The stock market doesn't operate on long-term principles anymore. So, in some sense, it is immaterial that the rosy scenarios where AI is responsible for >$30 trillion in the next decade are unlikely. If you bet against the hype and it goes on for a few more months/years you lose as much or more than if you went along for the ride.
Burry is well aware of this, he has written about how passive investing is contributing to this problem.
> The stock market doesn't operate on long-term principles anymore.
It has been broken since ~2008 (ZIRPs, etc.) and has really gone off the rails since BTC and memestocks have taken off. Now everything's a memestock. It's all vibes-based.
The internet absolutely went bust. That’s what the dotcom crash was.
Same with a lot of physical infrastructure. The UK has a robust railroad network today, but it was built during a bubble that was so insane people would take loans from banks to invest in railroad stocks.
And the counterpoint is that META, GOOG, AMZN, MSFT are all betting their companies that AI is the next move. Just yesterday, GOOG lent another $80 billion to be invested in their AI hardware, and they're also investing their own stock in AI hardware, for a cumulative investment of already over $1 trillion. Clearly the tech sector thinks this is worth it.
And of course the people deciding in FANG companies actually have numbers, Michael Burry has the same numbers you have. So these investments are "worth it" according to people with inside information. What Burry is doing, in one perspective, is calling out the leadership of FANG companies. Now that's the job of a short-seller of course. But that's the bet being made.
It’s very risky to make these kind of claims. If the denominator (USD) gets obliterated they might very well be worth $1T. It’s all about liquidity and central banks have a whole bag of tricks at their disposal. They never want to see deflation shocks again and they prefer asset inflation.
I think you could argue Anthropic could be worth $1T. With AI becoming an essential work utility, every global knowledge worker would want a claude subscription. There are 650M to 1B of such office workers. 300M workers × $50/month × 12 = $180B/year. The genie is not going back into the bottle, I have seen what claude code can do when properly connected to tools.
But Micheals arguments are valid. There could be competition, or even local models, thus indeed becoming 'commoditized'.
What probability do you assign to that, especially since CC harness code leaked?
Because I used frontier models this weekend (I had 78% of my assigned tokens for this month left, I wanted to burn them before June 1st, ended up with 24% left), and tbh, I don't see much of the improvement compared to the models I use day-to-day. I'd rather pay less for a slightly worse model. Stacktrace analysis (or any bug analysis really) is where LLMs have the most success rate imho, and free models are good enough since last year. As for coding/architecture tasks, frontier models seems to hallucinate less, but I wonder if it's the guardrails or the he model themselves.
thing is, we have both local models and local hardware and a true evaluation would do a calculation before openai inflated thw market, before nvidia made circular deals and the other distortions.
i think youd find the ROI is nowhere near the API rates are the "price support" is entirely a figment of billionaires and their parasites trying to corner the market by horde logistics
Literally adjusting my pension funds and ETA's away from this ... I don't even care if I lose out. I want stable growth. Not a nation obsessed with memestocks.
The real question is how we are defining "worth." Much of the market has decoupled from traditional fundamental data, with Tesla's P/E of 380 illustrating this perfectly, but the Tesla stock price refuses to collapse.
We all know the market can stay irrational much longer than you can stay solvent if you bet against it. If you watched "The Long Short" (excellent movie btw.) you know how close Michael burry came to capitulation before his subprime bet paid off. He seems to have a tendency to be too early with his predictions, even with his genius GameStop investment. So while he may be right again fundamentally, his timing may be completely off and those companies could be "worth" significantly more than a trillion dollars, at least temporarily, in stock valuations.
My personal prediction is this: The hype will go on longer than people think, just like with the New Economy. There is this quote from market analyst Larry Wachtel in 1999 who said: "Everybody's happy, everybody's making money - something's wrong here"[1]. Ironically, even Wachtel eventually succumbed to FOMO, capitulated, went in late, and lost a lot of money[2]. I am trying to not make his mistake, but it will be tempting to do so, I am sure about that.
