Let's imagine that OpenAI goes bankrupt. It could happen. The host of the podcast The Real Eisman Playbook. I don't know about you, but I know I don't have 100 billion dollars to spend building data centers. You must know our next guest from The Big Short. Character.
Had to throw that in, right? I like being written that way. It's going on my tombstone. The US would go into an immediate recession, Senator Andrews, I would say.
And wow, he's crazy, he's demented, making my programming imitate data. This sector, despite the billions invested, has... Steve Eisman is with us now. Let's bring in Steve Eisman.
Steve, welcome back. You're known for identifying bubbles before everyone else. Michael Lewis wrote an entire book about it. So my first question to you is: where the heck are we in AI now?
There was a great movie with Sean Connery and Andrew... Forrester. I remember the young man asks a very complicated question and he, while eating soup, answers: 'Ah, that's not exactly a soup question.' I never forgot it. I thought it was one of the best lines in cinema history. It's not a soup question. It's not exactly a soup question, right? I see it this way: the concentration risks here are impressive. You know, if you take a step back and if someone told me: 'Why don't you analyze this software company?' Forget what it does. It's a software company, a consolidated software company. And if it so happened that the company had thousands of clients, that would be great. If it so happened that the company had only two clients, you'd say: 'Whoa, I don't want to invest in that, because if something bad happens to one of those clients, that company is done.' There's something like that happening in this whole AI story.
So, let's start just with Nvidia. Nvidia, God bless them, and I own shares, okay? When they reported earnings a few weeks ago, I think revenue growth was like... was like 110%.
So let's just take a step back for a second and say to ourselves: wait, the biggest company on planet Earth just had... forget the earnings growth, which was also great, just the revenue. Revenue growth of more than 100%. Man, that's complete madness. That would indicate that the AI story is great, until you read the quarterly report that came out that night. And I'm going to impress your viewers by saying that if you look and go to Note 7, okay? Note 7 says that 70% of Nvidia's accounts receivable at the end of July came from five clients. That's a warning sign. It's not the end of the world. Warning sign. Now let's go to the cloud giants. We're talking about Google, Meta, Amazon, Microsoft, and let's include Oracle to be safe. Gigantic companies spending massive amounts of money buying Nvidia chips and everything else. 70% of their AI revenue, which amounts to something like 25 to 30% of all cloud revenue, comes exclusively from Anthropic and OpenAI. Is everything okay? Yes.
Let me... let me repeat so your viewers understand. If you look, leave Oracle out for a second. Microsoft, Amazon, Google, 70% of their AI revenue, which amounts to 25 to 30% of total cloud revenue, comes only from Anthropic and OpenAI. If you go to Oracle, then Oracle reports a figure called RPO, which is basically a form of order backlog. When they reported earnings last year, in October, for the August quarter, their RPO went from like 150 billion to about 400 billion in 3 months.
So it was a huge jump. It was a huge jump. I mean, people went crazy. Yes. The stock skyrocketed, right? The stock... the stock skyrocketed, it went from 230 to 330 in about two days. Yeah, that's about right. And then some market analysts who are very good, who took a look, published reports saying that 50% of that RPO is from OpenAI alone. Oh my God. Now, today Oracle is worth more than 600 billion and still about 50% is from OpenAI. Basically, 50% of Oracle's future revenue comes from a company that loses a ton of money.
Well, that's what I was going to ask you. Maybe you can explain how that works, but from my point of view I don't understand where that money comes from.
