FULL TRANSCRIPT
Slava Rubin (00:00)
Welcome, My name is Slava Rubin. I'm one of the founders here of Vincent. We are your platform for all things alternatives. Feel free to check out our newsletter, The Alternative Investment Report. we also have our podcast, Smart Humans, and with us is none other than the famous and brilliant Jan-Erik from Sacra. Jan-Erik, say hello.
Jan-Erik Asplund (00:19)
Thank you, Slava
Slava Rubin (00:20)
Absolutely. So so many people have been waiting for this IPO. Will this be the biggest IPO ever? Having just come off the SpaceX IPO, I'm not sure. But if SpaceX is the Super Bowl, this might be the World Cup. So I'm not sure which one is bigger. This is uh pretty exciting and we have a lot of information to cover here. I'll just say it in advance.
there's a lot of data, there's a lot of numbers. yes, this will be archived. You'll be able to check it out afterwards. I suggest if you'd like, you could take screenshots, you could take photos, whatever, but feel free to ask any questions along the way. So let's just dive right in. We're gonna talk about the positives, the negatives, we're gonna talk about the price, how to get in, where do we think it's going.
And obviously none of this is investment advice, so you should all navigate this as you think appropriate, but we will uh jump right in.
So we actually did a snapshot here for who's in the room. So a bunch of you are accredited, which makes sense, but it's a really high accredit amount. Two thirds of you are accredited. there's quite a bit of uh investment experience. only about a third of you are what you consider yourself beginners. And about two thirds of you, again, are thinking to invest into pre IPO in the next 12 months. Now, are you currently an investor in Anthropic? 11% are yes.
I am as well. so 89% of you are still trying to figure out if you're coming in. Now, do you expect to come in in the next 12 months? More than half are yes, very small percentage are no, and almost 40% of you are looking for this conversation and more information to figure out what you want to do next. Now, the official word from our compliance department, which is nothing in this presentation should be construed as an offer to sell securities or solicitation of an offer to buy securities.
All investments involve risk and the possibility of loss, including loss of principle, neither past performance nor forward looking information is a guarantee of future results. so let's dive right in. We're gonna be moving fast because there's so much information to cover. If you do have questions, please lob those questions in and I'll do my best to try to see that QA and try to incorporate it into the actual conversation. All right. So, Jan-Erik, contextualize. What are we talking about here?
Jan-Erik Asplund (02:55)
We're talking about Anthropic, which is one of the more unusual companies, if not the most unusual company, you know, in this big of a way to go into the public markets, started five years ago uh by a group of researchers who defected from Open AI, uh, you know, focused on making safer AI systems. so if you yeah, think back to those old times, that was the that was the the thesis of the company. It has now become, you know, the largest one of the largest.
One of the fastest growing enterprise technology businesses in the world. 17 billion of revenue in the third quarter. They just you know, we just got, you know, the data that they hit 80 billion annualized in October. Um, thousands of enterprise customers and and and finally as of Q2, you know, they're have positive operating income. so giant software company. Um
But also a hyperscale infrastructure operator, which is why we're talking about an IPO in the first place. you know, people ask, will these companies go public? and for Anthropic, you know, you might ask, why do you need so much money to go public? if you have 17 billion in quarterly revenue, and we'll look at that more. basically there's an enormous amount of compute coming online in the next several years, and even
being profitable and driving this much revenue, they need to raise a ton of cash to secure those resources that they need. So it's it's a big sort of component of the model, one we'll talk about more and one that makes it very much unlike most you know, software IPOs where you know, the you know the the additional cost of serving customers is is very low thanks to the internet. AI is very different as we'll as we'll look at. But that's kind of the overview.
Slava Rubin (04:46)
Amazing. So this should be one of the largest IPOs ever. And this company's only five years old. To put in perspective, SpaceX, which IPO'd, it took over 20 years for it to reach this enormous valuation. and this company, Anthropic, doesn't even have the quote unquote Elon Musk you know, factor. So it doesn't get that extra bump.
So it's pretty remarkable that we're talking about this in such a quick period of time. Let's go on to the next slide. So we actually um analyzed the S-1 because I know all of you probably would like to read every you know sentence of the S-1, but you haven't had the time, maybe. So we've actually done that and we've bulleted it out into the bullish and bearish points. Now each one of these points
could go into significant detail. So this will feel rushed. And we're gonna double-click into some of these points in the following slide. But I first want to go quickly through some of the highlights, and obviously you should on your own research this more. But some bullish points from the S1. the fastest revenue ramp on record. So we're looking at huge growth rates coming from Anthropic. Number two, the they're actually adjusted profitable.
for two quarters running, which in comparison to some of the other players out there, there's nothing even close to this sort of profitability in the AI space. And they're serving customers at an 85% margin, which is again crazy margins for a company going public at this scale. The land and expand uh at scale that's never been seen, 900% net retention, uh revenue retention, and they have so many customers. We'll talk about that a little bit further.
They don't really have massive revenue customer concentration, which is uh kind of wild. Now they have quite a bit of cash, but we'll talk about is it enough? And the cheaper models are just growing the pie. There's just more customers out there that want to use AI. And subscriptions are taking off you know, in a really remarkable way. No single customer to lose.