One of the hardest parts of investing during a major technological change is the paradox that a technology can be truly revolutionary, yet every company can individually be a bad investment.
The problem is that every company is valued by the market as if it could capture all the value of the new technology, but obviously, not all of them will. This is what happened during the first Web bubble. All the big names were bad bets. Everyone lost a lot of money. Many companies folded immediately. Others, like Yahoo, bled out eventually. Amazon took 10 years to recover. MSFT took 15. And the biggest winner, Google, didn't IPO until well after the bubble popped.
So the question isn't "is Anthropic worth $1T"? The question is "is AI worth Anthropic's $1T + OpenAI's $1T + SpaceX/xAI's $1.8T altogether?" Maybe, but probably not. That implies you are betting on a horse, not investing.
Which one will be the winner in 10 years? Are you confident making that pick? Is it even one of these companies or some other Google yet to be formed or will rise up from China or whatever?
That's the paradox. AI can still change the world and create a ton of value, yet none of these companies might be a good investment at these prices. Or one of them could be a great investment, and the others terrible investments.
He's been wrong a few times, but has been right far more than ONCE and appears to be profitably making decisions. For example:
Short dot.com stocks (55% return), short subprime mortgages (Massive return), long Gamestop 2019, short ARKK 2021, The shorts on Palantir and NVDA are probably still running (PLTR 25% in profit, NVDA 20% loss).
Its frustrating; you grind out a living making consistent solid double digit returns for investors and all anyone talks about is a guy who made 5bn for their investors 20 years ago and has lost 25bn for them since.
Neither Anthropic, Open AI nor SpaceX in its current form is a good candidate for an IPO at valuations that all but guarantee hundreds of billions of dollars in "passive capital" aka pension funds and ETFs will have to buy in.
SpaceX might have been a candidate on its own core business (aka: launching spacecraft and Starlink), but ever since the weird side deals with all of the other companies in the Muskverse (Twitter, xAI, Tesla) it is far too contaminated.
Sooner or later the AI bubble will burst - and assuming that the pension funds and ETFs buy in as projected, they stand to lose a lot of money that will make Covid's first lockdown + dotcom + 2007 Lehman combined look pale.
Elon must be looking to merge the new amalgamation of SPCX into TSLA and join all his sinking ships in one. Starlink itself won't be able to save that, but make it too big and the government might. That is the way to get to the ridiculous valuations needed for Elon's pay package.
AI doomers really are punching the air these days.
They will be right eventually and inevitably. Until then, it's funny watching them build a personal "brand" just to say "I told you so" when the market drops in X years.
He didn't say AI is doomed or a fluke, he said that these two AI companies aren't worth 1T in capitalization.
The same way semiconductor, internet or railroad companies were not great investments regardless of how important the technology was going to be. It's still a financial investment and it's only going to pay off if bought at the right price, not at crazy multiples.
I will also add: if all your moat is your latest model, you're as good as your latest model and can be easily dethroned.
Strong moats are monopoly-like concessions (Verisign), exclusive technological edge (ASML), brands (Coca Cola), etc.
Anthropic is the one company where that makes no sense. If revenue really is about 50 billion annually and growing than that means that a 20x 1 year of revenue valuation is modest compared to the shenanigans that have been going in the market. It's almost classically textbook conservative in comparison. The moat with these corporate enterprise contracts is literally your conversation history and companies aren't likely to jump ship when everyone likes the tools so long as costs stay nominal. AI at most orgs isn't even the biggest line item.
We know nothing about Anthropic, they make a few money go round deals, announce a bit of revenue, extrapolate it into the whole year and people parrot that they may be making $50bn. Most likely the cost to remain competitive eats at whatever revenue they could hope to make. I expect them to fudge the numbers the last quarter before their IPO, dump on passive investors, and then go back to being officially unprofitable.
