Seems like there's a lot riding on Anthropic at this point. Going well.
Yes. And Anthropic is a peculiar company. I've been wondering what their credit rating is going to be because they're a negative EBITDA company no matter what monkey business they play with their valuation or with how they capitalize things and all that, which makes it—I'm not sure how they rate anything above CCC. And if the ratings agencies allow them to go above that, they're basically corrupt and disgusting. But fundamentally, these AI labs don't really make sense economically. Anthropic has $413 billion of non-cancellable commitments. There is no getting around that. And people are saying, "Oh, Microsoft will let them off. Oh, Google will let them off." No, they won't. That's—that's
Well, but even if they don't let them off, you know, if you look at the revenue numbers that have been leaked, right? I know there was a lot of talk about the revenue growth decelerating, but it's still growing. They are still, you know, they are still bringing in revenue. And of course, the hope on the part of investors is that that growth in revenue will perhaps reaccelerate, that it's lumpy quarter by quarter, so that it could see a reacceleration. In other words, that they're going to be bringing more money in the door.
Well, that's kind of the point though. We don't know their cash flows. And that whole $80 billion net loss, but $72 billion was convertible, what have you, there are rumors going around that some of that is credits from hyperscalers for compute. So, I think it's reasonable to believe that Anthropic is likely sheltering some of their compute spend in there. The reason I say that is they have a $3.7 billion—$3.75 billion compute agreement with SpaceX every quarter. That means that across Amazon, Google, and Microsoft, they're spending $6 billion in compute. That does not make any sense to me. That doesn't make—it's
It's way too low. OpenAI burned, I think, $12 billion worth of compute just in the first quarter. And that was on like four, seven—I forget the exact numbers of revenues reported by The Information. Something doesn't make sense about these numbers. And I think Anthropic has to put their S-1 out there today if they want to make it before Thanksgiving.
I really want to see what's going on with these cash flows because there's something off with this company, because let's say revenue accelerates, what have you. So will the costs.
And you've got this company—oh, they only burned, let's say, $9 billion in Q3, I think it was. How is that possible with the big commitments they have? Something doesn't line up with these numbers, and so revenue may accelerate. It doesn't really seem like it is. The reports I've seen have just the annualized run rate, which is a made-up number anyway, is just still—it's just barely above plateauing. And Anthropic isn't a company that is going to do well if it's just okay. This is a company that needs to be basically 10 times the size in about three years to keep up with its own compute commitments and with the ridiculous amount of building around it. I mean, they have, I think it's $161 billion just with Broadcom. There's an Apollo deal that was $35 billion with Apollo.
Like a hundred each with what, Microsoft? Or not Microsoft's lower—it's with Alphabet and Amazon. Yeah.
And it's just—it's astonishing amounts of money that don't line up with what's been reported so far. So I really can't wait to see this S-1. I think it's time we stop playing with run rates. Stop playing with non-GAAP adjusted.
Yeah, this is—this is—this is something I've said, that just give me some GAAP metrics. Give me a number that a first-year business student can understand what it means and what it means for the balance sheet. But then let me ask you this, because to your point, our team reporting that we could get that S-1 as soon as today if they want to make it before Thanksgiving. Probably going to have to, or at least get it in soon after that. What are you expecting to see when you do actually see those? What are you going to be like? What's your—what's your, you know, F search?
Command-F cash flow. I want to see how the cash has moved around the company. I also want to see how they actually classify some of these adjusted numbers because right now everyone's guessing. Everyone, everyone out there, myself included, and I've read basically every story about Anthropic's revenue for years, and we are still guessing. I think that the journalism around this company has been between horrifying and upsetting in the amount of times that adjusted stuff is just reported as clean profit. So, I think it's time we need to put away these childish things. It's time to stop talking about run rates. I never want to hear about them again. I don't think that they are a reflection. This is a company—these aren't cutesy little startups anymore. These are basically the inspiration for the large—what I would say is the largest misallocation of capital in history. Why are we talking run rates? These are not cutesy children. These are big startups that have made the largest
agreements of anything I think in history. I mean, I think Microsoft's OPEX is like $150 billion a year. How are they possibly going to keep up? And I don't know if you saw OpenAI had to drop the valuation, re-rate their
And you know how bad it is when you go to the Middle East and you're like, "Yeah, I want 1.5 trillion." And they're like,
Well, they've got some different—they've got some different capital pressures of their own. I mean, what—what—what do you mean?
What—what is OpenAI specific? Oh, you mean—oh, the Middle East. My bad. Yes. Yes. Oh, there's something less cash—there's something going on over there. Yeah.
Do you think there—do you think those two companies are sort of connected in the same problems, or you think OpenAI and Anthropic have their own specific issues that—that could cause their own problems within those two companies?
I think they're basically the same company. I think one of them is run by basically cultists, the effective altruist types who poisoned Anthropic. There was a 400,000-word Harry Potter fanfiction written by Eliezer Yudkowsky that Dario Amodei is in. So that's what Anthropic is, but they are basically the same kind of book.
Have you—have you read that whole thing?
No, I have—I could literally do anything else. I have run into a wall at high speed. But the thing—the thing is
I think people also need to realize how important these companies are because they are 70 to 80% of all compute demand. They are 70-plus percent of all of Microsoft, Google and Amazon's revenues. Microsoft, I think consensus estimates have 70% of their growth in the next few years coming from cloud, and most of that growth is coming from OpenAI and Anthropic spend. Google Cloud—UBS's Steven Ju said 48% of Google Cloud's revenue, not AI, all of cloud will come just from OpenAI and Anthropic spend. These are load-bearing companies. These are companies that we literally can't afford to have not spend. And that get—but it gets back to the things we've heard about Anthropic. It's like $9 billion of compute spend in the core. What do you mean by that? With all of these, unless every analyst is wrong, unless all of these compute agreements are so far in the future that we may as well not be talking about them, something does not line up.