Which goes back to the kind of customer concentration, which is, I think, a really important point. They barely have tapped enterprise budgets, which is kind of wild. you know, Claude is less than point one percent on average of the Fortune 500 total expenses, which is a bit nuts to be talking about over a trillion dollar business company or potentially two trillion dollar enterprise value company with such little you know, tap from the large enterprises.
And the gap losses um are mostly on paper. And we could talk a little bit more about that as the as we continue. Let me ask you really quickly here, Jan-Erik, and I know I just like went through so many important points in just like two minutes. What would you say is the most important number on this slide that's the most exciting, the most interesting, the most unique, or you're most bullish about?
Jan-Erik Asplund (07:59)
Yeah, I think, if I set aside revenue, which I think is the clearest kind of and we'll look at their 2026 numbers, which are just insane. I I really like the number nine, this point about how much of the budget they take up at Fortune 500 companies. Because if you look at startups, if you look at like Y Combinator, you know, the last class of startups, I bet you would see, you know, token costs that greatly have outpaced payroll.
I I would imagine it's threatening rent at some of these startups, you know, 20%, 30% of the of the total expenses. And you think about what's what's going on in the enterprise, you know, up market at these Fortune 500 companies, they've barely sort of gotten into the you know 2026 era of agents, you know, long running agents with sub agents. that stuff is still all to come and I think will be transformative.
And we'll bring that number up from, a tenth of a percent up to, you know, two tenths and three tenths. And the goal is to, you know, get to the point where it's substituting for humans and it is eating up payroll. And so I think that is one of those clear numbers that to me shows how potentially early we are. Obviously, this is like a high risk kind of bet, but I I think that's one of the clearest ones for me.
Slava Rubin (09:21)
We've been doing this for so long, it's crazy. But I agree with you. Uh really I kind of see the number four land and expand model, the eight, no single customer to lose, and nine, the barely tapped enterprise budget, just speaking to how diversified the revenue base is, which shows, in my opinion, how much upside there could be. Because if the revenue keeps on growing, the you know, the usage keeps on growing, it's
It's there's not really this customer concentration risk, which is pretty wild. Let's go to the flip side, which is let's go on the next page and show some of the risks, which is it's definitely pricey, and we'll talk about how pricey. It's not red here because it's not the priciest AI stock, but it's definitely not cheap. Growth seems to be cooling off, especially if you look at the most recent quarter. margins are already starting to compress. This really is shaping up.
To be about a two, three, not many horse race. So is there a price war that's going to commoditize this and to really erode margins? really a large risk is, you know, Washington and the courts and the political risk and really the the the concern of the public as it relates to AI, and what is the existential risk there as it relates to that? the there's this talk about the
nearly $900 billion of commitments for build and obligations for AI data center usage and compute, et cetera, versus the $33 billion of revenue. How do you make that all work? the stock comp you know, as part of the expenses and the lockup you know arriving early in terms of being able to have some employees trade early. The shareholders don't really control the board because the founders
have a special class of shares that let them control uh where the company goes. And it's mostly pay as you go. So it's not a lot of locked in revenue for forecasting the future. So I mean, again, a lot of details here, a lot of numbers here. And we've just tried to simplify the entire S1 for you into here are our concerns. What would you say is your top concern on this page?
Jan-Erik Asplund (11:43)
For me, it's gotta be the maybe not so much the growth numbers in absolute terms, but the reasons for the growth slowdown, which you know we'll talk about, but I think there's a sort of pushback against the mentality of using as many AI tokens and as expensive of AI tokens as you can get your hands on all the time for every task. I think that is a headwind. And then I think another one is about 10 as pays you go. You know, 80% of revenue is consumption. And
Also, 80% of revenue comes from models they released in the last six months. And so this is like a constant treadmill of needing to put out the best model every time, or things will slow down. And then you know, smarter people than me can can speak about the implications on their ability to you know deal with these sort of compute obligations that they have to to Broadcom and others. but yeah, it's not good. They have to sort of keep being
keeping the best in in in some ways.
Slava Rubin (12:45)
Yeah, so for me, I would say the tactical concern is the competitive nature of the market and the fact that I think we're gonna keep on seeing more competition on price and on stratifying uh the margin opportunities. So if you want an easier model, it's gonna have to be cheaper, which then don't let you make as much money on it. So that would be number four. My bigger concern.
realistically is I think AI is here to stay and I don't think it's really gonna go away. Um, my bigger concern is actually how big of an issue AI becomes for the US, the world, and whether there is any Washington or legal intrusion as to how uh Anthropic is able to act. So I really see that as number five in regards to what is um
What is Washington gonna say about this company? And I don't really know. Uh, I think there's some implication with, you know, the uh the blue wave that just happened, but blue or not, I think that people are feeling very uncertain as to what is AI doing to us or as how AI is impacting our lives. And I do think there's a more than zero chance that Washington steps in one way or another. I'm not a big fan of that.
generically. Um, you know, I I like kind of freer markets, but I do think that could happen.
All right. Let's start diving in into some of the pros and cons. So first let's talk about some of the pros. So let's talk about revenue growth. Take it away.
Jan-Erik Asplund (14:28)
Yeah, I mean, this is one of the most, astonishing kind of revenue charts to look at from any company. and as you'll see, obviously 2026 is where things really took off. the big change, as we talked about in the previous slide, is 80% of revenue is from consumption. And the amount of consumption of Anthropic you know, uh of their API took off in 2026, largely on the back of the development of agents.
