NCI: The Anthropic S-1: Cambrian Explosion or Extinction Event?

Speaker1: [00:00:00] Welcome to Non-Consensus Investing. I'm Ram Ahluwalia, founder of Lumida Wealth. On this show, we identify emerging trends, look at the world from a different perspective, and explore what it means for your portfolio. Through my own analysis and conversations with guests, we ask what is the consensus missing, and where's the opportunity?

Let's get to work.

Speaker2: Claude Code and Cowork from Anthropic became the number one software product on the history of the planet, growing 10x, 10x. We've never seen anything in the software industry in 70 years growing at this level . 

Speaker1: If only, if only there could be a $2 trillion blockbuster IPO. 

Speaker2: Every week we need one

Their revenue per employee is breathtaking, 6 million plus. That's comparable to Nvidia. 

Speaker1: Think of a number in your mind's eye. [00:01:00] How many customers pay Anthropic at least 100,000 or more annually? Answer is around 12,000. 

Speaker2: That's, it's nothing. 

Speaker1: That's nothing. 

Speaker2: We are not even at the beginning of the beginning again.

Today, the bosses are saying, "Use the AI." And most companies are like, "What AI? What, I-" 

Speaker1: Three months ago, my view was this is 1997. Now I believe we're 1995. You know, there's this comment that Ken Griffin made from Citadel about these LLMs. He compared them to airline companies. Is Ken Griffin right or wrong?

Speaker2: He, he's asking the right question for five years from now. We have not even invented Pan Am yet. Well, that's RAMageddon. Yes, the RAM has gone up 500, 600%. If you need a smartphone or a computer, go buy it today. It will not be cheaper for the next three, four years. Nvidia is the most valuable company on the planet.

A semiconductor company has margins better than Microsoft had at its [00:02:00] peak of Windows, at 85%. The world has gone upside down. Applications, a Cambrian explosion, so there's all kinds of life. In three to five years, 80 to 90% of the new life will die. So as these things go public, expect volatility. This is not all prosperity in a straight line.

Speaker1: Cambrian explosion, there you have it. Welcome to the next episode of Lumida Non-Consensus Investing. I am thrilled to have very special guest, Michael Parekh. I don't think Michael needs an introduction. He's been a familiar guest on the pod. He started his career at Goldman Sachs, uh, as an analyst and also advising these companies, and fast-forward, he's analyzing AI and also advising these companies.

We're gonna focus on the Anthropic S1. It's a big drop with a lot of insights to process. It's the main event driving markets. So thank you for joining. If you're listening, you can follow on YouTube or Spotify as well. [00:03:00] Welcome back, Michael. 

Speaker2: Thank you. Good to be back. Good to be 

Speaker1: back with the team 

Speaker2: here. 

Speaker1: So how's the gang?

Good, good to see you too. 

Speaker2: Nothing's going on. 

Speaker1: Um... Nothing's going on and nothing ever changes. 

Speaker2: It's so slow. Everything is watching paint dry. 

Speaker1: If only, if only there could be a $2 trillion blockbuster IPO. Every 

Speaker2: week we need 

Speaker1: one. Now, what's interesting in the S1 for Anthropic is the TAM. And introducing the story is, I'm sure what the bankers and the CFO spent a lot of time on.

Speaker2: Mm-hmm. 

Speaker1: So I'm sharing on the screen here the opening of the S, um, S1. You've got Goldman Sachs lead left. Goldman Sachs on the left. Morgan Stanley on the right. Classic tech IPO. 

Speaker2: My 

Speaker1: [00:04:00] alma mater. I got a- And then you've got... That's your alma mater. You're gonna get a lot of investment banking fees on that one.

Uh- Couple of dollars ... and then you've got biology. You've got the stars that are featured here. Uh, and then the opening line, "We believe that A- AI..." I'll say it again. "We believe that AI is one of the most consequential technologies in human history. If managed well, it could-" You know, this 

Speaker2: TAM, um- 

Speaker1: Go ahead.

Speaker2: No, you know, this TAM is a subset of what Elon claimed on his TAM in his S1 in June. That was over $30 trillion. But this one- 

Speaker1: So Elon's at 30. And what's Anthropic at? 

Speaker2: I think, I don't know if they said a number. They just used a lot of qualitative words and pictures, like you're showing. I didn't mean to interrupt.