We'll get there, okay? Let... let me just finish the sequence. We finally arrive at Anthropic and OpenAI. In my view, the whole chain, Nvidia's chain, the cloud giants, depends on their future growth, because they are the ones making the commitments to those giants. If they don't grow or can't pay, the whole chain slows down. Yeah, that's exactly right. So that's the risk. I would say between the two, OpenAI is probably the weaker entity, but it's not clear because we really don't have enough numbers. The only thing we know, because the Wall Street Journal reported it, so I assume it's true, is that OpenAI had something like 6.5 billion in revenue in the second quarter of this year. Anthropic was at 11-something, but that's nothing more than simple revenue. Understood. Just revenue. OpenAI lost something, had costs of about 12 billion, and the math was that in three months, OpenAI's revenue went up 1 billion and costs went up 3 billion, which we usually say is upside down. You want the opposite, not that. Right. They want it to go the other way. So, and their revenue grew 18% in 3 months, while Anthropic's revenue grew more than 100% in 3 months. So they are the weaker company at the moment. That can... that can change. My only point is that this whole sector makes me nervous because let's imagine that OpenAI goes bankrupt. It could happen. The entire United States would go into a recession overnight. If that... I mean, over time there will be much more diversification and much more press, but that will take time. So, you know, in the next year or so, these companies have to stay healthy. That's where I think we are. So, there's a concentration risk, there's a huge concentration risk.
And that brings me to the question I was asking before. What I understand is that, well, we just talked about this. OpenAI and Anthropic are not making money, they're losing money.
Ah. Ah. Ah. With all due respect, saying they're losing money would be kind. They're bleeding money. They lose a ton of money, they're kind of... we're pre-, right? We're in the pre-profit phase, right? Pre-profit. Something like that. They... they make money if you exclude all the costs. That's how they like to think about it. Exactly. That. That. Ah, we should start reporting that metric. Yes. We... we should.
Ah, but then tell me where does all that money come from?
Well, that's actually a very interesting question. I thought that most of the company... most of the money came from venture capital, and it turns out that's not true. Most of the company comes from Amazon, Google, Microsoft, and Nvidia investing in these companies, and from SoftBank. Very well. So, by acquiring equity stakes in the company, buying equity stakes, because they raise capital and they disbursed the money, whether they'll keep investing, I don't know.
That seems like a circle. It seems like the money or the commitments go one way and then the commitments come back the other way.
Yeah, there is that tendency, right? That kind of reminds me of... actually a scene from The Big Short with your character, Michael Burry. You had to squeeze that in, didn't you? I did, I had to squeeze it in. I remember you saying, ah, ah, it's like CDOs, and then they put parts of that in the CDO and then those two are put in the CDO squared. Some similarities, obviously, right? But in their defense, right, there is circularity in the financing, but as long as Anthropic and OpenAI keep growing very, very fast and people keep giving them money, the chain will hold. Yeah, if one of these two companies really drops the ball and people pull their money out or don't want to invest in it anymore, that's when the chain won't hold. That's where the concentration risk problem comes in. If the industry were much more... if... if I had said instead of 70% of the hyperscalers' AI revenue comes from Anthropic and OpenAI, if that number were 10%, we'd be having an entirely different conversation, because, uh, clearly there are many more significant clients out there.
Hmm. So, that brings me to another argument I've already heard you make on your podcast and with your guests, which is that if something... let's just plug this podcast for a second, it's called The Real Eisman Playbook. Yeah, it's definitely very, very good and it's gaining a lot of traction. I... I think it comes down to the authenticity of the thing. Thank you. I really like it. I think it's top quality. Well, my... my wife and I work on it every day. You two are the least part of this. And, um, so she does, God bless her, all... all the editing. That... that's to give you an idea of how things work. Yeah, it really is. True. So I... I'm a very linear thinker. The way I'd define it is 1, 2, 3, 4, 5, conclusion. I understood everything very well. Uh, I write like that frequently, so I write, I do, and we have the weekly summary that we launch on Friday, where I summarize everything. Perfect. Often when I write I do 1, 2, 3, 4, 5, 6, and Valerie, my wife and editor, always says, you hid the main fact again. She flips everything and says, but most people never get to the end. People get to... your conclusion is down there and, you know, it takes 10 minutes to get to the end. So she... she... she edits and she touches the business. Right. Okay. How cool. I didn't know that. It's a bit of behind the scenes. I liked that. Okay. Ah, back... back to the AI reasoning, before I lose the thread again. So, following what you were saying, if something happens with OpenAI or Anthropic, that's where the problems can be right now, when we have the concentration risk. How do you see things like the new DeepSeek model or Kimi, what's it called now? Kimi K2. Kimi K2 is definitely the best name in the market. Right, it's a very easy name to pronounce, Kimi.