And we'll talk about Claude code as sort of a primary driver, but in general, agents are, you know, instead of you're having a chat conversation with an AI bot, you're having an agent go off and and take care of tasks autonomously and spin up its own subagents that also do tasks. And you're not talking to it anymore, you know, it's performing actions on your behalf, and all those actions incur increased cost because you're paying for the consumption. And so
Basically, for most people, the amount of tokens that they consume, whether they're working in Claude code, they're a developer, or the tools that they're using are calling Claude agents. you know, for the most part, people have started consuming a lot more tokens. That's what led to the you know gigantic explosion of revenue in the first half of 2026. And that's also what helped drive the slowdown in in Q3, which is the flip side to that consumption case.
Slava Rubin (15:49)
So you just mentioned Claude code. Let's go to Claude code. What is Claude code and how is it, impacting this massive revenue growth?
Jan-Erik Asplund (15:56)
Yeah, Claude Code was basically the the Anthropics version of Chat GPT in the sense that it was their killer product, killer app, not that it was for consumers. So it's a, you know, for for software developers, for engineers, it's a tool that basically will uh autonomously change their code just through a natural language conversation uh in their terminal. So it was very early, probably the first.
big public show of this agenc trend where you know this is AI agent will just do tasks on your behalf and then it plugs into your your Anthropic subscription. you know subscriptions are 20% of revenue, but a lot of people use it through the API so it creates consumption revenue or, you know, they use a w when they go over their subscription, they end up paying there. So it was yeah you know at the time in May 2026
It was around eight billion. you know, I think at that point doing more than Cursor, which was the original uh huge AI coding product, which Anthropic was mainly the model that was being used inside Cursor. So they kind of took down Cursor at the time. And yeah, I I can I can imagine that by now in October, this number is significantly higher. Well, into double digits, you know, probably you know, over fifteen, over twenty billion.
and run rate because this is this is really their core kind of product, the way that Chat GPT is core to Open AI.
Slava Rubin (17:28)
So I just want to make sure people understand this. So this graph on the left is almost unbelievable. So um Claude Code is the software that Anthropic is using to help augment how engineers code to the point where, you know, you don't want to say it's being done without the humans, but it's almost being done without the humans, almost.
And this product is so strong and growing so rapidly that it didn't exist practically a couple of years ago, and now it's doing tens of billions of dollars of run rate. Is that right?
Jan-Erik Asplund (18:06)
Yeah.
Slava Rubin (18:06)
Yeah, it's it's kinda wild. Let's talk about customers. So um if we can flip to, you know, how is this all layering into customer growth?
Jan-Erik Asplund (18:17)
Yeah. So, you pointed this out on the original upside slide. And their, you know, customer adoption has been has been I mean, again, you know, huge relative to even other very successful software IPOs. And in a way, this is downplaying it. You know, we're we're we're pointing out just the amount of companies that are using them and driving huge revenue or or driving, you know, our huge accounts, half of the Fortune five hundred.
Is spending at least a hundred thousand. and so that's huge. 900% net revenue retention, which is insane. you know, it comes from when people adopt Claude, they maybe are first using it for one small use case and then it spreads across to, you know, the the the engineering team is usually where it starts, but then it spreads from there to marketing to sales product. It's a sort of completely horizontal product, so it can go everywhere, and then you get that consumption revenue and that.
Agent uh juicing of the of the consumption spend, you end up with that kind of retention where the the people who keep using it become you know utter power users. again, I would say the one thing that maybe is slightly misleading almost is that you know, this represents very little spend for a Fortune 500 company. So it less than a tenth of a percent of their actual total expenses are being spent here.
Slava Rubin (19:39)
A question about Claude code. How is it that the revenue is being allocated? Is it token usage or is it subscription? How is Claude code being recognized?
Jan-Erik Asplund (19:49)
Yeah, it's not it's not completely clear when they have disclosed revenue. we've worked on this quite a bit, but assuming it's it's kind of like a a mix of the API uh token revenue with maybe some portion of subscription that's been kind of you know, based on usage, like how much of the subscription has been used in Claude.Ai versus Claude code. that's been our that's been our operating assumption is that they're they're sort of doing a blend.
Slava Rubin (20:17)
Let's talk about the profitability.
Jan-Erik Asplund (20:20)
Perfect. so yeah, they they you know for years obviously people talked about AI. you know, this would never be profitable. this is a business model built entirely on subsidization or whatever. you know, I think Anthropic showed that the core business of you know taking in user queries, running them on on GPUs and serving the results back can be can be profitable. So, you know, like
That solves that answers a big question. that said, you know, we're we're again we're we're adjusting here, stock-based compensation, all that kind of stuff. But it does suggest that there's a business here that can be sustainable. and you know, this is factoring in, for example, expensive model training, which has to happen you know over time as they're training the next generation of models. So
you can look at this as an indication that, you know, there is a profitable profitable business here. You can also see potentially vulnerability with the with the Q3 result, you know, a slightly thinner profit margin that any drop off in growth, you know, puts them back, you know, thins the margin again. So if you don't think that they're gonna get more profitable from here, then that would be a concern.
Slava Rubin (21:39)
One of the things that I read was that they're getting more revenue per every dollar of compute that they're spending on. So is this showing up here in the margin?
