Bloomberg Tech is live from the heart of Silicon Valley, with Ed Ludlow in San Francisco.
This is "Bloomberg Tech." Anthropic has filed publicly for an IPO and intends to list on the Nasdaq under the ticker ANT. This is the deepest look yet inside one of the world's biggest AI companies. The growth has been extraordinary, but Anthropic tells its investors plainly that growth will slow. At the same time, it's looking at hundreds of billions of dollars of future compute spending, even though most of its revenue depends on how much customers use Claude. It's a big step for an IPO that we report is on track to happen in mid-November.
Joining us now, Bloomberg AI reporter and chief Wall Street correspondent. Finally arrived. What did we learn? What is new?
So we know that Anthropic is listing on the Nasdaq. It's official. We have some solid revenue figures in the past as well as losses, in line with what Bloomberg has previously reported there, so no big surprises. But this is the first real financial set that we are getting from a pure-play AI company, right? So this is huge for people trying to analyze both extreme potential but also risks and weaknesses in the GenAI business.
$16.3 billion versus $1.2 billion in the same period a year ago. We think that they will try to match or exceed what SpaceX did in its IPO. Bring in the Wall Street perspective. We know who the banks are on this deal and in what order. Tell me what I need to know.
It leans left. There was a bit of a surprise. There was some expectation Morgan Stanley could be leading this IPO, but Goldman Sachs led left on SpaceX, then they brought the Google offering into the market. Now look at the slate of announcements, and I need to point that out because it is a banner day for capital markets. You know, when you think about investment banking, we go through periods of feast or famine, but Monday morning seems like a period of feast. Goldman was lead left on this Anthropic S-1 announcement, the $95 billion Meta fundraising announcement that came through, as well as structuring the $40 billion chip financing for SpaceX with Apollo. That is over $200 billion of deals that's been arranged led by one bank. Yes, there are others, Morgan Stanley and others are playing a lead role in many of those. But the fact that Goldman Sachs takes top billing is a big day for the bank.
Honestly, as we go through the next few weeks and head into the middle of November, this will certainly be a big test. If they can execute this properly, there were a lot of question marks around SpaceX and how it would trade after the IPO. It seems to have done just fine. If they could pull it off with Anthropic, that would be a success.
There was a clear warning in the risk language about the impact of AI on humanity. Please explain that.
Sure. Anthropic has long been a company that's raised risks about the consequences of this technology if it goes wrong, and that was a message loud and clear in the risk factors. The company said that advanced AI could pose catastrophic or existential risks to humanity, that its models have exhibited behavior including resisting shutdown, concealing and manipulating information, and behavior resembling blackmail. So in essence, Anthropic is saying our technology, if it does not have the right safeguards, if we do not figure out the security, could end up wiping out humanity — existential risk. That is quite notable for an IPO, not something you normally see in an S-1.
There are no third-quarter financials, really. There is no addressable market figure, which SpaceX did include. It says, here is the world we see in the far future. But I think it's important for us to look at the founding group. So my read on it: Dario Amodei stays CEO, Daniela Amodei stays as president and chair of the board. This idea that the founders have voting control, essentially. Explain the confirmation of the management team.
That's right. Anthropic is a company that has seven co-founders. They remain at the company, a tight-knit group. This S-1 solidifies that control and power that they have, and that is what we are seeing in this new share of stocks and that LLC that you mentioned.
I am looking at some of the financing language. I know the revolving credit line, there is an update there. Go back to what you were explaining a moment ago. There isn't much in the filing itself, but based on Bloomberg's reporting, this is now a pretty tight window for Anthropic to do this IPO by Thanksgiving.
When you start getting into the second week of December, that's when markets shut down. You really don't bring new deals then. Now we know that Anthropic is trying to pull this off the week before Thanksgiving. Going back to your point on the S-1 filing, when I reached out to investors this morning, asked them for their first take, a lot of the top-line numbers had been leaked into the market. We had a sense of what they would look like. What they really want to get are specifics in Q3, and considering that this could be a mid-November IPO, will we get a whisper number on the Q4 revenue expectations and what is the pace of revenue growth? Is there going to be significant deceleration? Because if that's the case, there will be pressure on Anthropic to further trim their valuation expectations and perhaps raise even more than $100 billion if this is what the window allows.
Okay, our AI correspondent and chief Wall Street correspondent, thank you very much. Let's keep the conversation going. Joining us now, Gil Luria. Start with what you feel you learned. Is anything new or changed the story around Anthropic?
