緊急特集：AnthropicのS-1提出を読み解く——Dan Primackと語るIPOの意味

This episode is brought to you by Kribble. Quick question, is your IT budget growing 30% a year? Because your data is. And now AI agents are piling on, running investigations around the clock, and running up three meters at once. Storage, queries, and tokens. Kribble, the AI platform for telemetry, was built for exactly this. Own all your data, query it wherever it lives, and let Kribble route every AI request to the right model at the right price. The best part? AI inference for workloads on Cribble is included. Models will change. Agents will change. Your data is the asset that lasts. Own it with Cribble. Visit CRIBL.io. It's never been more challenging to get front to your compute. How do you know where to look, who to trust, or how much to pay? Akash is an open cloud that provides easy access, flexibility, and peace of mind. That's how companies like Razer were able to cut their inference costs by 15 times. Visit akash.network.com for a free consultation.

That's akash.network.com for a free consultation. Well, the Anthropic S1 is here, so it's time for an emergency rapid reaction, and we're joined by the perfect guest to do it, Dan Permac is here. He's the business editor at Axios and the author of the ProRata newsletter. Dan, it's great to see you. Welcome to the show. Thanks for having me. Busy day. Definitely a busy day, Dan. Really appreciate your time. You have joined us in the middle of crazy, breaking financial news, including the dissolution of Silicon Valley Bank. Hopefully, this is an event that doesn't lead to economic catastrophe. And I think that's the right place to start. I mean, people were saying that the tech trade and really the full market hinged on the numbers we were going to see from Anthropics S1 document, which dropped at 9 a.m. Eastern this morning on Monday, October 26th. So now we actually can see their revenue, their expenses, how much money they make per employee, the risk factors, the whole deal. So after taking a look at the document, where do you think we stand on that question?

Is this economy going to hold up? It will, although I think the circularity of it is still like the issue, right? You know, you look at the contracts in here that Anthropic has with SpaceX and Broadcom and all the rest of it. All of these companies are so incredibly connected to one another. I think those concerns are still there. But look, what we got out of this is kind of what we expected, right? There had already been leaked financials. You know, obviously, revenue is going through the roof. So is spend for Anthropic. And the only thing we didn't get in here was the Q3 numbers, which have been leaked. My guess is before it actually goes public, those end up in an amended filing. Okay. And so from what you've seen so far, is this basically the company that you expected to see? And, you know, when you think about the IPO, I mean, obviously, we don't know how much they're raising right now. So that's a key question. We don't really know what their valuation is going to be. People are assuming it's going to be $100 million raised at a $200 billion valuation.

Where do you think Anthropix stands on the ability to do that based on what you saw in the document? Well, I don't think its ability is based on what I saw in the document. I think the document is not irrelevant to that, but largely irrelevant to that, right? I mean, I think it's going to be able to get that because that's what it wants, and the market is going to give it whatever it wants, just like the market was willing to give SpaceX $40-something billion to buy GPUs today. I think the capital markets have a lot invested in the AI story, and Anthropic and OpenAI are kind of the headlines of that. I think they're going to be able to get it. Now, it's worth noting, I wrote this morning that I think the Anthropic IPO, its success or failure or something in between is going to be a referendum on the AI market. I think that's true. We're on the business of AI, although I don't think it's necessarily fair. Anthropic is a company. This is a offering. If it comes in a little below that trillion or a little above, it's not that big a deal in the big picture. Okay. But you also wrote something else in your newsletter that I think we should talk about, which is that this is – and so you mentioned just now that you think the market will give Anthropic whatever it wants.