But, uh- Yeah ... they're directionally both the same, but Elon basically made [00:05:00] sure he wasn't leaving any terrestrial, uh, areas unclaimed. 

Speaker1: And the $30 trillion is about the size of US GDP. 

Speaker2: Correct. 

Speaker1: So you got multiple firms going after this massive GDP market, and there's that comment from Gavin Baker that seven of these Mag 7 companies, well, five out of seven will survive, and two out of seven will disappear.

Uh, so what stood out for you in the S1? 

Speaker2: The efficiency at which, uh, they are pr-producing their tokens and, and serving the extraordinary demand. The, the biggest takeaway, you know, we-we've been following technology companies for, for such a long time. You know, I've been tracking these things through at least four tech waves over forty decades.

The The growth rate for Anthropic, if we look at the first dollar to [00:06:00] today, we're looking at fourteen x, twelve x, ten x through two to three periods in terms of revenue growth. Um, and especially, you know, since they caught lightning in a bottle with, uh, Claude Code and, and then Claude Cowork about a year ago, the revenue numbers, you're going from about a billion or so to now fifteen, sixteen billion in the last quarter.

There's-- there are several blanks in the prospectus because they're still-- uh, there are some, some redacted numbers, so we'll see those things filled in. But if you chart this out and I'll have... I, I've had some, um, detailed charts on all this already done on my Substack at Michael Parekh, uh, uh, Reset to Zero.

And I, um, actually got a-- I've got a detailed one coming out tomorrow with about nine or ten graphs. And if you look at all that, the-- what just takes your breath away is the ten x growth on the revenue side of the equation, which just shows that the demand picture and then when the compute is [00:07:00] available is massive.

It's tens of thousands of businesses around the world spending hundreds of thousands of dollars per period, um, quarter, month, whatever, in terms of their applications, and the demand is, is very much there. To that, Anthropic is, is-- has to make sure... Remember, there's a variable cost element to this. They have to build a compute to serve the demand.

And Anthropic, over the last couple of years, was behind OpenAI in terms of rolling out their compute capacity, renting other people's servers mostly, et cetera, and they've accelerated that. And so those commitments are in-- also breathtaking in the prospectus, now over eight hundred billion dollars in commitments.

But the efficiency at which they serve-- seem to be serving this ten x growing demand is improving, going from about, by my calculations, two, two and a half dollars, uh, per revenue, uh, dollar to now fifty cents a revenue dollar, which allows them to show these... After all the adjustments, there are some big numbers, [00:08:00] seventy-two billion dollars in, in, uh, in, in costs and charges, but those are mostly having to do with the technical charges of converting convertible to equity, which gets into a little bit of the, uh, non-financial person's detail.

But it's important because you take that out and the margins on this, on this growth are impressive at eighty-five percent, et cetera. The last thing I'll just say is there is an apples and oranges in which, uh, Anthropic computes their revenue versus, uh, when we see OpenAI's S one. They-- This stuff had been leaked over the last few weeks.

I've written a lot about that If people are interested, OpenAI has been my view of the more conservative, my, my preferred way of measuring revenue, which is when, um, when their hyperscalers sell OpenAI capacity, the dollars that come to OpenAI are counted at O- as OpenAI's revenue. You know, so there's a gross merchandise value, as it were, as I did in the eBay IPO in '98, [00:09:00] and then the dollars that come to OpenAI.

Anthropic computes it a different way. They take all the dollars that they're getting from the hyperscalers, take out their payments to the hyperscalers and their other costs and so on, and then compute a net revenue number that way. So it gets into the weeds there. So there is an apples and oranges difference.

So when you see the OpenAI numbers in the next, say, few months, OpenAI is gonna, you know, presumably go public next year. There is a, a difference there. And then, you know, people like we-- you and me will adjust all this and make it all pari passu, et cetera. But those three, uh, are the numerical takeaways.

And then on the risk factor, which is about a, um, you know, 95-plus pages, there's a l- you know, very uniquely, uh, Anthropic-style, doomer-oriented, uh, risk factors. Like, "Hey, there is a chance that this AI may kill us all, that we may not be able to contain it, control it, uh, interpret it," you know, all of those kinds of elements.

So there's a lot to go in there. And then the final thing [00:10:00] is the regulatory concerns. "Hey, please, governments of the world, regulate us, um, make sure we don't go amok," that sort of thing. So those three at a high level I would highlight, and then I'd, uh- Yeah ... open to all the things that struck you. 

‍