Jan-Erik Asplund (21:48)
Exactly. If if you if you look at that graph, which it looks exactly like this graph in a way, in terms of the improvement. and so yeah, that a lot of that is internal improvements. no, actually I didn't include it because it actually is redundant in a way with this graph. Really what it's just showing is before they were making less than a dollar for every dollar of compute. Now, you know, as of Q two, they were making two dollars for every dollar of compute. So they were improving a lot.
Um a lot of that is based on when you have a company like Anthropic and you you have to build out all the data center capacity to serve customers in the future, you end up with these very negative operating margins in the moment. but they had to invest ahead of time now that demand has has appeared in 2026 and they've been able to show they've been able to be profitable because they had the sort of you know the the the forecasting down.
for what they would be dealing with in in twenty twenty six. Plus there's all kinds of internal improvements, all that stuff that Deep Seek and these Chinese AI companies were doing to be super efficient and cheap. Some of that stuff has made its way into, Anthropic and Open AI. So so we're getting we're getting smarter about how to serve these models more cheaply as well.
Slava Rubin (23:04)
So we keep on saying the uh low amount of Fortune five hundred spend. do you have a sense of what amount that is in absolute terms as some of these organizations? And how much upside do you think it realistically has to go up?
Jan-Erik Asplund (23:19)
the upside is huge, I think, for for for everyone. I think it's the question is maybe where we'll see it first. you know, I think law is one where there's been a ton of investment um from Anthropic and Open AI in trying to create kind of tailored products and services for them. So so law might be one where we will see it go up significantly from there. But I mean I would would you know the I think the pie in the sky future vision.
i hit w is one where you see it, go from tenth of a percent to to one to ten percent, you know, or more in the in the very long term. could be more than that, but in absolute terms, yeah.
Slava Rubin (23:58)
Let's use like a tactical example and you can say I don't know and that's a perfectly fair answer. But like Salesforce as a company, you know, how much do you think they're you spending in absolute terms on Anthropic and how much do you think they could spend for Anthropic?
Jan-Erik Asplund (24:11)
Yeah. yeah, pr I think probably in the in the hundreds of millions, is what I would guess. They're probably one of the biggest in the world. Yeah, yeah, yeah. Probably in the hundreds of millions. And I think probably they could probably ten X from their you know, in the U.
Slava Rubin (24:29)
Think that they could be spending billions of dollars on Anthropic.
Jan-Erik Asplund (24:32)
In the in the maybe ten year bulk case. I
Slava Rubin (24:36)
Yeah.
Jan-Erik Asplund (24:37)
think that's feasible, yeah.
Slava Rubin (24:38)
No, that's super helpful. Let's go back to the bullish points from the S-1. So we just hit a bunch of the positives into greater detail with slides. Is there anything, Jan-Erik, that you want to touch on that we didn't double click on that you think is a really important point to cover here?
Jan-Erik Asplund (24:57)
Yeah, we I I think we hit everything here. I I feel good. No, I
Slava Rubin (25:01)
So
good. Okay, perfect. Let's talk about some of the concerns. Let's go to the growth scare.
Jan-Erik Asplund (25:05)
Yeah. So forty-seven percent year quarter over quarter growth is pretty good, but when you just grew 132 and 144% respectively in in Q one, Q two, it looks bad. it's very scary. everyone thought Anthropic will accelerate forever. Um the reality is that you had customers sort of going back on this, you know, vision of you know, token maxing where you had
Meta had a leaderboard for who could use the most AI tokens. and and the person on top was the winner. Uh or you know, that's the way it was framed. because everyone wanted to use more and more AI, really get, you know, it into their workflow. There was a big sort of pullback from that in you know, end of Q2, in Q3, and a lot of companies started started
reducing their usage of Anthropic, a lot of companies started going to other models that were cheaper and because of just how much the expense was going up with that agentic move that we talked about before. And so I those were sort of the two big you know, secular trends that definitely slowed down growth. And then there wasn't a huge model release in Q3 until actually the very end of Q3 is where we got the latest Anthropic you know, sort of
top of the line model, which will be uh something I wanted to mention too on the plus side is Q four, you will we will see the effect of Opus five point five. So so really their their latest and greatest model came out just a few days before Q three ended. So we didn't see any of that in Q three.
Slava Rubin (26:45)
Why is this concept of the latest model so important? Because it seems to be very relevant on the expense side because of how much money has to go in to get to that model and let's call it the RD, which is very painful on the red side. But then it also seems to be a major catalyst to then drive top line and revenue. Why is this concept of the latest model so important?
Jan-Erik Asplund (27:09)
Yeah, so there are a few reasons. one is anthropic's entire uh position in the market is based on being the best and and on having premium models. So, from the middle of 2025 onwards, you know, they were the premier coding model, and that's what drove Claude code and and drove revenue in general. And a Open AI was not far behind and and and today, as of this stage, you know, they're very roughly very close. Um
But if they stop, if if Anthropic were to stop training new models, get really profitable suddenly, they would be, you know, pretty dead in the water in six months to a year because other companies would keep developing better models. And if Anthropic doesn't have sort of a premium model, it's better than everything else, I'm not sure what their sort of value prop is. because there are you know Chinese models that you can use through Open router for one thirtieth of the price that
are are also improving at the same time. So they really can't
Slava Rubin (28:13)
why aren't we all sorry to interrupt? Why aren't we just using the Chinese Open router version models that are one thirtieth of the price? Let's ignore, you know, which is not so easy to ignore, but let's ignore the fact that it's Chinese. And um let's say it's just an Open source model. why aren't we just using those models?
