Anthropic Files for IPO in Global Race to Raise Cash for AI

Bloomberg Audio Studios.

Podcasts.

Radio.

News. This is Bloomberg Businessweek Daily. Reporting from the magazine that helps global leaders stay ahead. With insight on the people, companies, and trends shaping today's complex economy. Plus, global business, finance, and tech news as it happens. The Bloomberg Businessweek Daily Podcast. With Carol Masser and Tim Stenebeck.

Okay, so for the past, I don't know, two years, three years, we've been talking to Shireen Ghaffari.

We have.

And Carol's been asking over and over again, what are the numbers?

Sorry, Shireen.

What are the numbers at OpenAI say? What are the books at Anthropix say? Like, what have we learned? And we're always like, well, we want to see the books. And guess what?

We know.

We saw some of the books.

We haven't seen everything.

But the S1 has been done.

The S1 has dropped in this company.

Yeah.

Might go public in November.

Yeah, I assume, right? Let's see what Shireen Ghaffari has to say.

She is Bloomberg News AI reporter and has been covering everything on this company. Shireen, it is good to have you here with Tim and me. And we do kind of pester you, like, what do we know? As you have gone through the S1, it's long. We know there's a lot when it comes to risks. We just talked to one venture capitalist who kind of likes that they are like laying it all out there for everybody.

And she's invested.

And she's invested. She is definitely invested in this company.

Or she has exposure is how she put it to us.

And not open AI.

Yeah.

But she kind of likes that they're laying out Anthropic, some of the risks and putting it out there. For someone who's covered it, as you've gone through this filing, I don't know, what's interesting, what's important for you and what you think is important for the Bloomberg audience? Yeah.

Yeah, I think this is a major moment where we're seeing the real financials for the first time of a pure play AI company of this big generative AI boom that's been, you know, boosting the markets and has drawn a lot of excitement and fear in the past several years. And so we've had bits and pieces through reports, and we've seen the SpaceX IPO. But of course, Anthropic is not satellite business plus an AI business. It is a purely software AI business right now and the first really monumental generative AI business to go public.

Is it still an AI safety and research company, in your view?

I think it's much more than that.

It is now one of the fastest growing startups of all time. As this S-1 confirmed, there was something like over 1,000% increase in revenue in 2025, year over year. So those numbers are staggering. However, the losses are very large and notable. And the key question for investors is going to be, will this revenue growth keep pace in a way that is profitable eventually with these losses? And so an anthropic did disclose is that it doesn't uh you know it lists it out in its risk factors doesn't expect the growth to necessarily be 1000 year over year that they do expect this this to slow down but and we have reported on already uh some financials prior to the s1 uh showing some of that revenue growth slowdown however it's still a fast growing it's still

a highly revenue um in.

You know, intensive business.

So that's what investors are going to be looking at, how that measures to all those very, very expensive compute deals that they're also committed to at the same time.

So if you were sitting down with the folks over at Anthropic, what would you want to be asking them? And now that you've got, you know, you've been going through the S1, like, what is it that's top of mind? You're like, well, wait a minute, can I get a little bit more on this or what?

Definitely there are, you know, full free cashflow statements. Um, We also want to know, of course, what they're going to officially value themselves at, right? Also, we would like to know, I think there's so many questions, the exact details of some of their compute agreements. While some of that's in there, there's also still a lot to be parsed out. So I think investors are going to have many questions ahead, but this is the clearest picture that we have yet of, again, a kind of pure... generative AI software company and certainly one of this magnitude actually listing out its financials in this kind of audited manner.

So I've been asking all day, mostly rhetorically, because I don't think anyone knows the answer to this. What Sam Altman and the executives like over at OpenAI thinking, Sarah Fryer, Greg Brockman and the like, what are they thinking right now?

That's a great question.

To be honest, I have been so busy chasing the anthropic financials and analyzing and parsing those that we haven't had time yet to fully report that out. But yes, I think that it's fair to say that not just OpenAI, but the entire AI industry is looking to this as a litmus test of how investors, how Wall Street is going to react to the economics, the business model of generative AI.

And if this goes well.

It could usher in a whole new era of AI startups that are more encouraged to go public. If this doesn't go well, it could cause other AI startups to take pause and assess if they should delay their IPOs or take a different strategy altogether.

