing.' But if there's a problem with those two companies, then I think the whole chain is in trouble.
I'm wondering what you make of current valuations in tech because, you know, I have been thinking about and looking at this circular financing problem for a long time and the concentration risk problem, which to me, as you say, it seems like a very big deal. But I wonder the extent to which that is actually priced in. And something we've been talking about on the show is the fact that Nvidia, on a forward earnings multiple basis, is actually quite cheap compared to the past several years. Do you think this is a dynamic that Wall Street understands quite well? Do you think that investors have their heads wrapped around these risks and do you think that they are pricing it in?
I love when people ask me that question because my answer to that question is I have no freaking idea. I mean, my answer is like we would need to have like a massive group therapy session where we're all one million of us are in a room and we pass the baton and say, 'Hey, what are you thinking?' I literally never know what is and is not priced into the market. I never... You know, what I do think is that there's a narrative about AI that's positive and that there's a narrative about AI that's negative. And it's not clear to me which one of those is going to succeed. But I do think at some point within the next six months we'll have a better idea.
Would you not say that in 2008 you did have an understanding of what was priced into the market, specifically that there was a lot that wasn't priced in?
Oh, absolutely. I learned this because I remember when Bear Stearns, they had like this fixed-income fund that blew up in May of 2007. And what that fund had done was invest in subprime paper. And I remember there was a sell-side analyst who I was very friendly with, who worked at a firm that was very heavily involved in subprime. He was a financial services analyst. And so he asked me what was going on. And I said, 'Well, why don't you just get in the elevator, go down two floors, and go talk to your own desk?' And what I realized was that the reason that the equity people didn't know what was going on was because it wasn't an equity story. It was a fixed-income story. And unless you immersed yourself in the world of fixed income, there was no way you could figure out what was going on. That's not true here. You know, this is a tech story which will, you know, I think everybody understands there's massive concentration risk. I mean, I think even the people who are very bullish would say that. And at the end of the day, I think the statement that at least for now it depends on OpenAI and Anthropic being successful, I think is true. Now, maybe it'll be successful and it'll all be okay, and then again, maybe not. I don't know yet.
You've also talked about some of the off-balance-sheet debt issues that are becoming more and more pervasive, it seems, in the AI world, in the data center world. Could you speak a little bit to what we're seeing there in terms of debt issuance, what the risks might be, and how systemic and important it is to the AI buildout right now?
Well, is this something like $500 billion worth of AI debt being raised this year? You know, how much of that is off-balance-sheet? I don't know yet. I don't think it's insignificant. I think the reason why they're doing some of these off-balance-sheet shenanigans, is the best way I could put it, is they're trying to preserve their credit ratings as much as possible. So if you can get it off balance sheet, it's like puff magic. It doesn't exist. And, you know, the rating agencies won't count it. I've seen this movie before. I've been in that movie. So, I'm kind of appalled. I mean, there was the... Meta did a last year, did an almost $30 billion deal where they're building a data center in Louisiana. And they created some off-balance-sheet vehicle where basically all $30 billion of debt or something like that doesn't show up on their balance sheet. And if you go into Meta's 10-K, there's an entire like three or four paragraphs where the auditor goes through a tortured description. I mean, torture doesn't even begin to tell you what's going on here, but a tortured description of this transaction and why it's off balance sheet. So, in my wrap for last week, I quoted the entire, all four paragraphs. I mean, it took me about three or four minutes to read the whole thing. And then I said to my viewers, let me translate this into plain English to tell you what's really going on here. What the auditor is really saying is we really don't know if this thing should be off balance sheet or not. We probably think it shouldn't, but Meta told us that it's okay. We're really nervous about it, but Meta said, 'Chill, bro.' And so we said, 'Okay, we chilled.' And that's basically what the 10-K said. You know, that's what happened in Enron. That's what happened in the SIVs that Wall Street created to get a lot of stuff off balance sheet. Like I said, I've seen this drill before. I mean, presumably it doesn't end well.
It seems that the problem whenever something goes wrong, whenever there is a problem in the market, it's because there is a lack of accountability on anyone's part. Whether it's the companies not taking accountability for their own, as you call it, shenanigans, or the auditors and the ratings agencies whose responsibility is to accurately assess what level of risk these companies are taking on. And if it's your view that the auditors are saying we don't really know, they told us it's fine, so it's fine, I guess the question is how large of an issue is that?