I think the money's there. Yeah, I think the money's there. And you've got so many big investors who are so, I hate to say invested, but figuratively invested in this industry and in this story. And again, everyone knows it's going to end up in the indexes, etc. Okay, but this is what you wrote today, so we kind of hold that up in contrast, and you can explain why I see a discrepancy there. So you say, Anthropics News comes on the same day that SpaceX secured $40 billion in debt financing to buy GPUs, and Meta said it will seek up to $95 billion in equity financing for its AI build-out. All this poses a capacity test for Wall Street. So basically the question is, is the money there? Now, you've already answered it, but explain how we're seeing both a capacity test and Anthropic will effectively get whatever it wants. Yeah, well, I think it's a capacity test, and I think Wall Street is going to pass said test. I think it's going to pass it with an A or maybe a gentleman's B. Yeah, so I think there's plenty of money there. Look, there was questions, I don't know if you remember, when SpaceX went public, there

was all these questions of, well, you know, SpaceX was public, can anyone else get any money for anything, right? Like, look at all the money. But, you know, folks rotate out of other trades and they put money. I think there is plenty of cash here, but it is – I think if you add – if you assume that Anthropik gets their $100 billion and you add those other two deals in, it's nearly – it represents nearly 1% of U.S. GDP. Not quite. It's like 0.7. But like it's an extraordinary amount of money. But yeah, I think there's the capacity. And part of that is because those three deals you talk about, SpaceX hitting the investment-grade debt market, Meta hitting kind of the regular equities market, and then Anthropik, which is kind of going for this new equities market. they're all going for different pools of capital, even if they're within the same institutions. I think if this was three $100 billion IPOs coming out at the same time, then I think the passing grade would be a little more questionable. Now, how is the Anthropic raise that different from the Meta raise? Because they're both trying to raise off equity. Yeah, different. Basically, it's just kind of how people are allocating.

A lot of the investors who are going to be in an IPO or the Anthropic IPO, You're talking the Fidelity's, the Wellington, the T-Rose of the world, who are really putting this in as a new name into their books, whereas Meta is an existing name. So even though they are both equity from a structural perspective, it's usually coming out from different managers who are literally managing, if not different funds, at least different pools of capital within funds. Okay. And then one more question about this, specifically on the finances. You know, you said basically, you know, to get the money to anthropic companies will rotate out of other companies. I mean, to do that, it takes like a real belief in Anthropics business and Anthropics itself in its risk factors. Hey, look, we're nascent. We're trying this out for the first time. We don't really know what's going to happen. So talk a little bit about why it why you know investors will feel so compelled to give this company what it wants based off of the fact that obviously it growing tremendously but there is that risk There is that risk. And look, I mean, I've never seen more risk factors and more pages of risk factors in my life.

I started writing them down on Twitter. I got to 80, and I just couldn't stop. I had to stop. It was so many. I mean, I think they could have just stopped with we could end humanity, and thus the capital markets are kind of a secondary concern. But the reason I have confidence in it is a little bit because of SpaceX, right? Now, SpaceX comes out, and you could argue it's this diversified business that has launch and has Starlink. But the valuation that SpaceX got, the money SpaceX got, was really based on the AI compute piece of things. And the market was very happy to give that to them, even though they didn't really have a demonstrated business on AI compute. And Anthropik's a little bit different, right? It's buying the AI compute more. But it's still it's the same story in general. Everybody is buying growth. And Anthropic has shown it's got extraordinary growth. Right. I want to talk to you now through talk with you about the expenses here and the free cash flow side of things. So Rittenhouse Research pointed this out today. They say for the first six months of 2026, Anthropic generated positive free cash flow of $3.4 billion.

So it's $3.9 billion of cash flow from operations, less $482 million of CapEx. So talk a little bit about Anthropic being cash flow positive. And is all the compute, the money that they're spending on compute factored in that doesn't seem like it? No, it doesn't seem to. And the compute spend is complicated when you go through the S1 a little bit because so much of what you have to think about as a prospective investor, if you care to really be doing this as a traditional investment, not just a momentum play, is the contracts going forward, right? It's the compute spend that they have pledged over the next six months, year, two years. I think in the S1, it goes out to the end of 2028, I think is maybe the longest. So it's a little hard to figure that stuff out. So you've got what they spent in the first six months, great. But with these companies, and when you look at the way the revenue is growing, the way the business is growing, the way the valuations are growing. I think the trailing indicators are almost become irrelevant, right?