How would you define going well in this case? Would it be an IPO that is priced and ends up closing somewhat higher than the way it's priced on the first day, and then shares consistently stay above a certain level for six months. It's hard to measure this, I think.

Yeah, it can be subjective, right?

And there are famous examples, like with Meta's IPO, for example, back then it was Facebook, how initially, right, some hiccups, but long-term, a great investment. So I think it may be too soon to say, even in the short term, how this IPO goes in the history of time. But, you know, I think investors will be watching the reactions regardless, both in the short, medium and long term.

Yeah, it's just I mean, is it's fair to say still, Shereen, like I keep saying TBD, TBD. I mean, there's still a lot to be known about the impact and how this all is ultimately used and where it really.

Pays off in terms of return on investment.

Yeah, I think, again, a lot of people, you know, questioned even just the pure kind of revenue growth and whether we also see whether things like agents would even materialize in any meaningful way. I think we are seeing with coding certainly being the main use case that's bolstered anthropics growth that that is real.

People are using AI agents to code.

I think how that's going to be picked up by the rest of the economy, how people at work and other professions are or aren't going to start automating their workflow using AI, that's going to be the key question. And that's what Anthropic has to predict, which is a tricky balance as they're planning their compute spend and their very real debt obligations.

Stay with us. More from Bloomberg Businessweek Daily coming up after this. Okay, so I'm looking at the most read stories in the last eight hours. Number one, Anthropic files for IPO and global race to raise cash. You've got number nine, Anthropic publicly files for NASDAQ IPO under ticker symbol ANT. Number 11, Anthropic seven founders are worth more than $ 13 billion each. And the 12th most read thing on the Bloomberg terminal in the last eight hours is the actual S1. And all 94 pages of risks.

Kind of anthropic state.

I have it right here.

We're just living in it right now, to be sure. And trying to make sense of it. It's a lot.

The risks alone are many, many pages, some 70,000 words or so. There's a lot to it. And I have to say, a voice that we have... consistently gone back to to try to make sense of some of what's going on in ai this is an investor uh in smaller startups in the ai world we're talking about uh rudina cesari she's a long-time venture capitalist she's found her managing partner of glasswing ventures and she does invest in companies where ai is the core architectural foundation and has been investing in identifying and the next generation of winners as the AI disruption spend and impact continue to evolve.

It's hard to keep track.

But, Rudina, we do lean on you because we feel like you have given us heads up on things when it comes to AI and trying to understand where we are in this cycle, maybe where we're going, and the questions still to be asked.

Great to have you back with us. We know you're in Boston.

Anthropx filing, I don't know if you've read all of it, the S-1, but tell us from your vantage point, what's important?

So good to see you as always, Carol and Tim. So a lot of interesting tidbits come out of that very long filing. One, of course, noteworthy is that Anthropx revenue grew about 10x for the same period over last year. And there's a sense that it will continue to accelerate. And I don't know, speaking of, you know, paying attention to some of the details, for the next six months or so, most of the revenue for Anthropic is going to come from their most recent models, which you should hear thus as, you know, the most expensive ones, most likely, and probably the higher performing ones. Prior to the break, you talked about with a prior speaker around the notion of compute and what that means. It's interesting because revenue

has grown faster than compute spend, but compute spend has continued to go up. At some point, one would hope that the model training and pre-training becomes more efficient so that compute starts to taper off. But for sure in this case, oh, go ahead.

No, well, you know, we're showing for people who are watching on YouTube and are streaming the anthropic cost of compute as a percentage of revenues. And it used to be, go back to 2023, it was over a thousand percent of revenue. So it has come down dramatically, but it is still 55% of revenue. And, you know, that is a tough path to profitability in a big way.

No?

No. So two thoughts. One, just to double click on your very point, keep in mind the revenue grew 10x. So if you normalize it for the same amount of revenue, the spend has gone up. Number two, profitability. What's profitability in a tech company? So, you know, most of the time tech companies grow fast and they optimize for top line growth. I'm not sure that anyone is optimizing in this juncture for profitability. And that's the reality. And equally important going with this point of compute, it's the fact that they have over $ 600 billion in compute commitments. So they're securing their compute going forward. And about $ 400 billion plus, they cannot get out of those contracts. Put differently, they

have to continue to grow their revenue and expand their footprint to be able to sustain sort of the spend and then the innovation that they're pulling forward. Back to the S1 point, the risks were an interesting read and unique in that way. When was the last time we read risk and said, Customer concentration, litigation. Oh, and we might just wipe out humanity.

