Yeah, it's a crazy number, but they got the AI business right, just like Google got internet marketing right 25 years ago, as long as they keep the momentum (and prices).
> google launched at i dunno a 1bn market cap and grew as the internet grew
Google closed its first trading day "worth more than $25 billion" [1] with $1.5bn revenues, up 234% from 2002's $440mm [2]. That's a 16x revenue multiple, 0.07 growth adjusted [3].
On a similar revenue multiple, Anthropic's $16.3bn half-year earnings are worth $543 billion. Adjust for Anthropic's 1,400% YoY growth and you're well over $3 trillion.
Obviously, it's a lot harder to grow at a thousand percent from a $16bn baseline versus a few billion. But purely on a multiple basis, Anthropic is in a similar ballpark as Google.
As another check, SpaceX is trading at ~62x sales with a growth rate in the 30% range. Assigning the same sales multiple puts Anthropic at ~$2T market cap. Add in a factor for much faster growth (and non-GAAP profitability) and $3T is a really easy place to end up.
Sure, you can throw any numbers around, if you wish to get any valuation you want.
In reality, a valuation based on revenue (sales) is bonkers to begin with, but even worse when you consider that Anthropic has so much debt that it doesn't have positive equity value, it's in a deficit ("When liabilities attached to an asset exceed its value, the difference is called a deficit and the asset is informally said to be "underwater" or "upside-down""). At least SpaceX has more assets than liabilities.
There are a set of mistakes in what you wrote, but
> SpaceX has more assets than liabilities.
is definitely a choice thing to write on the day SpaceX announced a $40B debt deal to purchase compute to run other companies' models. Based on available information, this will put SpaceX into a place where it has more liabilities than assets.
Larger point stands. SpaceX is commanding a 62x sales multiple, whether or not you personally believe price:sales should be relevant. SpaceX's AI operation is smaller and growing slower than Anthropic's. It's reasonable to expect investors to benchmark Anthropic against SpaceX, it would be silly not to do so. A normal thing one would expect investors to do is assign a higher multiple to a company that grows faster.
But the question is not so much if they can IPO at those prices but rather if the valuations are sustainable over a couple of years. If they are priced to perfection there isn't much left for public investors besides huge risk that something might not work out 100% as expected.
> Some people will benchmark Anthropic vs. SpaceX, that doesn't make it a smart approach.
Yeah, I would not pay 62x sales for SpaceX. And I don't buy IPOs. I'm not endorsing these values, I am saying benchmarking against peers is what investors commonly do.
> A normal thing is to consider risk, liabilities, cashflow, how diversified revenue streams are etc., and other things which you seem to be missing.
I'm not missing those things, I am not including them here because being "right" is rarely how markets value companies. (Also because the rest of that stuff is in the S-1.)
It's possible to build a "correct" model of a high-growth high-tech company that looks primarily at risk, liabilities, cashflow, and diversity of revenue streams. That model might put Anthropic at (say) a $100B valuation. That "correct" model is going to be less predictive than a model that starts from what similar firms are trading at in public markets. I personally am more interested in predictive power than correctness when evaluating stocks.
> I am saying benchmarking against peers is what investors commonly do.
You are benchmarking against an outlier stock with a cult following (Elon Musk) and diversified sources of revenue. Anthropic does not have either.
If it was as easy as comparing the hottest new IPO with Elon's companies and applying the same multiples, people wouldn't have lost so much money on electric car IPOs that pretended to be Tesla (see PSNY -99.5%; NKLA -99.99%; FSRNQ -99.99%; etc.).
> being "right" is rarely how markets value companies.
Conventional metrics are how most stocks are valued most of the time. What you want to say is that market can be exuberant sometimes. That's fair. When the exuberance goes away, valuations come back to earth (see PSNY -99.5%; NKLA -99.99%; FSRNQ -99.99%; etc.).
> That "correct" model is going to be less predictive than a model that starts from what similar firms are trading at in public markets.