We're talking about through June 30th. Those numbers don't seem to, to me, matter all that much. I mean, they matter, but they don't matter that much. The last three months is going to matter a lot more, and then everybody will get new numbers. You know, we're already in Q4. This industry is changing more rapidly than any other industry we've ever seen, right? Usually, you know, you hope for small growth quarter after quarter, year over year. The year-over-year numbers become irrelevant for a company like this. What Anthropik did in 2025, honestly, who cares? Right. But the growth is pretty crazy. But again, just to get back to this, because I feel like it's important to ask the critical questions about this, even though the numbers are big. If they have these massive compute outlays in the future, again, if you're an investor, how do you think about this company? I mean, clearly they're opening up a new category, but to me, like the sort of determination of what box to put them in, how you think about it, when so much of the action is going to happen next year and the year after and the year after that.

And like you said, if they're going to move money to go invest in this company, how do you square that? I want to say it should be hard, except I don't think you're going to see any investors view it as hard because I don't think – and this is kind of more pragmatic than balance sheet, right? I think most investors are going, most institutional investors, you know, have clients and clients are going to be outraged if they're not in the, not only just the hottest IPO, but one of the, you know, what's going to be one of the, you know, 10 most valuable companies in the world, most likely when it goes public on day one. So I think that unless there is some flashing red light in there, that was something that people weren't expecting yesterday and they saw today, I don't think it's going to impact whether investors go into this and potentially leave other things. And I understand the market goes up, so you don't necessarily always have to rotate completely. But I just think everyone's going to still buy into this because everybody is buying a story more than anything else. And it's the same story, arguably, to some extent, they bought with SpaceX. And I think it's the same story they'll buy next year with OpenAI.

And I think investors will look at this and say, all right, we're going to be wrong about one of these, but we need to take the basket because the basket as a whole is going to be extraordinary for us. And if one of the groceries falls to the floor and slaps like a tomato, we'll take that hit. And it's really to know which one that's going to be. And it's not going to be because of some of the numbers we're seeing today. Yeah, I mean, what is the story? You said it's the same story they're telling about all three. What would you say the story is? I think the story is that Anthropic or OpenAI, pick one of these companies, is kind of going to become, we talk about foundation models, but I think the story is this is going to be the foundation for business and society going forward. Truly, truly foundational, much like it's complicated because people always talk about the Internet and there's these, you know, the dot com boom and the question, are these really kind of the C-Lex of the time? You know, the companies that were really putting fiber in the ground, which didn't work out very well for most people. It was stuff built on top of them. That's possible.

But I think the difference here is that these companies themselves are actually building a lot of applications, whereas the C-Lex of the day and the fiber layers didn't build anything on top of it. They were just actually the utility. In this case, Anthropic is trying to be both, and so is OpenAI. Right. I mean, the revenue growth itself is extraordinary. I mean, they went from $41 million in revenue three years ago to $16.3 billion in the first half of 2026. Correct. Yeah. That doesn't happen. It doesn't happen. And what's, I think, almost more remarkable is that it's only probably about maybe 18 months ago that when I would speak to investors in OpenAI, they felt that this race the frontier model race was already determined they they did not see they knew anthropic but they didn't see it coming because when you would talk to open ai investors about this and i'd say yeah but doesn't anthropic you know it's going to be kind of going after more of a b2b marketplace you guys are more known you know for chat gpt and you know my mom using it or me using it you know as a consumer the answer would be yeah but owning the consumer is what

matters they would point to google right you know from years ago owning the consumer is what matters The B2B place is going to be small, and they misread that extraordinarily. Yeah, $6.6 million in revenue per employee. How does that rank with what you typically see technology companies IPOing with? I haven't done the math on it, except it's more. And by exponents, I mean, $6.6 million per employee, that's an insane number. It's just an insane number. It's even insane for Anthropic, right? I think last year didn they say it was like the same number was 1 I saying this now off the top of my head but it was something like that It was in the ones It a crazy number I mean think about that million in September 2025 Yeah Yeah no it a ridiculous number It just shows how fast they are scaling on the top line. Yeah. And what's interesting also is that Cloud Code and Cloud Cowork have been a big part of their ability to scale revenue. And I think that Cloud Cowork itself is actually growing faster than Code did.

