Anthropic’s IPO filing: $4.6B in revenue, a $42B loss and a warning about humanity

Anthropic. We got the news late last night. [music] Reuters has some of the reporters have reviewed a a copy of the IPO prospectus for Anthropic. Long awaited. We thought we weren't going to get that from the company officially until at least post the midterms. >> Well, we're not we're not I mean it's not official. >> not. That's true. We're really not. Reuters has apparently reviewed an early copy of it and pulled out some numbers that are are really quite drastic and really quite stark. I just want to run through a few of those at the top line. 2025 revenue we keep in mind they don't include 2026 figures here. 2025 revenue nearly $4.6 billion up 12-fold from roughly 400 million in 2024. That is a massive growth. Operating loss however has grown to 8.06 billion in 2025 against 2.98 billion in 2024. More growth there in the other direction. Something else I want to point out compute and infrastructure expenses. $7.33 billion in 2025 which is about three times what it was the preceding year and more than half of the total 12.65 billion in total operating expenses. This is an expensive company to operate. They're projecting $518 billion basically half a trillion dollars in planned compute and cloud and infrastructure obligations in the next few years. It's an expensive company. >> So so to me there there are sort of two parts to this to this story. Well, there are let's call it three parts. Part number one, the financials which yeah, I mean we knew the company was losing money and I do find it a little curious here that they didn't get the 2026 numbers but they only got the 2025 numbers. I'll come back to that in a second. Second is the risks. We knew that the risk factors in these IPOs would be excessive. And yeah, I mean it was 100 pages long. And third is who leaked the prospectus. Why you know, whenever something like this comes out before it's supposed to come out, you you have to wonder who did it come from, what were their motivations, what is this do for them? >> That's right. >> Um and so those are kind of the three tracks I'm thinking about for Anthropic. >> Yeah, I mean I I this this kind of like free association here, right? This is a lot in this thing that kind of hit us yesterday. Um so big the big thing that kind of that you just touched on was who leaked it and why. What's happening today? What could be happening later today? Something in a San Francisco, maybe it's a Open AI develop develop dev day, right? Interesting that this is >> spinning some con- >> Oh, that's [laughter] >> I don't know. >> That's some heat. >> No, no, no, no, no. I'm just bringing up facts, okay? I didn't say anything about it. >> Okay. >> Why can't I spin up conspiracy? Who cares? >> Wait, correlation is not causation, so that's one thing, right? You mentioned where are the 2026 numbers? Why are those Was it a selective leak of the prospectus? It sounds like when you read the reporting that they had the full prospectus. It's not saying we saw part of it or we only saw a piece of it. We saw the >> So why isn't 2026 >> Where are the 2026 Which are the numbers that I'm really more curious about? >> We're three quarters of the way into the year, but you would think that those numbers would be there. You mentioned the spending on compute, right? 7.3 billion last year. I mean, we know they're spending a billion a year alone on SpaceX compute, right? So we know that number obviously is going up, etc. etc. But also you the the actual loss, what, 42 billion dollars for 2025? >> loss. >> Net loss. >> Yeah, sorry. Now some of that hasn't shown. It includes a chunk that's possibly will be converted to open sorry, to Anthropic stock. Some people are are saying this is a real expense. This is not some fairy tale expense. This is actual real money that could go out the door for them. This is not a small loss. This is a large loss. So it's a very interesting kind of peek into this world. It's kind of scary from for a lot of reasons. If you're an investor, this is a big bet on what the two trillion dollar valuation, what's going to happen in the future? So, I'm not saying it's a this is a you stay away from it, but it's kind of these big numbers are kind of scary to see for a company becoming a public company. >> Right. >> And we knew that the numbers would be like this, but it's another thing to see them in black and white. The one the thing I would bring up about about the the the loss that you were talking about. Okay, so again, $42 billion according to Reuters, it's includes $34 billion in an accounting charge that reflected an increase in the estimated value of financing that could eventually turn into Anthropic shares. >> That's right. >> So, my question about that is, is it recurring? In other words, is that a rolling situation over time that we could see repeated? That kind of accounting charge. >> Or was it the it was it the full charge as of right now? >> Right, right. Or is that just is that just yeah? >> of >> Um so, I mean, so I think investors, I mean, we saw from SpaceX, it doesn't matter how much money you lose. >> That's right. >> As long as you're going to make money in the future. >> And as long as the story's good. >> And that's what we don't know. One other thing I would point out um just from the financial perspective, and then I think we can get into the risk factors. From the financial perspective, a quarter almost a quarter of its revenue came from two customers alone. >> Yep. >> Now, a quarter is not I mean, it's not half. >> It's not half. >> worse. >> But it's still it's two anchor customers, and they point out the prospectus warns per Reuters that many of their largest customers are not on long-term commitments, which means they could pull their money and their spending at any moment. >> Yeah. >> That's a bit risky. >> Same with the AI compute deals. Like that's why six-month I think it's six-month window for you could pull out of the deal. Um noting that the revenue of 