Even if we take your statement at face value, you're still not doing it right. A valuation based on comparisons has to have more than 1 peer to compare against. Check out the multiples of other model providers listed on the stock market. You'll probably be shocked that it's not a double digit multiple of revenue.
And then, if you really want to have a proper model, isolate the AI revenues of SpaceX and other hyperscalers, then we can argue that you have a model worth something.
> You are benchmarking against an outlier stock with a cult following (Elon Musk) and diversified sources of revenue
So in this case, diversified revenue is a drag on SpaceX because all the non-AI revenue is growing considerably slower.
But your point about Musk is taken. Cerebras went out in May, and despite being a hardware company is trading at 37x sales. Palantir trades at 78x, CrowdStrike at 50x, Cloudflare at 47x, etc. Anthropic is growing significantly faster than any of those companies. I'm not making the case that Anthropic will/should trade at any of those, just that Anthropic trading in the range of the P/S of SpaceX would not be crazy, and it would not require Musk to be involved.
> A valuation based on comparisons has to have more than 1 peer to compare against
Yes, the best peers are "public pure play frontier model providers trading in the US." There are none. SpaceX is closest, despite not really being a frontier provider and definitely not a pure play. But you are correct that investors will use a basket of stocks. It is silly to suggest SpaceX will not be in that basket.
You absolutely can make an AI doomer basket to model Anthropic. Use "real infrastructure companies" as the connective tissue since AI wants to be infra. In your basket put Caterpillar, Kroger, The Southern Company, AT&T, Comcast, and ExxonMobil. You can absolutely make the argument that Anthropic should trade more like them than like CrowdStrike or Cloudflare, that is a coherent position. Then based on that infra peer group, calculate a valuation. I think this model will be much less accurate in the near term than a model that uses other high-growth tech companies.
Price discovery is part of the fun of IPOs! The doomer model could be right, the stock value could collapse toward irrelevance as fast as the exchange's circuit breakers will allow!
> So in this case, diversified revenue is a drag on SpaceX because all the non-AI revenue is growing considerably slower.
Your argument is like saying that having gearbox in a car is a drag on speed, why not remove it? Just take out the transmission and the windshield while you're at it too.
> Cerebras went out in May, and despite being a hardware company is trading at 37x sales. Palantir trades at 78x, CrowdStrike at 50x, Cloudflare at 47x
None of those companies are AI labs, you're just cherry picking random expensive stocks and doing the whole price/sales nonsense again. All of them have positive shareholder equity (Anthropic has shareholder deficit).
> But you are correct that investors will use a basket of stocks.
That's your approach, I was just quoting it.
> You absolutely can make an AI doomer basket to model Anthropic. Use "real infrastructure companies" as the connective tissue since AI wants to be infra. In your basket put Caterpillar, Kroger, The Southern Company, AT&T, Comcast, and ExxonMobil.
At this point you're either being silly or arguing in bad faith. Your best case scenario is for Anthropic to trade like a stock with cult-like following and your worst case scenario is for it to trade like a ketchup stock. Completely ignoring the most obvious tech companies that are traded publicly and also provide models of their own. This conversation became so absurd that it's not worth answering anymore.
The mental gymnastics you're doing here are appalling.
Someone on a different forum fed the S-1 in Fable and it said "Shows a massive “loss” on paper, but that’s mostly an accounting effect from the company’s value going up"
Did they get it right? I'm geniunly curious because from my point of view the only company to get AI right so far is Microsoft. This is from the point of view of an EU energy company married to Microsoft services mind you, but Microsoft is the only AI company which actually sells AI to us in a manner that can be considered useful. 7 months ago I thought Copilot (the m365 one, not the github one) was useless. Now I think the chat part is as good as what you'll get on the smaller subscriptions to ChatGPT or Claude, and it's Cowork is very good.
Obviously it's not like having Claude Code or similar, but those would be very hard to run legally and safely in EU regulated industries like us in energy production. You could run Codex but it would require your IT department to have yet another place to manage security and software integrations. Which is a fairly big issue when you have around 1 IT staff member per 300 employees, and an organisation which will buy five different sharepoint solutions for the same problem.