Jan-Erik Asplund (28:31)
Yeah, it's a great question. I feel like there's a bunch of reasons. One is like it it's sticky. You know, you build everything around your Claude models because they were first, and then it's hard to actually just switch over. There's a fair amount of testing and all that iteration involved. They're still the best as well. So a lot of use cases, maybe the majority of use cases, you're still edging up against the edges of, you know, what an LLM can do, and you need to squeeze every uh
Or you want to squeeze every last bit of intelligence out. So you do want to use for a lot of use cases, you still want to use the best possible model. and for those models, people are just going to use whatever's best, like Anthropic. but again, with the stickiness point, like yeah, it's only been a few years. So a lot of companies are not really ready to um, you know, fully migrate over, but they are beginning to for some use cases to these Open source models.
Slava Rubin (29:22)
Challenge that one more time just to show a risk for Anthropic. So Formula One cars are totally awesome. Okay. And buying a Ferrari is totally cool. But most of us are driving around in like Hondas and Kias, etc. Like, why is it that when it comes to these AI models, we need the, as you're referring to, the Ferraris or even the F1 formula car. And now and why is the Hondas?
And the Kia is not enough.
Jan-Erik Asplund (29:53)
Yeah, I guess I I think of it as like you know, caliber of employee, you know, if want to hire someone and you have, you know, a person here who's a hundred IQ and a person here who's ninety IQ, you know, you can you know these are very crass and and blunt terms, but you know, you could do a lot more with the with the person who is who is brighter and and and you know, more eager than can with a person who's, you know, maybe a little bit less uh bright and eager. Um
And you wanna pay, you pay a premium if it means you have if you can move faster, grow faster as a business, you know, do more. and so yeah, I I'm not sure how that aligns exactly with Formula One, but I think fair enough. Yeah.
Slava Rubin (30:37)
Fair enough. I think people understand the point. So in regards to this is a big one that people like to talk about, especially in the media, the compute bet. There's this talk about anthropics making all this revenue, but actually their commitments of how much they're gonna spend far outweighs their revenue. So this business model is just upside down. So can you speak to that?
Jan-Erik Asplund (30:58)
Yeah. this is, you know, things get very these are very, very big numbers. But if you go back a few years, you know, you could probably have the same conversation if you had access to their the financials around what they had already created in terms of obligations for themselves for twenty twenty six. And you would have said, you know, all this compute, do you think you're gonna have demand in twenty twenty-six to meet this? you know, and do you have any kind of financial modeling that shows that this is feasible?
that I don't know how you would have a reasonable conversation that would lead to, saying $80 billion annualized is a is a is a reasonable expectation of revenue uh as of this month. But that's what happened, right? And and so when you look at $895 billion of committed spend over the coming years, that is a huge number. Um but this is you know this the idea here is that they need this to sort of serve future demand. And if they get it, then they can serve that future demand.
profitably. Um if the future demand doesn't come through, obviously then we are looking at a situation where Enthropic has lost, you know, significant ground to Open AI and other companies. So if you think about, you know, this number standalone in terms of spend, yes, it's huge, but also the revenue ramp that they've had has been uh you know obviously insane. And this is a you know sort of necessary element of of building something that can serve
exponentially growing demand um in the future. Obviously they need to make that happen. The demand needs to come in, but that's the that's the bet, so to speak.
Slava Rubin (32:34)
What happens if the demand doesn't cause what you said is we could have looked at this in twenty five and we could have like, my gosh, there's no way the twenty six commitments are gonna get fulfilled, but then the twenty six commitments got fulfilled 'cause the revenue is so big. What if it comes up short in twenty eight? What happens?
Jan-Erik Asplund (32:50)
yeah, so there's, like I I think the SpaceX contract is is sort of loose, but I'm not sure about the Amazon, Google and Broadcom. I I you know, assuming that these are not, you know, cancelable infrastructure commitments, then yeah. I mean then there you have a huge, you know, problem with gross margin and with profitability where you just go, you know, very negative very quickly because i if you have Anthropic having to pay for these idle servers
regardless of demand not catching up because everyone's going to Chinese models or whatever, the then I'm not sure how, you know, I'm not sure how they surv how that I'm not sure how they survive a situation like that where you know all the cash is going out the window to pay for this and and all that. I will say, however, you know, Amazon, SpaceX, Google, these are all investors in Anthropic, these are all partners. So I'm not sure if there's any kind of, you know, how those contracts are built, but assuming they are sort of uncancelable contracts, that's what would happen.
Slava Rubin (33:50)
Let's go back to the S-1 reds. so we covered a lot of the downside issues on I I'm gonna segue the if they all go to the Chinese models and the Open source weighted models. You know, that's one of my top risks for sure for the business. This kind of speaks back to the price wars, right? This is about trying to find, let's call it, lower IQ, easier, more efficient car, whatever it is, getting to do the work enough, good enough work at a more efficient price.
We covered quite a bit of the risks. Is there anything that we didn't double click on that you just want to make sure we cover here?
Jan-Erik Asplund (34:27)
well we'll talk about Open AI. so I'll leave that for when we talk about competition because I think that that's important.
Slava Rubin (34:33)
Perfect
segue. Let's do that. Let's go to competition.