Exactly.

That was the first one. Let's just say WeWorks S1 didn't say that.

No, how do we take that in stride?

How do we really, as a society, be like, wait a minute, folks, this is serious.

It is serious. And I mean, we've had a number of exchanges and this type of dialogue together. And we've said that it's an incredibly powerful set of technologies that it is not inherently bad or good, but AI is extremely powerful. And I honestly appreciate Dario and Anthroping getting it all out there. And this follows, you know, prior announcement around slowing things down to put the guardrails around AI, because we've seen plenty of instances when even at this juncture of its development and perhaps immaturity, it is actually getting out of whack with attacks on other enterprises.

So it's real.

And we should take it seriously on the guardrail side rather than panic. And there are other areas, like you talked about climate, or this one, I should say, talking about climate and other facets. I mean, this is a group of people that have largely been... subscribers to the notion of effective altruism. And you could see that as it manifested in the S1.

Right. Yeah.

A lot of talk of the way that the founders will pledge 80% of their equity stake to each of their equity stakes to charity or give it away, I should say. Rudina, and just to lay it all out there, did you invest in Anthropic? Have you invested in Anthropic or in OpenAI?

Yeah.

We have exposure to Anthropic, yes.

You have exposure to Anthropic. Take us through how you did that and when you did that, how it worked.

So unfortunately, it's all confidential, but let's just say we expect upside from this IPO.

Okay, you do.

Okay, this is good to know.

Wait, so upside in terms of the share price? So the investment will grow?

And is it TBD in terms of how long you will hold on to it?

So again, I'm a private markets investor, not a public markets investor. So we have our own sort of approaches and strategies for how we liquidate and when we liquidate when we're holding public stock. But if you step back, fundamentally... Exposure to Anthropic is a very good thing. You see, not for any VCs, and there are many, many, many VCs that have invested in it. It feels at times like the entire market. And not only because of the returns that it can generate, but also a lot of the startup community, if they have to pick, Anthropic has been the trusted, more ethics-oriented partner of choice.

So for you, will this be the best return on an investment that you've made?

No, it will be one of the best, but no, we have other returns that have been, you know, when we went in and, you know, the company wasn't raising a hundred million right off the bat, you know, there is something to be said about going in early stages, much lower valuation. So we'll see if the markets come back to that.

And you said you have exposure to Anthropic, but no exposure to OpenAI.

Correct.

Okay. Why did you choose Anthropic rather than OpenAI? Or why did you, maybe the opportunity presented itself in a different way, but take us through that.

Honestly, because of some of the ethics considerations and the approach that they have taken, I don't want to look back 10, 20, 30 years from now and have been a negative contributor. I want to contribute to society that can leverage AI to do transformative things from reinventing how we conduct business in much more efficient ways to curing major diseases, to discovering new materials, to perhaps you know, going to outer space. I'm sounding like Elon now.

No, but I like your optimism that we'll be around in 10, 20, 30 years and the anthropic spots won't have wiped us out by then.

Well, I also cautiously warned you that the risks exist because it's a very powerful technology.

Regina, where are.

You on AI bubble, not bubble? There was a big debate among some, it's a bubble.

Oh, it's a bubble for sure. I mean, every company I see, it might be an overstatement, but many, many, it's two founders, brilliant research scientists, and they're raising $ 100 million.

Why?

Well, that's what it warrants. Why not a lot of explanation? When you get to that juncture, you have to think carefully also, how many neo labs do we actually need or does this world need? And, you know, will there be a point where maybe the top five or even 10, you know, sort of capture the market? And then there is a long, long tail. We will have a lot more visibility in 18 to 24 months on that point.

And whether a lot of these neo labs that have raised large amounts like we've never seen before, oftentimes mirroring that of a public offering, but done in an early stage private market, when they come back to market, will they actually have delivered enough and created enough value to warrant much higher valuations?

You said with the understanding that a lot of the private market investing you do is confidential and the terms are confidential, but can you contextualize for us the return that you'll see from the anthropic investment?

Wow, you're very determined. It will be a venture-level return. What do you expect from venture capital?

Well, the reason I ask is because some of these investors, you know, invested at a $ 5 billion valuation, you know, and if it's valued at $ 2 trillion, like a venture capitalist will literally have made a billion dollars.