4.6 billion for 2025 not great. Again, 2025. >> Right. >> going to be interesting. Maybe this is this is just a lot of us you know, bloviating here. And the really the true number is at 2026 Q2 or Q3 numbers. Like that'll really sell the case for for this IPO because these all seem kind of like not great. But again, you know, it's funny. I always [clears throat] want to mention the existential risks, right? That would be Yeah. >> Yeah. Yeah, let's talk about that. >> those risk factors. >> SEC said 80 pages of the 261 pages in the filing were were regarding >> So, it wasn't a hundred. >> You were a little You were a little >> [laughter] >> in comparison regarding risk factors. >> 80 pages against 48 describing the business. >> Right. And it's also well a lot fewer than Space X than the risk factors in Space X's >> uh prospectus. But yeah, that quote was that AI could Advanced AI could pose quote catastrophic or existential risks to humanity. >> last time we saw the word existential in a risk factor? >> When's the last time you saw an IPO filing say, "We could end the human the human race as we know >> Yeah, Ross, correct me if I'm wrong. I don't think the SpaceX IPO prospectus laid out a risk of ending humanity. >> No, I didn't see that in there. Um so, >> [laughter] >> you're telling me like this is just I don't I can't believe we're in this world right now. I'm sorry. I'll take a take a step back. We're in a world right now where we see an IPO filing that talks about "Oh, we might end the human race. But hey, you still might want to invest in us, too." >> Well, and this >> You're going to make a lot of money along the way. >> Yeah, I mean well listen, listen. If you had some milk and you drank the milk and the chances are it could wipe out 10% of the human race, wouldn't you drink the milk if it would make things better? Sorry, that was the SNL's That was the >> Yeah, yeah, yeah, that was incredible. It was a great bit. >> No, you wouldn't drink that? Really? >> But then but then >> He's like, "Do the milk bit." I already did the milk bit. DIDN'T WORK. >> [laughter] >> ON A ON A ON A SERIOUS NOTE, I do want to note something we've kept talking about here, which is the idea that like, "Okay, fine. So, slow down. So, stop your progression." The problem though is that your enterprise customers, especially, keep wanting that next model, which makes it harder to slow down. Anthropic per Reuters specifically points to that and says its customer usage and as a result its revenue is driven by new models that develop, quote, at a continuous and overlapping cadence that that inherent to remaining at the frontier of AI development. Translated, we've got to keep pumping out new models. They've got to keep getting better or else the business fails. So, how do you balance that against, if you take uh Jacob Cox's word, the Anthropic researcher, a 10% chance that we're all going to die. >> So, you have all of this and I feel like there So, so the numbers here are not great. The risk factors are alarming as we know. And I feel like there's also this increasing drumbeat in the background of not just for Anthropic, for this whole industry. >> Yes. >> How much they're going to need to make over time >> Mhm. >> to actually get a return given the amount that they're spending. So, in the last week alone, Goldman, Bain is the latest. >> Brookings. >> Brookings. Like we Jake Bernstein. Jacob Bern- No, not Jacob. >> Jared Bernstein. >> And it was a Jake. All these people have come out with these assessments of this and the numbers are huge. And then today, our friend Torsten Slok over at Apollo also did an analysis and he looked at, okay, if you look at the bottoms-up analysis, in other words, if you look at the analysts who cover the tech sector specifically and they say, okay, what is operating cash flow for tech going to be? $2.4 trillion by 2028 is the number they came come up with, those individual analysts. So, that's an increase of over $1.2 trillion. However, if you look at the other sectors in the S&P 500 that are not tech, you would think that they if they're going to be paying the tech sector for this AI, their numbers would also have to go up. >> Be commensurate. >> But they're not. >> They're not. >> If you look at those estimates. So, that's just another way of slicing the question of like how is all of this going to get paid for? >> I mean, it's interesting to see if we you know, obviously we haven't seen the filing but if they mentioned the risk of open models, right? Open weight models. >> One would have to think that they would have would have mentioned that as well. >> Um because of the fact that could just take the head off that that trade, right? Talking about all this investment into physical build out for AI and there's no commensal return on the other end, who's left holding the bag, you know? It's >> Well, that so that is the most important question of this whole thing. Right? Like we we've been debating and you know, we've had this ongoing debate, okay? About the concerns about the future of AI, you know, people like Ed Zitron have been pounding the table on the capital concerns and the financing concerns for a while, okay? And then you have the investment community that is still all in on AI in terms of investing in these companies in this stock and the private credit of all of it. So, the question of it if it doesn't work, who gets left holding the bag? That's the question. >> And this is this is what I keep thinking about with all of the debt that gets issued. >> Right. >> At some point that debt comes due. And there's got to be someone holding the bag at the end of that. >> And maybe it all gets paid. >> Maybe it does. >> Maybe all of this works out and everything is fine and AI makes a makes the world amazing. We don't die from it. >> Yeah. Yeah. >> It cures cancer, it makes us all more effective, it improves our lives, everybody makes tons of money and everything goes great. >> Yeah. >> That's the that's the you know, that's the optimistic scenario.