Jan-Erik Asplund (34:37)
So obviously OpenAI was the big name in 2025 still, and was the leader in this space. And what they didn't foresee is just how important coding was gonna be as a sort of as really the first great, you know, sort of product version of AI. And so they missed it. And then, you know, they didn't release their version of Claude Code until months later and
They still, they haven't caught up to Claude Code, but they have had a dramatic resurgence around coding. you know, they've trained their models for that use case now. They have shown a lot of eagerness and willingness to discount and offer better pricing than you can get on Anthropic. So, you know, for example, if you're me and you're not really an engineer, you know, you're just using OpenAI's codex because.
is so much cheaper. You get so much more access than you get with a Claude code, right? And OpenAI also has you know 900 million or sorry over a billion users of of Chat GPT which is their main you know linchpin this consumer um this consumer platform that they have so you know I you can't count Open ai uh out at any point obviously this whole model release thing is often a game of leapfrog where the last model is always the best one
And I I think Open AI is is they're they're incredibly close to Anthropic and you know so it it is a two horse race again, even if maybe for a a brief period there, people sort of forgot Open AI existed in the first half of this year.
Slava Rubin (36:21)
Yeah, I mean, I think it's more than a two-horse race, personally. I think Google with their distribution is massive. I would not, you know, discount Google being a massive player in this a couple few years from now. I think they're already doing well, but you know, I think they're just gonna more lean into being able to have the AI product. I think it is impacting their search products, so they're gonna have to figure out that solution in regards to how to monetize their AI product.
But I believe in them figuring it out just like they figured out YouTube, which is now crushing. And I do think the Open weight models and the Chinese labs, that's a legit situation. I don't think that's going away at all. so I think it's more than a two horse race. So I think competition is very real. Um, which actually, as it relates to SpaceX, you know, SpaceX almost feels like a monopoly. It just didn't feel as competitive here, as much as Anthropic.
feels amazing and is so bullish, it does feel like a more competitive market. Let's dive into that competition more head to head with Open AI. I think we have some great information in comparison here.
Jan-Erik Asplund (37:26)
Yeah, exactly. So you can sort of see here, the very first row reflects twenty twenty five revenue, where Anthropic was a third of the size, but then from there on out, Anthropic sort of, you know, took over, growing absolutely tremendous amount in in in in the first quarters of twenty twenty six. you know, but in in uh there's also sort of things here where you could see that they're not that
Different. For example, gross margin is relatively similar because these two businesses are pretty are pretty close. They both have tons of compute commitments. OpenAI was super sort of ahead uh on compute for quite a while until 2026. And now Anthropic has sort of caught up. As you can see, the numbers there are not too far off, 900 billion to 1.4 trillion. and then valuation wise, you know, we're right sort of around the same numbers with Open AI.
in talks or around last I checked around this 1.25 trillion mark Anthropic likely that's the where around where their range will be 1.25, 1.3. so yeah, very similar businesses in some ways. I think Anthropic has a significantly larger usage based or consumption percentage of revenue, whereas OpenAI is a little more half and half between consumer and enterprise, although that might have shifted too in the last few months.
with the growth of codecs. so yeah, it yeah, really interesting sort of two businesses to look at, especially once OpenAI gives us an S-1. I think that'll be great.
Slava Rubin (38:56)
Yeah, so I know we're just picking these specific metrics. and it seems like there's more anthropics here on the end for edge. but you know, now that the valuation seemingly is coming down for Anthropic around a similar valuation to Open AI, or who knows, maybe the Open AI valuation is coming down. Who knows? Maybe it's a public versus private situation. is there one that you feel is getting the edge overall at the moment?
Jan-Erik Asplund (39:22)
I used to kinda talk a lot about Anthropic on on these webinars. I kind of at the at the moment I feel a little more inclined to to look at Open AI just because I feel like they're they're sort of uh their surge in twenty twenty six has been has been impressive and I think maybe underappreciated, maybe because Anthropic has sucked a lot of the conversation up around itself. But I do think they have a very competitive um
software engineering product. and obviously the the biggest, you know, sort of consumer distribution in the space besides Google. and so yeah.
Slava Rubin (40:00)
Nice. So let's talk about valuation. Uh going back. So now we're starting to talk about I, the listener, you know, should I be getting into this IPO or not? So let's start to frame that. So what's the valuation? How do we get here?
Jan-Erik Asplund (40:14)
yeah, so similar to the revenue chart, 2026 was sort of when things uh really took off. And we saw them hit you know, touch almost a trillion in terms of valuation back in the summer when this sort of growth curve was was really taking off. And now, you know, from what I understand, we're looking at 1.2, 1.3 as the most likely uh trillion in terms of the price.
which would be yeah for this IPO later in November. So um yeah that remains to be seen remains to be seen exactly where we'll end up. But it it's been absolutely, you know, an insane takeoff over the last year, obviously, which has been great for some of these investors that we pointed out on the right. And maybe not great if you're looking to buy a house in San Francisco anytime soon and you're looking at the high end of the market.
Slava Rubin (41:09)
It's supposed to like mint something like 10,000 10 millionaires. Meaning 10 million or more, is that right? It's something like
Jan-Erik Asplund (41:17)
Yeah, that sounds right.