What does revenue have to do with it? Mercedes has a revenue of $130 billion, profit of $5 billion and a market cap of $56 billion.
According to your logic, it should have a market cap of $2.6 trillion.
Conservative is to look at P/E, which is 10 for Mercedes.
Anthropic isn't even a growth stock, since it has already been force fed to everyone with one of the largest marketing and coercion campaigns in history.
I've had economics professors tell me that a "normal" business valuation is 10 years of profit so your example is in line with my thinking there. I'm just as curious to see the final numbers as you but if they are even approaching them it's not very out there to consider a high valuation. I don't wish to speculate on what the numbers actually are. I want to see them too.
All competitors need are their latest model to be either better or similar but much cheaper. And Anthropic has no less than 2 big competitors in the space in US alone providing similar quality models.
There's no moat in LLMs when you're as good as your latest model.
Companies out there aren't in the business of throwing money down the drain.
Take DS4, you can use Deepseek APIs directly with Claude Code, and you're unlikely to notice a difference for the overwhelming majority of your use cases. But your bills run in few $ per day. I'm talking 2 magnitudes less.
You forget the institutional inertia of how these things get negotiated from year to year. If something is working for people they tend to wanna keep it. The existing curation of how everything works is cheaper than rolling your own. Sure you can get something running yourself but integrations for a lot of people are worth some of this cost. Also for AI heavy customers (multimedia, video, etc) the sky is the limit and there's not enough processing for it right now.
> Strong moats are monopoly-like concessions (Verisign), exclusive technological edge (ASML), brands (Coca Cola), etc.
Agreed with the exception of Verisign. Many a "security" company went bust like DigiNotar after mishaps or hacks. Being a globally trusted root CA or DNS operator is a strong moat - but also an incredibly brittle one.
And brands... brands aren't as safe as we thought either, as "store brands"/"private labels" are taking up more and more market share [1].
The stock market doesn't operate on long-term principles anymore. So, in some sense, it is immaterial that the rosy scenarios where AI is responsible for >$30 trillion in the next decade are unlikely. If you bet against the hype and it goes on for a few more months/years you lose as much or more than if you went along for the ride.
Burry is well aware of this, he has written about how passive investing is contributing to this problem.
> The stock market doesn't operate on long-term principles anymore.
It has been broken since ~2008 (ZIRPs, etc.) and has really gone off the rails since BTC and memestocks have taken off. Now everything's a memestock. It's all vibes-based.
People have seen others make insane, life changing money at this slot machine and are wanting to take their chances.
There are no fundamentals.
There is very low signal to the noise.
>People have seen others make insane, life changing money at this slot machine and are wanting to take their chances.
Isn’t this the exact same sentiment from the late 1920s when people were making “insane, life changing” money by buying equities on margin?
(Agreeing with you) In the 1980s Gary Shilling said:
The market can remain irrational longer than you can remain solvent.
A long-term principle that I think does still apply.
The extent to which you’re exposed to long term principles is directly related to the time you’re in the market. Ie trader vs investor
“In the short run, the market is a voting machine, but in the long run, it is a weighing machine”
- benjamin graham
Can you elaborate on how passive investing contributes to this?
What's the rationale offered by NASDAQ for their new "15-days and you're in the index" rule for these massive IPOs?
Seems equivalent to removing road safety rules for the least-tested, most-powerful new vehicles only.
Of course not, this is pure speculation just like in past cycles, until it goes bum.
And this time the boooom has a larger blast radius than all the previous booooms combined.
The largest so far. We have always been in a boom/ bust cycle. The difference is that they are coming faster and faster.
I need to figure out the next one.
Minus, of course, all the times where it did not go bum, right?
Such as?
Do you live in a cave? I guess not.
Do you have access to the internet? Seemingly yes.
The booms just tend to be much bigger than the busts.
Great examples.
The internet absolutely went bust. That’s what the dotcom crash was.