Microsoft sells something that seamlessly integrates into everything else you buy from them. It's easy to manage credit spending and when I say that I don't mean that it's easy to set it up. It's easy to track and budget so the right departments pay for it. It can read your emails, your onedrive, your loop, your sharepoint, your powerbi/fabric and anything else you have access to. Sure it's annoying to be confronted with a point you've made 7 months ago (like how you think Copilot is utter trash) and since changed opinions on it's frightingly good at finding relevant stuff, even in some obscoure legal document on your intranet.
I assume that Anthropic and OpenAI makes money of this somehow, but for the most parts our users wouldn't care if it was the GPT and Claude models or Mistral or Deepseek. Which is why I asked if they had really gotten it right.
Yeah, it's a crazy number, but they got the AI business right, just like Google got internet marketing right 25 years ago, as long as they keep the momentum (and prices).
Isn't the issue tho, that google launched at i dunno a 1bn market cap and grew as the internet grew.
what you're buying into here is a company that is already large on the idea of growth over the next 10 years.
Modern IPO's just reek of investors cashing out
> google launched at i dunno a 1bn market cap and grew as the internet grew
Google closed its first trading day "worth more than $25 billion" [1] with $1.5bn revenues, up 234% from 2002's $440mm [2]. That's a 16x revenue multiple, 0.07 growth adjusted [3].
On a similar revenue multiple, Anthropic's $16.3bn half-year earnings are worth $543 billion. Adjust for Anthropic's 1,400% YoY growth and you're well over $3 trillion.
Obviously, it's a lot harder to grow at a thousand percent from a $16bn baseline versus a few billion. But purely on a multiple basis, Anthropic is in a similar ballpark as Google.
[1] https://www.nytimes.com/2004/08/19/technology/shares-of-goog...
[2] https://www.sec.gov/Archives/edgar/data/1288776/000119312504...
[3] https://www.investopedia.com/terms/p/pegratio.asp
If only it was that simple, to switch multiples around between random companies and come up with valuations
Google was profitable, free cash flow positive and had no debts basically
Anthropic is burning money by the millisecond, no profits, free cash flow negative and has lots of debts & commitments
Correct does that poster actually know what comparables are?
Comparables are not by peer group or sector.
It’s cash flows, growth and risk bud. If you want to do peer group - go ahead - just don’t forget to control for differences..
Price to sales multiples are internally inconsistent by definition.
Is there an internet rule for folks asking if someone else knows something following it with a completely wrong definition of that thing?
As another check, SpaceX is trading at ~62x sales with a growth rate in the 30% range. Assigning the same sales multiple puts Anthropic at ~$2T market cap. Add in a factor for much faster growth (and non-GAAP profitability) and $3T is a really easy place to end up.
Sure, you can throw any numbers around, if you wish to get any valuation you want.
In reality, a valuation based on revenue (sales) is bonkers to begin with, but even worse when you consider that Anthropic has so much debt that it doesn't have positive equity value, it's in a deficit ("When liabilities attached to an asset exceed its value, the difference is called a deficit and the asset is informally said to be "underwater" or "upside-down""). At least SpaceX has more assets than liabilities.
There are a set of mistakes in what you wrote, but
> SpaceX has more assets than liabilities.
is definitely a choice thing to write on the day SpaceX announced a $40B debt deal to purchase compute to run other companies' models. Based on available information, this will put SpaceX into a place where it has more liabilities than assets.
Lol
You're speculating what SpaceX balance sheet might be, like they haven't just released MRQ earnings report last week
Just go read it and you'll see you're wrong even after you subtract that 40 billion SPV https://www.sec.gov/Archives/edgar/data/1181412/000162828026...
Fair, I misread the filing.
Larger point stands. SpaceX is commanding a 62x sales multiple, whether or not you personally believe price:sales should be relevant. SpaceX's AI operation is smaller and growing slower than Anthropic's. It's reasonable to expect investors to benchmark Anthropic against SpaceX, it would be silly not to do so. A normal thing one would expect investors to do is assign a higher multiple to a company that grows faster.