I think you're right about that. I don't have the numbers in front of me, but I think you're right about that. I mean, I think that you're going to – well, it will be interesting to watch with this company over the next several quarters, assuming they go public and we continue to get quarterly earnings from them as a regular public company. I think a lot of this stuff is going to keep flip-flopping back and forth, just like we've seen flip-flops back and forth between them and OpenAI when it comes to who's actually in the lead, either in terms of model performance or in terms of revenue. I think you're going to see the same thing internally at a company like Anthropic. You know, they spent an enormous amount of time, to me, in the S1 talking about biology and bio efforts, et cetera. And not that that's making them any money really yet. But, you know, you almost read this and think, man, in three or four years, Dario kind of thinks that might be the core of their business in terms of where their money is coming from. Yeah. I mean, to me, the most interesting thing is in the last version of this that we got, there was two customers making up 25% of their revenue. Here's the update today that I think was the most newsworthy. No single customer represented more than 10% of our revenue for the nine months ended September 30th, 2026.

Approximately 79% of our revenue came from customers outside of our top five customers for the three months ended September 30th, 2026. I mean, there's a real diversification of revenue that they have going on there. Which they need, and that's, I mean, it's a very long tail, but I mean, which is what a company like that's going to need, right? If you want to be the foundation of business and of society, as I said, you need that, right? You need everybody using it or at least close to everybody using it. Yeah, but they do have some concentration in the hyperscalers. I don't know what you'd make of this. 48% of total revenue comes from Anthropic Google and Microsoft marketplaces. I guess that's still coming from a diverse base of customers, but it is interesting that they rely so heavily on the hyperscalers, two of the three of which have been invested in them. Yeah, I mean, I'm not terribly surprised by that when you think about what the business is. But you're right. But what the question becomes, and it's not really in there, but I don't even know how you do this, what are really the end users of that? As you say, Amazon and Google marketplaces, okay, that's everyone, right, essentially.

And so, I mean, it's not surprising they're concentrated with the hyperscalers. Who else is going to sign giant contracts with them up front and who's going to kind of go into them? So it doesn't surprise me, but it's, again, it is this broader diversification, though, of everything from consumers to the business use. And, again, just to go back to the business use thing or the B2B piece of this, they made a bet that most people weren't making a couple years ago to start there. Or if not start there, to really hit there hard from a sales perspective. And it has clearly paid off. And again, it's worth remembering the decision to go public before OpenAI or the decision to go public now and not push it in the next year was not a sure thing. There's all sorts of reasons. People for two months have been doing this. Will they or won't they? Should they or shouldn't they? Every time revenue comes out, obviously, when the safety stuff came out, the fact that they're going forward, I think, is, at least from their perspective, is a show of strength. Yeah. And I mean, one of the things that, you know, obviously you point to the revenue growth, which we've discussed.

And then there's the compute efficiency, which is very interesting. This is a chart that came out in terms of the revenue per dollar of compute that they're making. In 2025 Q1, it was $0.40 per dollar. So they're losing money on every dollar of compute. Today, it is $2.08. That's wild to be able to reverse that the way that they have. I mean, obviously, that's the cloud code and cloud co-work effect. It is. It's also, it's interesting, you know, there's been so much talk or a lot of talk, you know, with the data center build out and the enormous amount of money that's gone toward it and gone toward energy, towards data centers. You know, what if all the, you know, what if it becomes, what if data centers become just so much more efficient, not incrementally, but become so much more efficient that we, or GPUs do, that we don't need all the stuff or if the models do. And, you know, for Anthropic, that's a big bull case for them going forward, right, that this is all going to become so much more efficient, which means that their spend is going to get to go way down, at least as a percentage, right? Obviously, their spend will continue to go up as more people use it,