I don't know how much is off balance sheet at this point. That's my only caveat to it.
Meaning you're not ready to determine.
I don't know the size of it yet. I'm trying to figure out, you know, of the $500 billion, how much of it is off balance sheet and how much of it is on balance sheet. I don't know yet.
And depending on how large that number is, what will that mean for your analysis?
I think what it would mean is that these companies are very, very nervous about their credit ratings and they're trying to do everything they can to preserve them. And if that means creating off-balance-sheet vehicles that really shouldn't be off balance sheet, but they get their auditors to agree that it's off balance sheet, so be it. That's what I think it means.
We'll be right back after the break and we have some exciting news. We have been nominated for three Signal Awards. So, please go vote for us at vote.signalawards.com. Type in Prof G Markets in the search bar and there you can vote. We've been nominated for Best Money and Finance Podcast, also Best Daily Podcast, and also Best Video Podcast. Please vote for all of them. We really appreciate it. We'll also leave a link in the description to make it easy for you.
Support for this podcast and the following message is brought to you by E*TRADE from Morgan Stanley. Simplify your finances and discover the convenience of investing and banking all in one place with E*TRADE from Morgan Stanley. Choose from a wide range of investment choices and award-winning banking solutions together in one platform. Plus, get up to $1,500 when you open a brokerage account with a qualifying deposit today. Learn more at etrade.com/offer. Banking products and services are provided by Morgan Stanley Private Bank, National Association, member FDIC. Terms and other fees apply. Investing involves risks. Morgan Stanley Smith Barney LLC, member SIPC.
Support for the show comes from VCX, the public ticker for private tech. For generations, American companies have moved the world forward through their ingenuity and determination. And for generations, everyday Americans could be a part of that journey through perhaps the greatest innovation of all, the US stock market. It didn't matter whether you were a factory worker in Detroit or a farmer in Omaha. Anyone can own a piece of the great American companies. But now that's changed. Today, our most innovative companies are staying private rather than going public. The result is that everyday Americans are excluded from investing and getting left further behind while a select few reap all the benefits. Until now. Introducing VCX, the public ticker for private tech, now available wherever you buy stocks. VCX by Fundrise gives everyone the opportunity to invest in the next generation of innovation, including the companies leading the AI revolution, space exploration, defense tech, and more. Visit getvcx.com for more info. That's getvcx.com. Carefully consider the investment material before investing, including objectives, risk, charges, and expenses. This and other information can be found in the fund's prospectus at getvcx.com. This is a paid sponsorship.
We're back with Prof G Markets. Increasingly, the story of stocks seems to really be actually a story of bonds or a story of yields and rates. You wrote that the 5% on the 10-year is the quote, 'Rubicon for the market.'
Could be. No, nobody... Let's be clear, nobody ever knows until after the fact.
Yes. You said, quote, 'Unless something happens that pushes rates back below 5%, a market correction seems imminent.' Could you explain what you mean by that? Why is it that these high yields could result in a market correction? What does a 10-year at 5% plus mean for everybody?
So obviously it hurts housing because, you know, you were taking out a 6% mortgage before, now it's 7, 7 and a half percent. You're not taking out a mortgage so fast, you know, you're not going to buy a house. So that hurts the entire housing sector. Now, the housing sector hasn't been doing too great anyway, but it's kind of like, you know, one more nail in the coffin. You know, then any corporation that's going to raise debt, it's more expensive. And, you know, all this data center debt, you know, CoreWeave raised debt six months ago at 9%. Today, it'd probably be 11. So things get more expensive. They just get more expensive. The economy gets creaky when rates start to go higher. So I think that's what's making everybody nervous because, you know, it makes the deficit bigger and it makes AI debt more expensive. It makes everything more expensive.
Do you see the AI debt problem as the largest problem in terms of how higher yields could affect the economy? Because as you point out, you know, increasingly this AI data center buildout is becoming dependent on debt issuance. And I guess these companies have a lot of money, but there are companies like CoreWeave which have a lot less money than the big tech companies and the hyperscalers, the true hyperscalers. And so you'd have to assume that if they're paying more and more in interest for the debt that they are issuing to build the data centers, at some point that might not be sustainable anymore.