Yep, and there are some much bigger funds. You know, who did I see? Spark Capital, who's actually our neighbors here in Boston and the West Coast, they've invested... I believe in six rounds. So it actually has favored the bigger funds quite dramatically because they've gotten bites at the apple multiple times.

Rutina, you know, there are bubbles and then there are bubbles. And I just wonder, just got about 40 seconds here. Is this a bubble that brings down the economy if it bursts?

Because of its reach?

Yeah.

Yes, so I will separate the valuation piece versus is there real potential in tech here? On the real potential in tech, absolutely. This is the most transformative technology we've ever seen. On the valuation, they will need to correct.

Is it a crisis, though? No, no, no.

We're about to actually get things done in ways that we haven't managed before. You should be excited.

Yes.

Stay with us. More from Bloomberg Businessweek Daily coming up after this.

Hey, so when the Fed chair reaches out to you for a one-on-one meeting, it's safe to say you probably say yes.

I would say yes.

Sorry, can't make it.

There's not a lot of people I would say no to.

There are tons of people you would say no to.

Okay, fine.

But probably not Fed Chair Kevin Warsh. That's what our next guest did back in early August. He did not say no. This was after the July FOMC meeting. It was before Jackson Hole. Let's get to it with Axel Merck. He is President and Chief Investment Officer over at Merck Investments. They've got roughly $ 2. 6 billion in AUM assets under management.

He joins us once again from Palo Alto, California. Hey, so good to have you here.

We followed some of the posts you did out on social, but Axel, I want to take a step back before you kind of get into the meeting and maybe the misperception that people have about this Fed under this new Fed chair. What's your background with Kevin Warsh?

Do you guys know each other? Do you play pickleball together or tennis?

Give us to the ranch?

Exactly. Do you share scotch together? What's going on?

At heart, I'm a monetary policy buff.

I am a nerd. And ever since the financial crisis, I have had views, as you know better than many, on monetary policy. And so I have attended some nerdy conferences, mostly attended by academics and policymakers. And I've gotten to know Kevin in that process over the years. I've probably known him for about 10 years now. And in Jackson Hole, at the very last sentence, I think he said he had solicited and unsolicited input. And I had the honor in being in the former camp. And just for those who know me through other channels, we did not talk about gold. I'd just like to point that out.

I'm shocked.

Well, okay.

What did you talk about?

Well, about lots of generic things, I suppose. It was a very good two-way conversation. The context, I think you framed it right. It was after the July press conference that I believe confused many in the markets. And he reached out to a variety of people. And then you've seen a change in a focus, I believe, that Kevin had at Jackson Hole and also in the September press.

I don't think that's a coincidence that the focus changed. I think what, if you just take a step back, he came to the Fed, he wants to change everything at an institution that has a lot of inertia and where many at the institution don't think anything needs to be changed. And so he came out with his task forces, And then in July, he kind of nibbled a bit in various places, and that created some confusion. And so I could just take one example from his, because it's one, I think, misunderstanding in the market. Each presser, he now mentions that the ample reserve regime is persisting. Well, it doesn't mean he likes it, but it means he's not just going to say, I'm going to do this or that without having everybody on board, because these things do

take some time.

You do need to.

Develop a consistent understanding on the Fed to introduce that and many other things. And obviously, one of the more important issues, and we talked about kind of many of the, I can't talk about the various positions, we talked about what we talked about kind of the perception of some things and my opinion and his opinion on some of these things. But obviously, as everybody knows, he is not a fan of forward guidance. And just earlier today, there was a very unfortunate op-ed by Alan Binder in the Wall Street Journal that talked about how somehow the Nobel laureates were embracing forward guidance. And I think that's confusing forward guidance with a reaction function. And those who criticize

Walsh for not having published a reaction function, well, you can't publish a reaction function if you don't have everybody on board. And so it's an amazing opportunity that he has. And I was rooting for him as he was running for the job. I continue to root for him. It puts me in an odd position because I have been very critical of the Fed under Bernanke and Yellen and Powell, most notably as well. And so it's very interesting to watch him.

So is he, you know, based on your conversation, based on what you think maybe people are misreading, is he the right, we always talk about, Axel, the right Fed chair at the right time.

Is he the right Fed chair?

Absolutely, absolutely.