But the question is not so much if they can IPO at those prices but rather if the valuations are sustainable over a couple of years. If they are priced to perfection there isn't much left for public investors besides huge risk that something might not work out 100% as expected.
I mean, yes. You are raising points, using Wall Street lingo, that are rightfully raised about every IPO, ever.
To be clear: I do not have a financial interest in this, and will not be buying the IPO. I very much am just describing the world as it is.
Larger point doesn't stand, price/sales is as useful of a metric like measuring speed as mass/distance. "Yes, my speed is 100kg/10 meters".
Some people will benchmark Anthropic vs. SpaceX, that doesn't make it a smart approach.
A normal thing is to consider risk, liabilities, cashflow, how diversified revenue streams are etc., and other things which you seem to be missing.
I'm not trying to dissuade you from buying it. I'm just saying that the reasons you give are wrong.
> Some people will benchmark Anthropic vs. SpaceX, that doesn't make it a smart approach.
Yeah, I would not pay 62x sales for SpaceX. And I don't buy IPOs. I'm not endorsing these values, I am saying benchmarking against peers is what investors commonly do.
> A normal thing is to consider risk, liabilities, cashflow, how diversified revenue streams are etc., and other things which you seem to be missing.
I'm not missing those things, I am not including them here because being "right" is rarely how markets value companies. (Also because the rest of that stuff is in the S-1.)
It's possible to build a "correct" model of a high-growth high-tech company that looks primarily at risk, liabilities, cashflow, and diversity of revenue streams. That model might put Anthropic at (say) a $100B valuation. That "correct" model is going to be less predictive than a model that starts from what similar firms are trading at in public markets. I personally am more interested in predictive power than correctness when evaluating stocks.
> I am saying benchmarking against peers is what investors commonly do.
You are benchmarking against an outlier stock with a cult following (Elon Musk) and diversified sources of revenue. Anthropic does not have either.
If it was as easy as comparing the hottest new IPO with Elon's companies and applying the same multiples, people wouldn't have lost so much money on electric car IPOs that pretended to be Tesla (see PSNY -99.5%; NKLA -99.99%; FSRNQ -99.99%; etc.).
> being "right" is rarely how markets value companies.
Conventional metrics are how most stocks are valued most of the time. What you want to say is that market can be exuberant sometimes. That's fair. When the exuberance goes away, valuations come back to earth (see PSNY -99.5%; NKLA -99.99%; FSRNQ -99.99%; etc.).
> That "correct" model is going to be less predictive than a model that starts from what similar firms are trading at in public markets.
Even if we take your statement at face value, you're still not doing it right. A valuation based on comparisons has to have more than 1 peer to compare against. Check out the multiples of other model providers listed on the stock market. You'll probably be shocked that it's not a double digit multiple of revenue.
And then, if you really want to have a proper model, isolate the AI revenues of SpaceX and other hyperscalers, then we can argue that you have a model worth something.
> You are benchmarking
Comparing != benchmarking. The only folks acting like anyone upthread is benchmarking to SpaceX are those who never made that comparison.
> You are benchmarking against an outlier stock with a cult following (Elon Musk) and diversified sources of revenue
So in this case, diversified revenue is a drag on SpaceX because all the non-AI revenue is growing considerably slower.
But your point about Musk is taken. Cerebras went out in May, and despite being a hardware company is trading at 37x sales. Palantir trades at 78x, CrowdStrike at 50x, Cloudflare at 47x, etc. Anthropic is growing significantly faster than any of those companies. I'm not making the case that Anthropic will/should trade at any of those, just that Anthropic trading in the range of the P/S of SpaceX would not be crazy, and it would not require Musk to be involved.
> A valuation based on comparisons has to have more than 1 peer to compare against
Yes, the best peers are "public pure play frontier model providers trading in the US." There are none. SpaceX is closest, despite not really being a frontier provider and definitely not a pure play. But you are correct that investors will use a basket of stocks. It is silly to suggest SpaceX will not be in that basket.