Same with a lot of physical infrastructure. The UK has a robust railroad network today, but it was built during a bubble that was so insane people would take loans from banks to invest in railroad stocks.
Most people can't get past the kind of logic that declares Facebook doomed for 20 years because Myspace.
Beyond GenX and Boomers who is still using Facebook? We don't live forever.
The last time Facebook reported its monthly active user base they were at 3 billion. [1] So I guess the answer to your question is… everyone?
[1] https://www.sec.gov/Archives/edgar/data/1326801/000132680124...
But what's being discussed isn't usership: it's stock prices.
This is what they're actually talking about:
https://michaeljburry.substack.com/
And the counterpoint is that META, GOOG, AMZN, MSFT are all betting their companies that AI is the next move. Just yesterday, GOOG lent another $80 billion to be invested in their AI hardware, and they're also investing their own stock in AI hardware, for a cumulative investment of already over $1 trillion. Clearly the tech sector thinks this is worth it.
And of course the people deciding in FANG companies actually have numbers, Michael Burry has the same numbers you have. So these investments are "worth it" according to people with inside information. What Burry is doing, in one perspective, is calling out the leadership of FANG companies. Now that's the job of a short-seller of course. But that's the bet being made.
It’s very risky to make these kind of claims. If the denominator (USD) gets obliterated they might very well be worth $1T. It’s all about liquidity and central banks have a whole bag of tricks at their disposal. They never want to see deflation shocks again and they prefer asset inflation.
That seems like a hollow win.
Matt Levine made an interesting point that the SpaceX IPO looks like a short squeeze/pump and dump [1]:
1. Do an IPO.
2. Sell a small amount (5%) to price insensitive Elon Musk-fans at a $2 trillion valuation
3. Get in all the indexes, because you are huge.
4. Unlock more stock, which the index funds have to purchase
[1] https://www.bloomberg.com/opinion/newsletters/2026-06-01/the...
Please, don’t buy the stock. It’s a classic retail investor trap
The issue is that once it's in the S&P500, Russel, Nasdaq100 etc indices - everyone with a pension or tracker fund will be buying it without thinking
Yes. I am severely reducing my exposure to these indices finding alternatives.
My impressions is that US investors also have a special love for the S&P500 and could likely benefit from a non-US bias.
This is why direct indexing is important. I can simply exclude these tickers. Will it skew it slightly from the index? Sure but I’m ok with that
I think you could argue Anthropic could be worth $1T. With AI becoming an essential work utility, every global knowledge worker would want a claude subscription. There are 650M to 1B of such office workers. 300M workers × $50/month × 12 = $180B/year. The genie is not going back into the bottle, I have seen what claude code can do when properly connected to tools.
But Micheals arguments are valid. There could be competition, or even local models, thus indeed becoming 'commoditized'.
What probability do you assign to that, especially since CC harness code leaked?
Because I used frontier models this weekend (I had 78% of my assigned tokens for this month left, I wanted to burn them before June 1st, ended up with 24% left), and tbh, I don't see much of the improvement compared to the models I use day-to-day. I'd rather pay less for a slightly worse model. Stacktrace analysis (or any bug analysis really) is where LLMs have the most success rate imho, and free models are good enough since last year. As for coding/architecture tasks, frontier models seems to hallucinate less, but I wonder if it's the guardrails or the he model themselves.
Why would we need 1B office workers when Claude is supposed to fire everyone anyways?
thing is, we have both local models and local hardware and a true evaluation would do a calculation before openai inflated thw market, before nvidia made circular deals and the other distortions. i think youd find the ROI is nowhere near the API rates are the "price support" is entirely a figment of billionaires and their parasites trying to corner the market by horde logistics
Literally adjusting my pension funds and ETA's away from this ... I don't even care if I lose out. I want stable growth. Not a nation obsessed with memestocks.
Are you trying to be sane and rational in a market that has turned into a casino? U crazy or something?
It'll be a wonderful crow feast to anyone who agrees with Michael Burry on this.