Absolutely, because the Fed had, starting in the financial crisis, the Fed had gone on a slippery slope from ever more micromanaging the economy. And by the way, Walsh has been super consistent. Some people said he changed his tune as he was trying to, quote unquote, lobby for the job. But the reason, to take a step back, A lot of people don't realize that monetary policy is simply about allocating credit. It's about the amount of money in the economy, about interest rates. It is not about allocating that credit.

That is fiscal policy. And I am a purist in that I say that buying mortgage-backed securities already, and that's my personal view, is already wrong. Because buying mortgage-backed securities... allocates money to the mortgage sector. And what that does is it attracts political interference. Monetary independence is not about a tweet of the president. It's about the foundational underlying approach that does not look at fiscal policy. And Ben Bernanke, he may be the worst offender in that. He published a book towards the end of the pandemic that was praising how the Fed is ever more micromanaging the economy.

And that just gets you more and more into politics. And the politicians then rightfully tried to step on the Fed's turf.

Is that the same case for financial markets then, too, that they shouldn't be looking– necessarily at financial markets?

Or should they?

I mean, they obviously have to look at credit and rates and stress, but I mean.

The Fed has this roundabout way. They look at financial conditions, right? And it happens to be that risk assets happen to be correlated with financial conditions. And rightfully or wrongly, mostly wrongly, then people think about how the Fed reacts when the markets tank. No, I mean, the Federal Reserve, of course, is looking at the bond market as they should. And you need to look at the financial conditions more broadly.

But that doesn't mean, of course, you need to have a Fed put that always jumps in when something happens. And that was the risk since the financial crisis. that nothing bad is ever supposed to happen. Well, you need to have the markets to function. And it goes way beyond monetary policy because a lot of the young folks these days believe the system is rigged and they vote towards the socialists. And I can't blame them because the system has been rigged when everybody gets bailed out and so forth. And one of the great things in recent months is that We've been talking less about the Fed, right?

We've talked more about Iran. We've talked about fiscal policy. When the bond market has had a hiccup, we talk about Besant. And that's a good thing because there's political accountability. And that's where the short focus should be.

So is that, in your view, a sign that the Fed chair is succeeding at his job right now?

Yes. Now, I would call it a success already if we stopped going down further on the slippery slope. that we have been. And I think, yes, he has taken the Fed out of the news to a degree. Bad habits die hard. There's still too much forward guidance in some of the communication and so forth.

It's a process.

One of the bigger issues, I think, is a very tough one. He wants to reduce the Fed balance sheet. Paying interest on reserves is a terrible idea. Politically, it's terrible because you're literally paying billions to banks. But the other thing that's bad about it is that you don't see small issues in the banking system. In the scarce reserve system, when the New York Fed intervened in the markets, you could see things that were wrong in the market.

My guess is that the institutional pushback is going to be amazing to change back to that system. So I don't know what's going to happen there. But there are lots of ideas. We, of course, always focus on interest rates. But there are all these small things happening in the background that matter a great deal. And I think there's progress being made on those issues.

Speaking of things that matter a great deal, we talk every day, I'm sure you know this, Axel, about the amount of money that is being spent to build out on the artificial intelligence world that.

We're in right now. And every day.

It's increasingly into the credit side of things.

But there's also equity.

And I just wonder, do you also see, and I'm curious if that came up at all in your conversation, in terms of how the Fed keeping a watch on something that could be potentially a strain or stress for the financial system? Because it just feels so intertwined.

And we talk about circular financing.

I'm just curious how you see that or he sees that factoring in.

So I'll tell you how I see it, because we didn't talk about that specifically. But I see the big difference between the tech boom in the 90s and today's boom, that today's boom is much more debt driven. I believe Walsh is on record for having said that he much rather deal with a boom that comes in the current environment than many other situations. I think it's pretty clear that if or when There is a downturn because the implications for the debt markets, the Federal Reserve will react to that, and appropriately so, right? But in the meantime, I think one reason I believe Walsh got the job, and again, that's from me, is that he overall believes in the productivity boom that is there as a potential

because of the AI growth. He, at the same time, is a hawk at heart. And so he has to kind of bridge those two views. But we have a great deal of debt in this economy. And so outgrowing the debt is one of the few chances we can do this in a way that is not too disruptive. And to the extent that the inflationary environment allows this.

What you're saying, AI is a way for us to outgrow this debt?

The fiscal debt?