You absolutely can make an AI doomer basket to model Anthropic. Use "real infrastructure companies" as the connective tissue since AI wants to be infra. In your basket put Caterpillar, Kroger, The Southern Company, AT&T, Comcast, and ExxonMobil. You can absolutely make the argument that Anthropic should trade more like them than like CrowdStrike or Cloudflare, that is a coherent position. Then based on that infra peer group, calculate a valuation. I think this model will be much less accurate in the near term than a model that uses other high-growth tech companies.
Price discovery is part of the fun of IPOs! The doomer model could be right, the stock value could collapse toward irrelevance as fast as the exchange's circuit breakers will allow!
> So in this case, diversified revenue is a drag on SpaceX because all the non-AI revenue is growing considerably slower.
Your argument is like saying that having gearbox in a car is a drag on speed, why not remove it? Just take out the transmission and the windshield while you're at it too.
> Cerebras went out in May, and despite being a hardware company is trading at 37x sales. Palantir trades at 78x, CrowdStrike at 50x, Cloudflare at 47x
None of those companies are AI labs, you're just cherry picking random expensive stocks and doing the whole price/sales nonsense again. All of them have positive shareholder equity (Anthropic has shareholder deficit).
> But you are correct that investors will use a basket of stocks.
That's your approach, I was just quoting it.
> You absolutely can make an AI doomer basket to model Anthropic. Use "real infrastructure companies" as the connective tissue since AI wants to be infra. In your basket put Caterpillar, Kroger, The Southern Company, AT&T, Comcast, and ExxonMobil.
At this point you're either being silly or arguing in bad faith. Your best case scenario is for Anthropic to trade like a stock with cult-like following and your worst case scenario is for it to trade like a ketchup stock. Completely ignoring the most obvious tech companies that are traded publicly and also provide models of their own. This conversation became so absurd that it's not worth answering anymore.
The mental gymnastics you're doing here are appalling.
If your hyper growth startup does not have more liabilities than assets, it has fucked up.
Hahaha
Someone on a different forum fed the S-1 in Fable and it said "Shows a massive “loss” on paper, but that’s mostly an accounting effect from the company’s value going up"
Funniest shit I've read all day
>Modern IPO's just reek of investors cashing out
99.9999% of the time someone sells something, the goal is to obtain cash. Throughout all of human history.
Did they get it right? I'm geniunly curious because from my point of view the only company to get AI right so far is Microsoft. This is from the point of view of an EU energy company married to Microsoft services mind you, but Microsoft is the only AI company which actually sells AI to us in a manner that can be considered useful. 7 months ago I thought Copilot (the m365 one, not the github one) was useless. Now I think the chat part is as good as what you'll get on the smaller subscriptions to ChatGPT or Claude, and it's Cowork is very good.
Obviously it's not like having Claude Code or similar, but those would be very hard to run legally and safely in EU regulated industries like us in energy production. You could run Codex but it would require your IT department to have yet another place to manage security and software integrations. Which is a fairly big issue when you have around 1 IT staff member per 300 employees, and an organisation which will buy five different sharepoint solutions for the same problem.
Microsoft sells something that seamlessly integrates into everything else you buy from them. It's easy to manage credit spending and when I say that I don't mean that it's easy to set it up. It's easy to track and budget so the right departments pay for it. It can read your emails, your onedrive, your loop, your sharepoint, your powerbi/fabric and anything else you have access to. Sure it's annoying to be confronted with a point you've made 7 months ago (like how you think Copilot is utter trash) and since changed opinions on it's frightingly good at finding relevant stuff, even in some obscoure legal document on your intranet.
I assume that Anthropic and OpenAI makes money of this somehow, but for the most parts our users wouldn't care if it was the GPT and Claude models or Mistral or Deepseek. Which is why I asked if they had really gotten it right.
What fraction of coding opex do they need to capture to justify this, whether they got it right or not?