I think the most important things I saw are, one, this is a profitable company. I think there's been a lot of assertion that Anthropic and OpenAI lose money on every query. That is not the case. Their revenue is twice as much as their total compute costs, including training. And that was unclear from all the leaks, and it is now very clear in an SEC filing. They have $130 billion of cash and equivalents. That means the IPO is not an act of desperation. They actually have a war chest that could last them for a while because they're actually not burning through that much cash. The cash is not in there, but this is a company with not that much capex, which is to say they're actually in okay financial shape.
The one thing I think the reading of the S-1 will take longer to understand is the public benefit corporation aspect of this. That means the company is saying, we are not just here to maximize shareholder value. We are also here for the benefit of humanity, which we get to decide what that means. That's the part that I think is going to be a little bit of a tougher read.
I want to put some numbers behind the points you are making. Cash and short-term investments, approximately $131 billion as of June 30 on the balance sheet. Total liabilities $137 billion. You said this was a company that is profitable, and what that looked like in the second quarter was GAAP operating income of $974 million and GAAP operating margin of 8%. The bit that I want to get to, that I feel like no one is talking about as much, was the completely stark and codified warning that growth is going to slow, and that it will simply not be at historical levels going forwards. Is everyone pretty calm about that?
I think we got the news that the third quarter decelerated from the second quarter last week, so we understood that. But just by definition, a company can't multiply itself by 10 times every year. Clearly it's going to grow less than it did last year. It was still disappointing that Q3 grew slower than Q2 in terms of the revenue they added. It was really just a feature of the fact that their market share was almost 100% of the enterprise in the middle of the year, and then OpenAI caught up, then open-weight models caught up, and therefore their market share is much more moderate than that. That's going to continue to be the case. So as you were discussing earlier, it's very important that the fourth-quarter guidance come in at a nice growth rate, that there be 2027 guidance for growth. Those are pieces that will come out during the roadshow. They're not in the S-1. It is important they have continued growth, that they're not peaking in terms of revenue.
The 3Q revenue number was not in the S-1. It's $16.8 billion, which is up from $11.6 billion in the second quarter and would represent 45% sequential growth. But the overall trend is decelerating growth. The numbers that jumped out at me is the cost of compute and economics. So in 2023, compute was 10x revenue. In the first half of this year, it was 55% of revenue. That is positive, right? That's the direction of travel you'd want to see.
Yeah, that's very positive. Again, it goes to the essence of the whole AI trade. It depends on any model company, regardless of whether it's Anthropic or others that will emerge with OpenAI, having to be able to sell that compute that they're paying for at a higher price. Right now it seems like not only is that the case, but the profit margin, that gross margin, revenue over compute, is actually increasing right now. And that's a very good sign for the AI economy because that has to be the case for them to keep buying compute from the hyperscalers and others, and so the hyperscalers and others can buy semis, so they can buy construction and energy, etc. The whole AI economy depends on that margin. That's why to me that was the highlight of this S-1.
I would say generally speaking, technology stocks are higher, right? While I don't draw too much of a causal link between the S-1 and the fact the Nasdaq is up, there has been a bit of life in the AI trade, I think. Sometimes I would ask you to draw on your experience in markets here a little bit, if you would. It's as much about what you don't say as what you do. There is no sort of big number in this document. SpaceX gave us a very big TAM number just for AI, which was 26.5 trillion. What do you make of that?
That could still come out during the roadshow process. I would say that when they talk about their TAM, they're not going to talk about the TAM for software. They will either create a new number like SpaceX did, or anchor it to something along the lines of what SpaceX did, or what they're more likely to say is the market for labor is tens of trillions of dollars. That is the market we are addressing because we are going to make people more productive. We will displace some labor, and that is the addressable market, and that's such a big addressable market that it allows them a runway for growth. I still expect that to come out during either the next iteration of the S-1 or during the roadshow.
In the SpaceX S-1, it was this sort of Elon Musk story. The risk factors very highly correlated to Elon Musk participation in the future endeavors of SpaceX. There's less of that key-man risk presented in this filing, based on the hours I've spent reading it so far. The seven founders get 50% voting role. The president and chair stay the same. Do things like that matter when you are trying to pull off an IPO like this?
It should matter to investors. Usually the controlling stake stays with one or maybe two founders, the case for Google. Here we are going to have these seven people, not just Dario Amodei and Daniela, but seven people that are going to have controlling interest. That's a whole other dynamic of who is going to want what, and how are they going to vote, and what's the best interest. So this is a unique permutation, but it has that big disclaimer that we often have with these types of IPOs, that you as a shareholder will never get to a controlling interest. That is always going to stay with the founders, and in this case there's going to be a new level of opacity around what the seven people agree on, as opposed to knowing what Elon wants or Mark wants. Here we will have to wonder what those seven people agree on, and whether there will be some dynamics there of them disagreeing on things and therefore changing the trajectory of the company.