Well, potentially, because I mean, the massive investments that are taking place does boost growth. Now, obviously, at some point, there will be a pushback and the like. But if you keep in mind, right, when the railways were built, the infrastructure persisted. Now, maybe these assets are depreciating faster, but there are also tremendous advances being made. And And if you look at the rest of the world, the U.S.

growth is far superior than the rest of the world. So I think it's that mindset that helped get him the job. But at the same time, yes, he is a hawk at heart. He says inflation is too high. And one thing maybe before I go I should mention, I think the statement that he has to remove a dose of accommodation is telling about where we are. No need to use the sledgehammer, but we need to have tighter financial conditions to get inflation down.

Stay with us. More from Bloomberg Businessweek Daily coming up after this. Can we talk about it?

Let's go.

I was sitting there waiting for Dave Lee's column to drop. It's like we were on the editorial call this morning. We found out the S-1 was hitting. We're like, okay, we got to figure out what Dave Lee's writing about.

It dropped.

It dropped. As Dave Lee writes, every S-1 comes with long warnings about how a business might go belly up, but rarely have potential investors seen such breadth of risk. Everything from access to energy, to intellectual property rights disputes, to political strife with the current or future administration, to the bare face fear that AI could be misused, replicated, weaponized, or allowed to operate with a high degree of autonomy.

The risk factors, 94 pages, and more than 70,000 words.

Okay, that's a lot.

On the other hand, Dave Lee, Bloomberg Opinion columnist, look at all that money.

It's an awful lot of money, and that's the counterbalance, isn't it? You know, this thing could end the world as we know it, change jobs, change the economy, but... what people were looking for in this s1 was just how that business was performing and so you know this they're big numbers 16.3 billion dollars in revenue for the first six months of of this year um it's going to be a need to be a lot higher than that to justify all the spending that they're doing and also you know all these other companies are in the same boat um but yeah it's a big moment it's a I saw it described as a referendum on the AI trade, which I think is a fair description.

Are we living in a bizarro world where we are just, okay, well, they laid out the risks for us. I mean, there's a lot of stuff in there. And we know certainly with the midterms just around the corner and people pushing back on data center builds and different things, like things that could slow this down, supply chain problems. I mean, are we living in a bizarro world that we just kind of, people can't wait to get a piece of this?

I mean, it's a bizarro world. It's Dario Amadei's world, I guess, as well. I mean, listen, I mean, he, the CEO of Anthropic, he has made a brand of sort of talking about the worst case scenario, right? That's been what he's been doing. I mean, Anthropic itself as a company was built out of this idea that what was happening at OpenAI, they weren't happy with, so they went to found this other company.

What's different, really, between the two, though?

Well, Anthropic would say, and people question this, I guess, but Anthropic would say they're sort of more safety-minded. And that kind of runs through this S1, right? There's several pieces in the document where they say to potential investors, guess what? If we think this isn't going to be safe, we're going to slow down. That might cost us. That might cost us revenue. It might put the very company at risk if we don't fill up. We're going to get this in a good place. The way I sort of see this S1 is it's, and they sort of reference this themselves in the document. They talk about when they have a new hire potentially joining them. They talk about the anti-sell, which is the angle of this, right?

The anti-sell, which is basically they tell the hire all the things that might go wrong and all the reasons why this might not be the job for them. And then so a lot of people say, well, okay, fair enough. I'll go somewhere else. I love it. Some candidates will say, you know what, I think I'm up to that. And I think what they're trying to do is find the investor equivalent of those employees, right? The investors that will say, okay, we trust Dario Amadei and Daniel Amadei and the rest of the leadership team. And we're willing to sit back as this company spends huge sums on more compute capacity, keeps on spending big on R & D, chases all these use cases. They're going to need time to do that,

and I think that's what they're trying to find with this prospectus.

Look, in the next hour, we're going to speak with Paul Sweeney, who did spend 20 years on Wall Street and lived to tell the tale of not just the financial crisis, but also the dot-com boom and bust of the 90s and early 2000s and reminded us it took 10 years for the NASDAQ to get back to where it was pre-bubble bursting. In terms of historical corollaries... Is SpaceX kind of the only one of these? Because just massive size, unprofitable company, a total addressable market that overlaps a little bit.

Only one in what sense?

The only other company that we can sort of compare the size and scope of Anthropix S1 and IPO2.

What I think what's particularly interesting about SpaceX, and we've seen Elon Musk on Twitter in the last couple of days talking about this, They have another business, right? Which is, as well as AI, they have rockets, right? And they have Starlink, which is doing.