and there's more problems, et cetera, et cetera. But if these models become much more efficient and, you know, don't need to use as much, that's their big problem, right, is their spend side. It's not the revenue side. It's the spend side. The efficiency is kind of the ballgame for them, and if they can point to increased efficiency going forward, that's going to make these actually viable businesses, which right now they are still unprofitable. And again, Anthropic isn't just going public because it wants headlines. Anthropic is going public because it needs money. Yeah. Although, let me read you this from Semi Analysis. I want to get your reaction to this one in terms of cash on hand. Semi-Nalysis pointed this out. As of June 30, 2026, we have raised an aggregate of $141 billion of capital and cash and cash equivalents of $112 billion in short-term investments. Oh, and had cash and cash equivalents of $112 billion and short-term investments of $19.3 billion. So that means that what Semi-Nalysis is reading there is that they've only burned $9.7 billion of that. And they also had a couple of profitable quarters.

What do you read on that front? I would have to look at those numbers, which I hadn't. Yeah. So assuming that... Yeah, go ahead. They are committing the money. They are committing the money. Well, they've got huge contracts, obviously, going forward, huge spend contracts going forward. That's interesting if they've got that much cash on hand. But they are still... They might have had a couple of profitable quarters in there. They are still unprofitable as a whole and are expecting to spend money. One thing that they're going to spend money on, and they've started doing this, OpenAI has done it more than them, is probably going to be acquisitions, right? As I said, for a company like Anthropic to be more than the utility, they're going to have to build stuff on top of it. And they are. They're building a lot of stuff in-house. They're also going to buy a lot of that. And you could do that via stock. It's one of the good things about going public. You now have a public currency to trade. They're going to use a lot of cash to buy stuff. And they're going to need cash to buy stuff. Yeah. Yeah, I just looked at the S1. It looks like that matches up with what SemiAnalysis was going to say. Okay, one last thing here. It's in the risk factors. The risk factors are very interesting. Way up top is the compute part of it.

If we are unable to secure adequate compute or if we fail to direct our compute effectively across our workloads, we may be unable to train models that remain competitive with those of our competitors that adequately serve our customers. You know Dan we I think over the past week or so just seen these up and downs with the neoclouds and the market even questioning companies like Meta it seems like who been eager to raise more money to build more compute Just talk a little bit about the – does the AI game just boil down to like, do investors trust you enough to buy the compute you need to survive? I think that's a piece of it. I think that's a good way to sum it up. I have no concerns that Anthropix is going to be able to buy the compute, right? because, you know, for starters, it's one of the biggest dogs on the block. They are going to be at the front of the line. It's not going to be, you know, a new startup. It's going to be them, particularly if they go public and have all this extra cash on hand. But look, I mean, it's not the most interesting risk factor in there. But no, I think it's probably the most practical risk factor in there.

Yeah, I mean, the most interesting risk factor is like our models can cause you harm, which they just, you kind of knew that was going to be in there. Oh, it can't cause you harm. It's end humanity. Destroy the planet. It's an existential threat to humanity. I mean, by the way, I love when you read it, and I'm going to paraphrase this, but it's kind of like, you know, that's a concern internally and externally. We could end humanity as we know it, which could impact our financial performance. Yeah, no shit. That could impact your financial performance. Yeah. Yeah, I'll read it. concerns have been raised both within our organization and by other researchers, regulators, policymakers, and industry participants that sufficiently advanced AI systems could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponized, or allowed to operate with a high degree of autonomy. Dan, in the history of financial reporting, has there ever been a sentence like that in S1? No, because I don't think Oppenheimer went public. So, no, I don't. No. No. No. No.

Nothing like this. And there was a lot. I will say there's been a lot of commentary over the last, what it's been, two months since the Coxon thing came out. Well, you know, they're going to have to resolve this issue before they go. Not possible. Well, whether possible or not, clearly not. They're not even really trying to. And that's not even a knock on them. I don't think they can. But, no, this is going to be in there. And again, whether you invest in the IPO or whether you are invested in the IPO or not is not going to impact or change the fact that Anthropic could not destroy the world, which will, again, render the rest of your portfolio meaningless. How does Anthropic's performance here trickle to OpenAI? I mean, it will, right, because it's the most direct comp. I mean, SpaceX is kind of a comp, and depending on how you want to suss out the meta piece of it, but I mean, this is the most direct one-to-one, right? These are essentially the same company. Again, SpaceX has Starlink, and it's got other things. Launch, it's got Elon.