Slava Rubin (41:19)
ridiculous ridiculous number, and they're all gonna be looking for a house. Exactly right. So let's try to put this in perspective. So you see these massive jumps on valuation, especially from February twenty-sixth to May 26. It actually makes the IPO range seem a bit muted.
comparatively. There was talk of two trillion and now we're bringing that down in terms of what the market's talking about one point two, one point three, one point four. Let's just put that in perspective in terms of like a multiple on what kind of price is that in comparison to the market.
Jan-Erik Asplund (41:47)
I think the multiple is ultimately not it's nothing crazy. you know, we're talking like yeah, at a sort of one point two uh we're looking at like a like a basically an eighteen, nineteen X multiple on their if you take their like quarterly revenue and annualize it. so
for all things considered, um, it's middle of the range uh as far as like comps go, um, for a multiple.
You know, more expensive, of course, than your huge tech companies, but roughly comparable to Open AI. And then like we talked about SpaceX AI is the full name of the company. Now, you know, that is it is up in the thirties. you have one of their Chinese Open model competitors listed in China, obviously a very different valuation, but at a higher multiple. The point is overall, you know, this the
it the multiple itself is not not the priciest.
Slava Rubin (42:49)
So yeah, I mean it's actually interesting because people talk about Anthropic like it's so expensive. But realistically from a fundamentalist perspective, as as it relates so far, it doesn't look like it's actually that pricey, especially if the number comes in to one point two, one point three, one point four.
Jan-Erik Asplund (43:05)
Yeah, exactly. And it it is the it's the denominator, right, is so big. Um the amount of revenue is so big and and you're paying for um that. But I I think, you know, the the flip side is that people are seeing um people are seeing the lack of you know, it's not a recurring software subscription that's gonna keep on coming in reliably every year. It is a astronomical, you know, amount of growth fueled by, you know.
this massive compute obligation and and having the best AI model at all times. So bit of a question around around how maybe durable or lasting the revenue is that might account for this lower multiple.
Slava Rubin (43:50)
What do you think about the the Jensen, NVIDIA, Spygate, all that stuff right now? What is that narrative have having impact on the Anthropic IPO? Anything? Very much zero?
Jan-Erik Asplund (44:04)
In the short term, I think it might have a little uh a little bit of an impact. I mean, it had an impact on Nvidia, and and that's a very important company in the space, given that, you know, virtually most of what Anthropic does runs on their chips. I don't know that I feel any any way about it, sort of beyond the very short term impact because I'm not it doesn't seem like that
real of a story. I don't know what you think, but I don't think she's a spy. That's my that's my personal my personal belief. And I think uh it's probably not gonna be a long term story.
Slava Rubin (44:41)
There will be no confirmation or rejection that she is a spy ever to be had. So I don't think that will ever be known. Let's go back to the revenue growth in the beginning. because we have it through twenty six. What do you think the revenue is gonna be in twenty seven, twenty eight, and twenty nine? Give me some predictions.
Jan-Erik Asplund (45:01)
Great question. you're testing my um my dad brain right now, which I noticed
Slava Rubin (45:10)
Jan-Erik just had a first child, two months old. so Jan-Erik doesn't sleep as much as he normally does, even though he is the most brilliant man on the internet.
Jan-Erik Asplund (45:19)
Yeah, I noticed my my hold of my my grasp on math it was the first thing to go, but you know I think I think we're gonna end up at something like fifty f between forty and fifty billion for the year. Um we're at thirty three point two for this year, sorry, starting with twenty twenty six. I think we'll be in the in the just around the fifty, slightly over fifty range, let's say.
And then maybe maybe we'll be in the hundred twenty range for next year. Um so not maybe a little over doubling or doubling so slowing down. Um and then maybe a doubling to to say two hundred um doubling and then
Slava Rubin (46:07)
Exit, what's the exit ARR in in in December twenty seven?
If the annual is much more
Jan-Erik Asplund (46:14)
Yeah, let's say we're at eighty. Um we're at eighty this year. Yeah, probably end up at like a hundred. So maybe yeah, maybe uh two hundred, two hundred exit ARR two two twenty, two thirty.
Slava Rubin (46:31)
just to put that in perspective, right? Let's just say it's a hundred and fifty forward twelve months, right, on the ARR. I mean, it's not even a ten X, right, on price.
I I mean you have to believe this slows down. That's why you think the price goes down, right?
Jan-Erik Asplund (46:48)
Yeah. Yeah. I I would think so just 'cause of the sort of the limits on how quickly this stuff can be adopted throughout the economy.
Slava Rubin (46:56)
What are the chances? This is a little bit more of a bullish take. What are the chances that people just can't process how big of a company Anthropic and Open AI and all this AI stuff is? And they even though it's obvious that it's growing at this rate, they just can't comprehend it. Let's call it NVIDIA a few years ago. 'cause people just didn't appreciate that either. And it like skyrocketed.
Jan-Erik Asplund (47:18)
Yeah, I think that's the bull case and that's why you invest ultimately is that you don't think this is a normal company. You think that, we're talking about building the next best model for, you know, software engineers or whatever, expanding coding, but you know, the company itself has the vision, you know, and and Dario Amodei the CEO, has been on the record recently saying that, you know, every major disease will be solved in the next 10 years. you know.
earnestly saying this. So, and he's not Elon Musk, so I think that there's a vision that, is that AI is going to fundamentally transform, every sector of the economy and life in general. And if it does that, then you know, if Anthropic or Open AI gets there and begins to you know reshape the world uh in their own image, you know, in the image of AI, then then we're, you know, it doesn't really matter what is going on with
cheap Open source models, whether in China or Europe or anywhere else, or compute costs, right, if they are sort of yeah, I mean, yeah, advancing beyond our sort of recognition or or ability to tie back to a comp, you know. so I think that's very possible. But you know not very
Slava Rubin (48:32)
Possible possible.