Again, can't emphasize enough, they went through a section of public benefits, which is, we are not always going to maximize value. We don't want our share price to trade above the value of the company. These are not typical things. This is highly unusual, and investors are going to have to really get comfortable with that, that they're going to trust this group of seven people to decide what's good for humanity, to decide what is good for the company, even if that doesn't mean maximizing shareholder value.
The mission or public benefit is the responsible development and maintenance of advanced AI for the long-term benefit of humanity. Everyone will try to model for how you run the economics of the company and the profit for shareholders with that public benefit in mind. Really great having you with us on "Bloomberg Tech." Thank you very much.
Coming up, how are markets digesting Anthropic's S-1 news? The broader competition for capital to fund the AI boom. There are lots of other financial transactions this morning. This is "Bloomberg Tech."
Anthropic has filed publicly for an IPO. Its S-1 is out today. Fresh numbers around one of the biggest bets yet on artificial intelligence. There is life in the AI trade. It's hard to know if what's driving market sentiment is specifically what is detailed in the S-1. We are also in this earnings period where there is a more complicated picture. Where is there still value? Where are the bottlenecks? Where is the next phase going to emerge? Lori Keith from Parnassus Investments is with us and has the unenviable task of trying to make sense of it all. Let's start with the S-1. It's a moment that equity markets have been waiting for, and I guess to start, what did we learn about the world that these AI and frontier labs are operating in?
Thank you. It's great to be here, Ed. I would say this is a tremendous milestone that we are seeing today, as you are really seeing AI move from these pilots and experimentation to now a focus on what are the companies that can truly capture that economic value going forward. So I think there is no doubt that AI is extremely transformational, and that's going to create a huge opportunity not only for some of the frontier players but also these downstream second-order beneficiaries that really are participating in the buildout of AI.
It is specific to Anthropic, but I think people had some concern about concentration risk, the number of pies. So when I was going through the data, I looked at the customer data they have. There are almost 12,000 customers that spend $100,000 a year with Anthropic, which is up from about 950 customers a year ago. Big growth in that cohort. There are almost 1,900 customers spending $1 million a year, and then you can extrapolate out there from the midpoints of the market. What does that tell you about where Anthropic's growth actually is?
It's a good question. Ultimately what it shows there is this insatiable demand. We have obviously seen it really on the forefront in terms of engineering and coding compute needed in those use cases, and as those broaden out into things like customer service and other applications, we are seeing a number of our companies that we are involved with really implement AI to help drive efficiency and strengthen those competitive advantages. So I do think we are still early days in terms of the adoption of AI, but we think there is a tremendous long runway not only for the enablers, those companies that are leveraging the compute and providing the chips, but also many of these downstream players such as in the infrastructure space, the power, water infrastructure. There is a whole host of companies that we think have very strong competitive advantages that will benefit throughout the buildout of this phase.
It is an extraordinary morning because Anthropic filed publicly for an IPO, but Meta also detailed its plans to seek up to $95 billion in equity, and on the credit or debt side, SpaceX agreed to a quite complicated chip financing deal with Apollo. The umbrella point: these companies will still need capital.
Absolutely. For sure. I think it speaks to where we are in the cycle. Certainly we are still very much in a capital phase of this buildout, but looking at who are the actual beneficiaries, the first and second-order beneficiaries will benefit from this spending. We see a whole host of particularly interesting companies in the mid-cap space, where companies that are essentially providing that infrastructure pipeline — be it the utilities that are participating and expanding their grid, some water technology players. We think all of them will really be well poised to generate compounded earnings over the next decade due to the spending that will be needed today. And certainly compute can be driven by Moore's Law, but when you are thinking about establishing or building out a new power plant, that doesn't align with the speed of the law. That takes much more time, and frankly that's a much longer investment cycle, but we see interesting companies in that area.
Another data point I will put out is, for example, Anthropic, which is a frontier lab, listed TSMC, world's number one chip manufacturer, as a risk factor — the chips they need to compute. We are in earnings. What have you learned about the AI story specifically from the earnings we have had so far?