The real question is how we are defining "worth." Much of the market has decoupled from traditional fundamental data, with Tesla's P/E of 380 illustrating this perfectly, but the Tesla stock price refuses to collapse.
We all know the market can stay irrational much longer than you can stay solvent if you bet against it. If you watched "The Long Short" (excellent movie btw.) you know how close Michael burry came to capitulation before his subprime bet paid off. He seems to have a tendency to be too early with his predictions, even with his genius GameStop investment. So while he may be right again fundamentally, his timing may be completely off and those companies could be "worth" significantly more than a trillion dollars, at least temporarily, in stock valuations.
My personal prediction is this: The hype will go on longer than people think, just like with the New Economy. There is this quote from market analyst Larry Wachtel in 1999 who said: "Everybody's happy, everybody's making money - something's wrong here"[1]. Ironically, even Wachtel eventually succumbed to FOMO, capitulated, went in late, and lost a lot of money[2]. I am trying to not make his mistake, but it will be tempting to do so, I am sure about that.
[1] https://youtu.be/uaK5tsH59UM?t=1188
[2] https://youtu.be/DSVPsP0Bfx0?t=456
One of the hardest parts of investing during a major technological change is the paradox that a technology can be truly revolutionary, yet every company can individually be a bad investment.
The problem is that every company is valued by the market as if it could capture all the value of the new technology, but obviously, not all of them will. This is what happened during the first Web bubble. All the big names were bad bets. Everyone lost a lot of money. Many companies folded immediately. Others, like Yahoo, bled out eventually. Amazon took 10 years to recover. MSFT took 15. And the biggest winner, Google, didn't IPO until well after the bubble popped.
So the question isn't "is Anthropic worth $1T"? The question is "is AI worth Anthropic's $1T + OpenAI's $1T + SpaceX/xAI's $1.8T altogether?" Maybe, but probably not. That implies you are betting on a horse, not investing.
Which one will be the winner in 10 years? Are you confident making that pick? Is it even one of these companies or some other Google yet to be formed or will rise up from China or whatever?
That's the paradox. AI can still change the world and create a ton of value, yet none of these companies might be a good investment at these prices. Or one of them could be a great investment, and the others terrible investments.
bro was right ONCE and now every tweet is a headline
He's been wrong a few times, but has been right far more than ONCE and appears to be profitably making decisions. For example:
Short dot.com stocks (55% return), short subprime mortgages (Massive return), long Gamestop 2019, short ARKK 2021, The shorts on Palantir and NVDA are probably still running (PLTR 25% in profit, NVDA 20% loss).
Its frustrating; you grind out a living making consistent solid double digit returns for investors and all anyone talks about is a guy who made 5bn for their investors 20 years ago and has lost 25bn for them since.
It was a hell of a once tho
He's still right this time.
Neither Anthropic, Open AI nor SpaceX in its current form is a good candidate for an IPO at valuations that all but guarantee hundreds of billions of dollars in "passive capital" aka pension funds and ETFs will have to buy in.
SpaceX might have been a candidate on its own core business (aka: launching spacecraft and Starlink), but ever since the weird side deals with all of the other companies in the Muskverse (Twitter, xAI, Tesla) it is far too contaminated.
Sooner or later the AI bubble will burst - and assuming that the pension funds and ETFs buy in as projected, they stand to lose a lot of money that will make Covid's first lockdown + dotcom + 2007 Lehman combined look pale.
Elon must be looking to merge the new amalgamation of SPCX into TSLA and join all his sinking ships in one. Starlink itself won't be able to save that, but make it too big and the government might. That is the way to get to the ridiculous valuations needed for Elon's pay package.
True, but the VC money will have been paid back, and what are we if not meat for the soulless machine of capitalism?
AI doomers really are punching the air these days.
They will be right eventually and inevitably. Until then, it's funny watching them build a personal "brand" just to say "I told you so" when the market drops in X years.