These are the most, so I think for OpenAI, I think whatever Anthropik is trading at, from even a multiples perspective, is going to impact what happens to OpenAI if and when it goes public, unless there's a stage or innovation or change in the interim between the two. Okay. You closed your newsletter with plenty of companies flip their S1s and then get cold feet. Anthropic can't afford to be one of them in terms of both its balance sheet and its reputation. Just elaborate on that. Balance sheet, I mean, despite the cash on hand, it needs money. And like, again, you see OpenAI out there right now trying to raise $30 billion from the private markets because they both need money. They're cash hounds. And then from a reputation standpoint, this is just too big for them to not go, right? You know, this if Anthropic weren't to go public outside of some, you know, catastrophic 9-11 hurricane sort of thing, you know, everyone agrees has nothing to do with them. I think it would be a massive stain on them. And then there'd be questions of what can you ever go out? And I think I think it could start snowballing on itself.

So I don't think they're going to let that happen. I think it is better for Anthropic to go public at a lower valuation or a lower price than what we're all expecting than to not go at all and to wait. This isn't a company like Aura. I don't think they can afford to wait. Okay. Now that they started. They couldn't be pumped, but now we're here. Yeah. All right. Really last one for you. We've seen an unprecedented run of fundraising from both OpenAI and Anthropic over the past two years. They're going to go public. They're going to get the IPO. It's going to be out there. They'll obviously have the opportunity to do, I guess, some more equity raises. like Meta is doing right now. Eventually, Dan, does the, you know, they have to, like, make a profit and, you know, start being able to fund, being self-sufficient and be able to fund these investments out of the money they're making? Or can this, you know, sort of treadmill go on forever? Eventually they need to make a profit.

I mean, you know, I think a little bit about Amazon with this, right? You know, it was unprofitable for so long, and Bezos was able to talk the story about reinvesting. And the difference, a difference, though, being the quantums of capital weren't nearly as much. Even when Amazon was growing, you know, by leaps and bounds, they weren't losing or they weren't spending this kind of money. So, yeah, do they eventually have to become profitable? Yeah, they have to prove that these can be viable businesses. I absolutely think so. I'll add one thing, you know, when you talk about the money and the money that these companies have raised. One thing that's interesting to me about the S-1 is that none of that's really in there with a couple of small exceptions. It doesn't talk about it. And so what that means is, for those of us who care about kind of the venture side of things and deals and potentially kind of the virtuous flywheel this could create, right, because funds get their money back and can raise more funds and they can fund other startups, it's going to take a long time for us to actually figure out what venture firms have made on Anthropic despite the price. Like, you can look and be like, oh, you invested a $200 million valuation or $200 billion valuation. It's now worth $2 trillion.

We don't know anything about preference tax. We don't know anything about conversions. We just don't have it. None of that's in the S1. For a smaller offering, it would be. But one of the reasons it's not in the S1 is you can be a venture firm that's literally invested a billion dollars into this one company, which is more than most VC firms that ever even raised total. And you're not mentioned in there. You're not even close. You could have done 5 billion. You're not in there because you're too small compared to the expected valuation. The only firm in there is Spark, and that's because they've got a board seat. So there's some related party transactions. Yeah, no investor had more than 5% of the company. which also it's worth noting despite you know that hundred billion dollars plus invested in it no investor has any actual real say again you have one VC on the board who in theory could be voted down she's one voice on the board uh nobody has any control not even control but really has any say over anything that happens at this company which is probably the most valuable investment for most of them the website's axios.com dan permac is the business editor you can find his pro rata newsletter highly recommend you subscribe dan great to see you thank you for joining us

Thanks so much. All right. Crazy times ahead. Thanks, everybody, for listening. We'll be back on Wednesday with another interview, this time with Ethan Malik from Wharton, talking about where the AI productivity is, and then Ron and I will be back on Friday. Thanks for listening, and we'll see you next time on Big Technology Podcast.