It's a great question. AI, there's been really a couple of things that I would point out. First of all, I would say companies that have been putting up really great numbers aren't being rewarded for that yet. We are still early in the earnings cycle. Less than 25% of the companies on the S&P, and even mid-caps are lower today. But I think ultimately the valuations do matter, and expectations are high going into the earnings. So we did see a company such as Lam Research, which reported last week, they had an excellent quarter, beat numbers. The stock actually still fell back. Ultimately, though, we think that is creating opportunities for some of these really strong players that have strong competitive advantages given their positioning around advanced packaging, around memory within the equipment space.
Lori Keith of Parnassus Investments, thank you. What a morning in the world of technology and finance news. Coming up we will bring you those details. Meta seeks to raise up to $95 billion to help fund its AI spending, and SpaceX looks at a very interesting financing mechanism with Apollo. Both of those when we return. This is "Bloomberg Tech."
Meta wants to raise up to $95 billion from selling common and preferred stock to support its AI investments. Joining us with the details, our Bloomberg big tech editor. This relates to what we first learned last week.
In earnings last week, Meta said it was going to be creating a new class of shares, a non-voting class through a stock split, and that would happen in January. So part of this new offering is going to be the common stock, and part of it is that new share class. What Meta said about that, beyond raising money for the AI boom, is this is going to allow them to bring in different kinds of investors, and it's also going to allow them to let some stock get sold without diluting or reducing the stake of Mark Zuckerberg, the CEO of Meta. To remind everyone, Google and Intel both made very rare stock offerings earlier this year that were both oversubscribed. Google's was $85 billion. So this is even bigger than Google's, and there's been a lot more skepticism around Meta's. It will be interesting to see this alongside the Anthropic IPO.
Sarah, thank you very much indeed. Real quick, Elon Musk said that SpaceX has closed a $40 billion equipment finance lease with Apollo Global for the chips needed to fuel its AI ambitions. The fixed rate is yet to be determined. The deal is expected to close before the end of the year. The lease is expected to start the first half of next year. Halftime here in the city of San Francisco. So much more to come on this. This is "Bloomberg Tech."
Welcome back to "Bloomberg Tech." Anthropic has filed publicly for an IPO, and it intends to list on the Nasdaq under the ticker "ANT." Without being too much of it directly, there is some energy in the AI trade. People have had little time to digest that. Right now, the Nasdaq up 1.5%, the stocks up 3.4%. We also have breaking news crossing the Bloomberg Terminal: Microsoft offering buyouts to 60,000 employees. Here with us with the story.
60,000 is a lot of people. We have seen a lot of tech companies try to shed their workforce, given Microsoft's and Google's and Meta's of the world are betting more they can do more with less people with AI. This is the latest strategy Microsoft is taking to get some costs off its books.
Buyouts, they do that typically? Reduction?
Very limited. They did a retirement buyout earlier this year, that was the first large-scale one we have seen. To me, this is a bit friendlier version of the Elon "extremely hardcore" Twitter buyout. A lot of people remember when he bought Twitter: if you are sticking around, it will get tougher, and if not, you can take a buyout. That is energy from Microsoft. They are saying we need to be focused on this mission, and if you're thinking of not being here, maybe you shouldn't be.
As of June 30, Microsoft has 223,000 employees, so that buyout number seems big. Bloomberg's Brody Ford with the news you need. Let's get back to today's big story: Anthropic's S-1. There's a lot we need to go through in terms of what Anthropic said about the future. Bloomberg's benchmark capital reporter is with us. I want to talk about specific disclosures they made, and the reason is they are on the precipice of a giant IPO, but they couldn't be more clear about how they see future financing, etc.
Anthropic has around $141 billion worth of cash right now, as of June 30. Quite a lot. But CEOs also talk about IPOs being a fundraising event. It is so Anthropic is reminding people it will need a lot of cash in order to keep funding its endeavors. At this same time, we have been covering this rapidly, and we are getting a heads up that will continue. It's clear they will need to fund it through a variety of means.
For the "Bloomberg Tech" audience, there is a lot not in this S-1. But what we look for is 5% ownership. You were on the show last week and detailed some of the big potential winners from this IPO. Actually, what is not there and what we do not [see]?
You will not see any — we will not see the big shareholders like Iconiq, Sequoia, GIC, Menlo Ventures. There is one venture firm in the filing, Spark Capital. Their general partner is the only board director from the venture capital world. They are still not above that 5% stockholder name we are looking for, but Yasmin is noted as one of the few board of directors. So the tiniest glimmer of...
Venture capital name in one of the biggest venture capital exits of all time.
So what is missing? We do not know the IPO size, target range, price range, valuation, no Q3 financials, no TAM number. Those kinds of things will come. Something interesting in the background, we just did the big Meta marketing raise — OpenAI is trying to do a private round while this is going on to remind us of that and why it is important.