Maybe they'll have cars soon, too.

Maybe they'll have cars once all of these things join together. But that's sort of.

And robots.

Sorry.

That feels like it's a tougher thing for you. That's another story altogether. But I think the point he was making, though, about some of the other companies, like the NeoCloud, you know, these companies that are set up just to be computing factories for these AI companies... It's that it's quite hard to do this if you don't have something else, right?

SpaceX has rockets. Meta has the rest of Meta.

Advertising.

Advertising, all those apps. Google is, obviously, Google has a ton of things. I do wonder how difficult it's going to be when you're a company like Anthropic and the AI is the entire play. It's the pure play. And so when you release a product, and we've seen in the last few days that Anthropic's launched. a competitor to Muse, Meta's AI assistant, they've called it Claude Duet, or Claude Duet, I've actually forgotten the name. They have to charge for that immediately. They can't subsidize it in the same way Meta can, because that's their entire business. And so I think that's, you know, whereas it can be compared to SpaceX in some degree, actually, SpaceX has got much more ways to sort of move the money around, which obviously

Elon's very good at.

And unlike OpenAI... which uses advertising in the free version, at least, of ChatGPT. Anthropic's not doing that.

They have very much committed to not doing that, haven't they? There was the ads around the Super Bowl, very, very effective, very cheeky, got under Sam Altman's skin, certainly. And they reiterated it again in this prospectus. They said, we've committed to not doing that. I would love a sort of crystal ball five years from now, whether Anthropic is going to kind of pull a Netflix and say, we don't want advertising ever. only to then think, oh, you know what, maybe we do.

Not so bad.

Yeah, yeah.

I still, though, you know, Dave, I want to understand, there's multiple large language models, and then there's what China's doing. There's like a lot coming at us, and I'm trying to understand, is it going to be a world where, I don't know, Claude is good for this, Gemini is good for this, the Chinese open source is good for this, for those of you who just want travel advice. Like, I don't understand how it's going to ultimately play out?

Do we need all of it?

I think it's a really interesting point because a lot of the early backing of these companies, particularly OpenAI, was the idea that they would be the one, right? Or that some one company would be the one. It's not really shaking out that way, is it? We're seeing different models to be good at different things. I had breakfast recently with the CEO of a legal AI company doing very, very well. And he was saying that they swap out the models every couple of weeks.

Wow.

Depending on who is doing well at certain things.

So their product is built in a way that you can just, these things are interchangeable?

The brains are interchangeable? Yeah, so for example, one model might be better at drafting a legal document. Another model might be better at finding citations, etc.

etc., etc.

But the models that are good at each of those things, it changes depending on when new releases come out and what frontiers are reached. And he was saying that it's simply a case of a line of code for them to change that out. And so I've got to wonder when you look at Anthropic and they spend billions for what, two weeks to be the model of choice? It feels like a tough gig.

And who owns all the data that they're getting? Like, where are they getting all this data? Like, I keep thinking about, to make these models better, they need to keep pulling in stuff.

Like, I keep.

Thinking about whether it's stories written on Bloomberg or, you know, research, medical documents.

Or when people are uploading to these things.

Right.

Like, we are helping make them all better.

Yes.

I mean, it's all of the above.

Isn't it at some point it's going to be like, well, wait a minute, that data set actually belongs to.

I mean, I think that's where some people would argue we're right there, right? I mean, in terms of what music's being scraped up, what articles are being scraped up. You know, there are a number of lawsuits between content providers, say like the New York Times, for example, which is suing OpenAI. And look, I mean, eventually there's the other problem, which there might just not be enough data left to keep on training, right? And so one of the things that some of these AI models are seeking to do is to have sort of synthetic data, you know, create new stuff to then be learned from instead.

This is the Bloomberg Businessweek Daily Podcast. Available on Apple, Spotify, and anywhere else you get your podcasts. Listen live weekday afternoons from 2 to 5 p.m. Eastern on Bloomberg.com, the iHeartRadio app, TuneIn, and the Bloomberg Business app. You can also watch us live every weekday on YouTube. and always on the Bloomberg Terminal.

Anthropic Files for IPO in Global Race to Raise Cash for AI

Kapitelklicken springt direkt hin

Shownotes

Transkriptvom Podcast bereitgestellt · Zeitmarken springen hin