Your comment suggests that you (1) didn't read the article, and (2) have no idea who Michael Burry is.
He didn't say AI is doomed or a fluke, he said that these two AI companies aren't worth 1T in capitalization.
The same way semiconductor, internet or railroad companies were not great investments regardless of how important the technology was going to be. It's still a financial investment and it's only going to pay off if bought at the right price, not at crazy multiples.
I will also add: if all your moat is your latest model, you're as good as your latest model and can be easily dethroned.
Strong moats are monopoly-like concessions (Verisign), exclusive technological edge (ASML), brands (Coca Cola), etc.
Anthropic is the one company where that makes no sense. If revenue really is about 50 billion annually and growing than that means that a 20x 1 year of revenue valuation is modest compared to the shenanigans that have been going in the market. It's almost classically textbook conservative in comparison. The moat with these corporate enterprise contracts is literally your conversation history and companies aren't likely to jump ship when everyone likes the tools so long as costs stay nominal. AI at most orgs isn't even the biggest line item.
We know nothing about Anthropic, they make a few money go round deals, announce a bit of revenue, extrapolate it into the whole year and people parrot that they may be making $50bn. Most likely the cost to remain competitive eats at whatever revenue they could hope to make. I expect them to fudge the numbers the last quarter before their IPO, dump on passive investors, and then go back to being officially unprofitable.
What does revenue have to do with it? Mercedes has a revenue of $130 billion, profit of $5 billion and a market cap of $56 billion.
According to your logic, it should have a market cap of $2.6 trillion.
Conservative is to look at P/E, which is 10 for Mercedes.
Anthropic isn't even a growth stock, since it has already been force fed to everyone with one of the largest marketing and coercion campaigns in history.
I've had economics professors tell me that a "normal" business valuation is 10 years of profit so your example is in line with my thinking there. I'm just as curious to see the final numbers as you but if they are even approaching them it's not very out there to consider a high valuation. I don't wish to speculate on what the numbers actually are. I want to see them too.
All competitors need are their latest model to be either better or similar but much cheaper. And Anthropic has no less than 2 big competitors in the space in US alone providing similar quality models.
There's no moat in LLMs when you're as good as your latest model.
Companies out there aren't in the business of throwing money down the drain.
Take DS4, you can use Deepseek APIs directly with Claude Code, and you're unlikely to notice a difference for the overwhelming majority of your use cases. But your bills run in few $ per day. I'm talking 2 magnitudes less.
You forget the institutional inertia of how these things get negotiated from year to year. If something is working for people they tend to wanna keep it. The existing curation of how everything works is cheaper than rolling your own. Sure you can get something running yourself but integrations for a lot of people are worth some of this cost. Also for AI heavy customers (multimedia, video, etc) the sky is the limit and there's not enough processing for it right now.
That logic works for 20 vs 40$ SaaS subscriptions, not for burning triple/four digits per day in APIs.
> brands (Coca Cola), etc.
Paul Graham doesn't think so
https://paulgraham.com/brandage.html
> Strong moats are monopoly-like concessions (Verisign), exclusive technological edge (ASML), brands (Coca Cola), etc.
Agreed with the exception of Verisign. Many a "security" company went bust like DigiNotar after mishaps or hacks. Being a globally trusted root CA or DNS operator is a strong moat - but also an incredibly brittle one.
And brands... brands aren't as safe as we thought either, as "store brands"/"private labels" are taking up more and more market share [1].
[1] https://www.nbcnews.com/business/consumer/shoppers-are-tradi...
So the company that actually makes computers programming computers work, ain't worth $1T.
Whoever you are Michael Burry, you don't know shit about the implications, and where this is headed, and that the party only just got started.
I sure do wish I had a big chunk of that overpriced Google IPO stock, and Amazon, and MS, and Apple, etc etc etc
makes computers work. makes computer programming work.
please go back to school
> the company that actually makes computers programming computers work
LOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOL