Anthropic's S-1

Bloomberg Audio Studios.

Podcasts.

Radio.

News.

How much is riding on this IPO, Dave?

I mean, I'm not sure I can put a number on it. It feels like everything, right? I mean, the AI business is propping up the economy.

I'm Sarah Holder, and this is The Big Take from Bloomberg News. Today on the show, Anthropic is going public. what we learn from the company's S-1, and what its IPO could mean for the rest of the economy. Bloomberg Opinion columnist Dave Lee and Mike Regan, managing editor for U.S. Equities, thanks for joining us to talk about Anthropic's S-1 filing, the paperwork the company has to submit to the SEC today.

before going for an initial public offering. Mike, I want to start with the size of the planned IPO, which could come mid-November. What kind of valuation is Anthropic seeking?

Well, so we don't know exactly right at the moment. We have reported that they plan for it to be even bigger than SpaceX, which was the largest IPO. It raised about $ 86 billion. So we're looking at possibly $ 90, maybe even $ 100 billion. raised at a valuation that is still sort of up in the air. I think a lot depends on the feedback the bankers get on the roadshow when they really take this to potential buyers and say, what are you willing to pay for this? But call it somewhere between $ 1. 5 and $ 2 trillion.

Wow. And Dave, $ 100 billion, $ 90 to $ 100 billion trying to raise. Give me a sense of what this company would need so much money for.

Yeah.

Well, this will come as little surprise. It's about compute, compute, compute, right? Getting hold of those chips or getting at least access to use those chips. We know that takes a ton of data centers. We know it needs lots of energy. And then also, you know, there's the continued sort of need to retain talent and have new talent. This is an intensely competitive area, obviously. And, you know, when you're anthropic and you're a frontier lab, You need to make sure that you're constantly at that frontier. Otherwise, the cascading effect of being slightly behind all the time is going to be a big, big problem.

Well, the S1 gave us some fresh insights into anthropics, revenue and spending, which I think a lot of people were really interested to see. It is bringing in a lot more money than last year and faster, but it's also spending a lot more money than last year. Walk us through the cash flow they reported.

Yeah, I think the operating cash flow was something like $ 3. 9 billion. I mean, obviously, the headline net loss number was like $ 80 billion, but that's due to basically an accounting maneuver to account for the value of some convertible securities they sold.

Why would they do that accounting maneuver?

Basically, it shows how much the company's valuation has expanded. These are notes that can convert to equity, so they're a liability to the company. But the only reason they ballooned out is because the company itself, its valuation has expanded so much. So it now owes these investors that much equity, but it's not paying out of pocket for any of that.

So counterintuitively, an $ 80 billion net loss actually makes the company look a little better?

It's not a huge deal. Yes, I know it is counterintuitive. But I think the main takeaway from a lot of investors' reading through the numbers on this, is how much they don't have a ton of debt. In other words, they have a lot of commitments, to Dave's point, to spend on compute. But instead of doing what, say, the Googles of the world or the Amazons of the world are doing and building out their own data centers, they're basically just planning to rent them for the most part. Now, that could change a little bit in the future, but their commitments are to these leases over the next five or 10 years. to the Amazons of the world, the Googles of the world, to the Microsofts and the SpaceX of the world.

And what is the scale of those commitments? That's hundreds of billions of dollars.

It's hundreds of billions of dollars, upwards of $ 800 billion over the next, call it, decade or so. So it is a huge commitment, but it's somewhat different than simply having that debt. They're not paying interest on it. It provides them some flexibility. Some of them, not most of them, are cancelable. It's known as being an asset-light business. which is kind of what was a lot of the attraction to the Googles and the Amazons of the world until they decided to really invest heavily in their own data centers.

Anthropic is also dependent on some of these companies in another way, right? Nearly half of its revenue is now coming from money that Amazon, Google, and Microsoft pass along when their cloud customers access Anthropic's AI models.

Why is that seen as a risk?

Well, because, you know, they're paying a commission to those hyperscalers to get that revenue. They had a great line in the S1. They described the cloud providers as, quote, investors, customers, distribution partners and competitors.

All at once.

You know, obviously, you know, Google and Microsoft and SpaceX are all competing with them in the A.I. space while at the same time, you know, providing them this computing power they need.

So partners and competitors sounds messy fast.

It's very much so. And investors as well. And in many cases, if you've seen those circular financing diagrams, you know, this is this is exactly this.

OK, speaking of risks, usually the risks in an S1 are like. we could face new competition from similar businesses, or there could be new regulatory changes that make our product harder to get. Or I was reading Jersey Mike's S-1 filing from earlier this year. It's like food safety risks. But Anthropics Risk Prospectus was 94 pages long and included some risks I have never seen in an S-1, like existential risk to humanity.

Dave, what stood out to you?

Well, I mean, the length, as you say, extraordinary, right? It's a novel's worth of risks. And the breadth of it is fascinating. This is a company that has sort of threats from all sides, right? It could be from the U.S. government, which has tried to designate them as a risk and take them out of being used for government contracts over a row with the defense secretary. Then you've got, you know, rows around intellectual property rights and whether anthropics should be allowed to scrape data from certain sources.

Then you have things about climate risk. Then you have, can we get enough energy? Then we have, can we just carry on being the best of this? But the most sort of, the one that gives you the most pause, I think, is just as a human being, is to sort of look and say, when they talk about existential risk, and so they're saying things like, our models pose catastrophic and existential risks to humanity, particularly if such systems are misused, replicated, weaponized, or allowed to operate of a high degree of autonomy.

And there's this idea that what they're creating could just go off the rails.

And Dave, you made the provocative point that including some of these caveats and disclaimers was essentially a tactic to weed out the weak investors.

Explain that.

Yeah.

So in the introduction to the S1, there's a passage in there talking about how they like to hire people. And they describe it as the anti-sell. And they bring in a candidate and they say, well, here's all the reasons why you shouldn't work here. It's going to be hard. There's risks. It might all fail. And they say... Many candidates might leave at that point, but the ones that remain, they're the ones that we want to hire, the sort of anti-sale process. They're sort of doing the same approach here with this S1. They're sort of saying, here's all the reasons almost why you shouldn't back this, because there's all these things that could go wrong, in the hope, I think, that the investors they do get will sort

of be.

Along for the ride.

Because one of the things the company's making very, very clear throughout this, the founders are going to have control over what the company does. They're going to have the majority of shares. They're going to call the shots very, very clearly. And this is, I mean, that's pretty typical of tech companies. We've seen founders, you know, keep control in the past. But then they also go on to say, you know, there'll be moments when we might have a model that's better than any model in the world, but we might be decided to hold that back because we're worried about the safety and that might harm us commercially as well. This company has built its entire ethos on being the safe AI company. Now, whether you take that

on face value or not is a different matter. But, you know, this is a company that was born out of a group of employees at OpenAI saying, we don't like how this is going, we're going to go and start our own company. So it's in the sort of DNA of the firm. And I think this is the case of seeing how that plays out once you hit the reality of quarterly earnings and the public markets.

When we come back, Dave Lee and Mike Regan on why Anthropic is going public now and how much is riding on its success. Our colleague Shireen Ghaffari made a helpful point in one of her recent articles that Anthropic is in this interesting position, right? Because on the one hand, it's pushing the industry to slow down on AI development, prioritize safety. On the other hand, it is moving full steam ahead on this IPO. A few weeks ago, New York City's comptroller sent a letter to Anthropic CEO Dario Amadei asking him to pause the IPO immediately. focus on, quote, addressing these unprecedented risks. Why isn't it doing that? Why is it pushing ahead with this IPO?

Well, one really important thing, and it's pretty simple, is that there's sort of an unwritten rule in the IPO world where if you have a big IPO, you do not want to launch it after Thanksgiving. You do not want to be competing with investors Thanksgiving, Hanukkah, Christmas, New Year, kids vacation plans. So they filed confidentially over the summer, meaning they provided the SEC with this document, but the public couldn't see it until yesterday. So once they did that, the clock was kind of ticking. I think what happened is you look at the stock market on any given day, if the chip makers are selling off and the AI stocks are selling off, it had a lot of people wondering like, okay, are they really going to do it

this year? Will they push it to next year? And I think had they done that, that would have been sort of a gut punch to the market because even if you look at the market reaction yesterday, it wasn't a huge, enormous rally in stocks, but it was a nice recovery in some of the chip makers and some of the other big tech companies that have been beaten down and the market indexes as a whole rallied simply because this company filed.

OpenAI has indicated it isn't planning on going public until next year. Why is Anthropic ready and not OpenAI?

I think with both companies, there's not necessarily a super urgency to access that equity capital. You look at Anthropic, for example, it's raised $ 141 billion. It has $ 131, almost $ 132 billion in cash and short-term investments just sitting on its balance sheet. So it has a big chunk of cash to sort of weather whatever comes in the near future. So, you know, clearly they decided, you know, let's get it done this year. I don't think we have a clear sense of why OpenAI is waiting.

Yeah, I mean, I think in any of these cases, I mean, timing is everything, isn't it, as you were saying? And in the case of Anthropic, I mean, they do warn that, you know, they've had this wonderful revenue growth. I think it was 1,300% year-on-year for the first six months of this year compared to last. But they warned that that growth might not be sustainable. You know, for all we know, they've waited another year's time that some of that tailing off might have been evident in the filing.

And then that makes going public much harder. Missing the window to do it can be really sort of fatal to doing it at all, actually, in many cases. And, you know, what would be interesting to see is how damaging it might be to OpenAI to go second. Because if this comes out of the gate, doesn't perform very, very well, that's going to make it all the harder for OpenAI.

And there clearly is a pretty fierce competition to be first to get this capital. Two hours before Anthropix S1, Meta came out and said, we're selling a bunch of stuff. And they upped it to like $ 98 billion. And you just had SpaceX raise $ 86 billion. Alphabet sold a big chunk of stock. So there is sort of a first mover advantage because At some point, I think a lot of people on Wall Street think, well, we're going to reach the limit of demand for all this equity.

Yeah. Is there investor appetite for all these equity raises?

So far, yes. But I think, you know, the general thinking is at some point we'll reach a limit and we'll start to see some pushback. And we already have seen some pushback on some of the smaller data center IPOs. There was an Australian company that actually postponed its IPO plans. And there's sort of jitters around the more minor players, the neoclouds, as they're called. So these big household name companies, I think, are one thing compared to the smaller players.

But at the end of the day, I mean, we're talking about hundreds of billions of dollars of new equity being sold.

What would it mean if the IPO didn't meet expectations for the economy, for the stock market, for the AI companies that are reliant on Anthropic and for Anthropic itself?

I think it would be seen as a big signal that the returns on this aren't going to be where we perhaps thought they might be. I think what's going to be an interesting litmus test for some of that would be when the tech companies report earnings in the next week or so. Because that's the other end of this, right?

It's all well and good sort of backing the companies that are building the technology. But when you get to the ones that are deploying it, if they're not seeing the return on that investment, well, then that upends the whole concept. I think for an anthropic, what's particularly intriguing is that this is the first sort of pure play AI company, AI frontier lab to go public. I mean, we've had SpaceX, which has XAI as part of it.

But when you're SpaceX, you can offset the cost of AI by putting rockets into space, right? And the other companies that are big in AI and on the market, such as Alphabet and Meta, they have legacy businesses that can support the massive expenditures on AI, which is just, especially in a company like Meta's case, just burning money, but they don't mind. They just want to get ahead.

Anthropic doesn't really have that option. The product is AI, and that's the whole thing. And so I think it'd be a really interesting test to see how a company like that survives, you know, because I think it's a fair assumption that after this IPO, they will go back and try and raise more money as they go because they're just burning through so much of it. And, you know, one of the questions about all the compute they've committed to.

to buying up. There's two different problems. One might be they don't have the money to do it. The other problem is they might not be able to do it because of the competition for that compute as well. So it's such a litmus test for the ability for an AI company that's just an AI company to survive.

Mike, how do you think about the stakes of this IPO's success or failure, or indeed the success or failure of Anthropic as a company when you think about the rest of the economy?

AI is, you know, the main driving force in the stock market and the economy as a whole. Now, to Dave's point about the, I think it was 94 pages of risk factors, you know, it's all laid bare exactly there. What could derail all this, right? The existential risk aside, you know, if we all do go extinct, we won't be worrying about the stock market. But, you know, the... Peace, finally.

Right.

But I think Obviously, the huge risk and the big elephant in the room is the midterms coming up. And, you know, if Democrats take control of one or both houses of Congress, how hard will they push back on this massive investment being done in data centers? How loud will they ring those bells about the environment, about the potential for job loss, about all these risks laid out in those 94 pages? Really, the whole letter was trying to acknowledge that backlash and claim that high ground as we're the one who's going to do this right. You know, whether that's enough to placate a Democratic Congress is a big question mark.

This is The Big Take from Bloomberg News. I'm Sarah Holder.

To get more from The Big Take and unlimited access to all of Bloomberg.com, subscribe today at Bloomberg.com slash podcast offer. If you liked this episode, make sure to subscribe and review The Big Take wherever you listen to podcasts.

It helps people find the show. Thanks for listening. We'll be back tomorrow.

Kapitelklicken springt direkt hin

Shownotes

Transkriptvom Podcast bereitgestellt · Zeitmarken springen hin