What's the bull case five, ten years from now? What's the bull case? What's the enterprise value? Let's say the IPO, let's use a round number, one point five. Five years from now, what's the enterprise value on the bull case?
Jan-Erik Asplund (48:45)
Yeah, I I'll try to rein it into reality a little bit and say, you know, maybe let's say like five X where Nvidia is now. So like
Slava Rubin (49:00)
Sorry, five X of where NVIDIA is now, which is NVIDIA is between four and five billion, depending on who Jensen dates. So um that's like twenty to twenty five billion. Sorry, trillion sorry, trillion.
Jan-Erik Asplund (49:12)
Maybe
yeah, yeah, yeah. Maybe let's distinguish you know, you got me started and then and then I lost track of if we were talking about the bull case or the hyper bull case, because I think the the hyperbole case for AI and and part of the reason for the bet um is that vision. But I think
Slava Rubin (49:29)
Well give me give me the bull case
Jan-Erik Asplund (49:31)
Yeah, I think the bull case probably is that it ends up looking more like in five years.
probably more in the realm of,
Five, six trillion. That's what I would say, roughly. So a little more than NVIDIA's worth today. I think that is more of a that's not a world where we have, super intelligence.
Slava Rubin (49:51)
So that's like a four or five X from here, uh depending on let's call the IPO price, the and where we're at. And that obviously does not coincide with the growth rates we're talking about in regards to revenue growth. It just obviously slows somewhat and it becomes a five. But you're still painting a, you know, let's call a four or five bagger. So let's talk about how you could get in. Let's go to the IPO, your options to get in, which is
you could buy it via proxies, which people have been doing. The easy proxies are actually just pick up some Amazon or Google, because they are gonna crush, crush, crush. Amazon's gonna make so much money um on their Anthropic stake. Obviously, there's the IPO. retail is getting 20%, which is not quite as much as SpaceX, but it's a nice amount. As a matter of fact, my Robin Hood account just told me on Monday.
that you could get your you could get queued up for Anthropic, which by the way, I queued up, so I'm competing with all of you guys. you could try to flip it on day one, uh, if you're allowed to. There's the buy and hold and then there's the ripple effects, which is try to stay in other players and try to make the money that way. Is there do you have predictions as to how this is gonna go and how would you play the IPO?
Jan-Erik Asplund (51:09)
yeah, I well I I think there's a potential, you know phenomenon like when what we saw with SpaceX, which is the opening day pop, and then maybe some kind of you know recession, like giving back those gains over those first months. I you know, I I we saw that happen in a big way with SpaceX, and then you know, potentially we get back to that air IPO price after the first six months. you know, people potentially getting
a little carried away. I like the wait and see approach, you know, when it comes to to to public markets. But yeah, I'm I'm not sure. I don't think my predictions were super strong with SpaceX. So I'll try to stay away from those.
Slava Rubin (51:51)
I like SpaceX as a long term safer investment better. I think that Anthropic, while I'm quite bullish and very pro, I think it has much more risk and volatility. I would not bet my entire children's future savings on Anthropic.
That does not mean I don't think there's a lot of money to be made on Anthropic. I just think you have to be ready for significant volatility. My personal opinion is if you are able to flip, meaning to get into the IPO and flip, I do think there's going to be FOMO and I think it's going to be the story of, you know, all business for a few days. And I think there'll be money to be made there. I would be shocked personally if it IPOs and just goes straight down. I think that would be quite negative.
not only for the company, but probably for our entire IPO trade, i.e. the stock market and the hyperscalers. So my opinion is it will price cheap enough that it will definitely pop. Whether or not you'll be able to trade it is really up to you or your brokerage house, whether or not they you know put you in the penalty box, etc. Now, I personally would hold, because I do think there's the potential for a two, three X, five X, 10X from here, I just think you need to size your position accordingly.
be willing to take the the downside. And I don't know you need to come in all in on the first day or the first month. You should DCA in just like I suggested with SpaceX, because just like with SpaceX, there are times when the market or idiosyncratic issues with the company just has the stock plummet for no good reason. And I'm still very bullish on SpaceX in the long run. And I would DCA into Anthropic. just keep in mind that Anthropic is a more competitive market. There are more risks. So just
Take that into account. Is there anything we didn't cover Jan-Erik you want to make sure we touch on?
Jan-Erik Asplund (53:47)
yeah, hopefully, you know, Dario doesn't meet with any espionage related Chinese women, you know, and and we can we can just enjoy the uh price action, you know, in its pure form.
Slava Rubin (53:59)
Dario has a whole entire future ahead of him to make all kinds of positive and negative stories. So I'm sure there'll be plenty of media to be had. when you run one of the most incredible, ridiculous, world changing companies, I'm sure there'll be stories to be had. Well, I'm actually quite impressed. We covered so much information. thank you everybody for all the questions. We had so many. Um and I guess that's a wrap.
So feel free to uh stay on top of it. we are actually archiving this and it's still a few days before the IPO. So I'm sure there's gonna be more information to come. Have a good one all.