No, I agree. I mean, think of it this way. Let's imagine, you know, a company that has an LLM model or an agentic AI model. You know, they've invested a ton of money in it, but they're expecting a return of multiples of what they invested. Otherwise, why do it? When you're building a... when you're CoreWeave and you're building a data center, there's only so much you can charge to whoever is going to, you know, employ the data center. You know, there's a return. This is real estate. This is a real estate transaction. So if rates are 100 basis points higher, it just makes the transaction more difficult. You know, you're going to have to charge your customer more. Is a customer going to want to pay? I don't know. But it just makes it more difficult.
There was some other data center news that was quite interesting to us. One of them was the fact that Oracle sent this force majeure notice to Blue Owl that was building their data center campus in New Mexico. Project Jupiter, I believe, was the name of the data center.
What did you make of that news? Essentially, what it meant is that they're saying that if something bad happens, we're not liable, is my understanding. What did you make of that news?
I think Oracle is petrified with their credit rating. You know, right now Oracle's credit rating is triple-B minus. Triple-B minus is one level above junk. You know, Oracle does not want to get downgraded to junk. They do not. And they're going to try and do everything they can to protect their credit rating. I think that's what that was about.
Is that company... would you say Oracle has sort of the weakest balance sheet in the AI trade right now?
I don't know if it's the weakest amongst the larger companies. It's the weakest. You know, if I were to compare it to CoreWeave or somebody else, I don't know if I would say it's weaker than those companies. But, you know, like I said, when you're rated triple-B minus, things can get ugly.
How close to those levels do you think some of these larger tech companies are? Because my understanding is that they're pretty well off unless this off-balance-sheet stuff is a lot larger than we think. My understanding is if you're Meta, Google, Microsoft, you're okay.
I think at this point they are. I don't lose sleep over those companies. I worry about Oracle, CoreWeave, and all the other companies at this point. But, you know, like I said, if something bad ever happened to Anthropic or OpenAI, there'll be no place to hide because they're just too important to the entire AI ecosystem.
Just in terms of macro, I mean, getting away from the AI stuff, we had inflation was last at 3.4%. We had our first rate hike. What were your takeaways from that rate hike? And it seems that we're probably going to get another one. What does it have you thinking about? Were you surprised by it at all? What are your reactions?
You know, for a time the inflation numbers were actually getting a little bit better, but, you know, oil prices are higher. So I think the Fed was basically taking out some insurance. They may take out some more insurance. I'm much more concerned by what's happening with the 10-year. And I think part of the reason why the 10-year is so high, other than that, you know, the oil prices are up and people are nervous, is that AI is a $500 billion issuance this year and it's having a crowding-out effect of the Treasury markets. You know, if there was no AI, none, the 10-year would be much lower.
Can you spell out why that would be the case for us?
If you're Microsoft or Meta or Google, well, not Meta, but Google, and you're building a data center, that's a long-term project. You know, you're not raising overnight debt, you're raising long-term debt. You're raising 10-year debt, something like that. So, you know, the Treasury has a $6 trillion deficit this year. Now I add on top of that $500 billion in AI debt and there's a bit of a competition which didn't exist before. So I think one of the reasons why the 10-year is as high as it is because, like I said, there's a crowding-out effect where AI debt is crowding out Treasury, which is a crazy statement because nobody has crowded out Treasury before, but now they are. And that's why, you know, Scott Bessent tried to pull a rabbit out of the hat by trying to buy $6 billion worth of long-term Treasuries in something called Operation Twist, where he would issue short-term debt to buy long-term Treasuries. And it's failed miserably. So, the Treasury Department has lost a lot of credibility because when he started, I think the 10-year was like at 4.75 and now it's at 5.21.
Yeah. What do you make of his comments about that buyback program? Because his view was that yields were wrong, that the markets were incorrectly saying something about the sustainability of the US economy that wasn't true, and he was trying to correct it back to the truth. Do you agree with him? What do you make of his comments?
I think it's a weird argument. I mean, the Treasury markets are pretty efficient. They're more than pretty efficient. They're really efficient. So to say that they're wrong, what are they wrong about? I mean, there's a war going on. The oil prices are higher. You know, diesel fuel is at an all-time high. I mean, what exactly is... I don't know what he's talking about. What is he saying, that inflation is about to come down? Well, if inflation is about to come down, then Treasury yields will come back down when that happens. I mean, I don't get the argument. I really don't.
Just in terms of this AI bubble story, which has been around for a very long time at this point.
Let's not get carried away. It's been around for six months, which in our world feels like an eternity.
