Big short Burry warns that U.S. stocks will repeat the "2000" bear market: capital expenditure is nearing its peak, and two years is enough for the AI bubble to burst!

Recently, Michael Burry, known for accurately predicting the 2008 subprime mortgage crisis, was interviewed on a podcast.

In the conversation, Michael Burry expressed an extremely pessimistic outlook on the current U.S. stock market, believing that a long-term bear market similar to that of 2000 may be on the horizon in the coming years. He judged that due to the dominance of passive investing (index funds) in the market (over 50%), future market declines will be "a unified drop," making it difficult to protect oneself with long positions in the U.S.

Burry compared the current AI investment frenzy to the "data transmission bubble" of 2000, pointing out that both exhibit a lag between capital expenditure and stock market peaks. He bets that Palantir will significantly decline within two years, with the core reason for his bearish stance being its unreasonable valuation and unhealthy financial structure, particularly as the company has created multiple billionaires through high-cost equity incentives despite not generating substantial profits historically.

Burry believes that AI poses a lethal threat to Google's core cash flow source—its search business. The success of Google Search lies in its extremely low costs, while AI search is exceedingly expensive. He judges that most users can obtain the services they need through free tiers, and the proportion willing to pay for large models will be very small, with real profits likely existing only within the developer ecosystem.

Burry holds the sharpest criticism of the Federal Reserve, arguing that it has "not done anything truly beneficial" in its hundred years of existence and advocates for abolishing the Federal Reserve and transferring its functions to the Treasury Department.

Highlights from the interview:

• I believe the current state of the stock market is not good, and the next few years may be very bad, potentially leading to a long-term bear market similar to that of 2000.
• Right now, I think the entire market will drop together, making it difficult to protect oneself by holding long positions in the U.S.
• For example, Palantir is currently at $200 per share, while I believe it is worth only $30 or even lower, so I would buy two-year put options significantly out of the money at an exercise price of $50.
• "In other words, you are betting that Palantir will significantly decline within two years." "Yes, it will drop significantly over a relatively long period."
• If you look at Palantir this way, it has historically made almost no real money. My basic conclusion is: your company has such a high market value, but in reality, it hasn't made much—or has only made a little profit—yet has created a bunch of billionaires.
• Palantir and NVIDIA are the two luckiest companies on the planet; neither initially produced products specifically designed for AI.
• What will this bubble look like? It is very similar to the "internet bubble" of the past, but that bubble was not entirely an "internet bubble"; it was more like a "data transmission bubble."
• In every previous cycle like this, when the relevant sector's stock market peaked, capital expenditure often hadn't even completed half of its cycle
• Our current level of capital expenditure is comparable to several previous peaks: for example, it is similar to the level during the shale oil revolution relative to GDP, and it is also close to the level when the Nasdaq peaked during the internet era. … I think a two-year put option is sufficient.
• We have reached a state where as long as you announce an increase of $1 in capital expenditure on AI, your market value can increase by $3 as a result.
• But since I started using ChatGPT and Claude, I basically don’t use Google anymore.
• For the average person, how much additional value can AI bring? Not much. … The problem is that the proportion of people willing to pay for large models is very small, and they are likely to never need to pay because these services will be highly commoditized and priced very low.
• I believe the Federal Reserve hasn’t done anything truly beneficial; it might be the easiest job in the world. … I think we don’t need the Federal Reserve.

The following is a transcript of the interview, with some content omitted:

Michael Burry: The high proportion of passive investment is a new risk in the market, and I have closed my external funding pool

Host:

Let’s talk about this. Your current institutional size is very small, with only a few investors. What are the regulatory disclosure requirements for operations like yours? What can the outside world see about what you are doing?

Michael Burry:

What they can see are U.S. trades, specifically U.S. securities that are stock-related. They can see stock positions and a severely “distorted” options position.

Host:

Why do you say it’s “distorted”?

Michael Burry:

For example, if I buy 50,000 put options on Palantir, with each contract corresponding to 100 shares of the underlying, then I nominally “control” 5 million shares. For instance, if Palantir is currently $200 per share, and I believe it is only worth $30 or even lower, I would buy two-year put options significantly out of the money at an exercise price of $50.

Host:

In other words, you are betting that Palantir will drop significantly within two years.

Michael Burry:

Yes, it’s a bet on a significant drop over a relatively long period. One day I saw a CNBC report in the gym saying I had a $1 billion short position on Palantir. In reality, that position was only about $10 million.

Host:

I couldn’t believe it when I saw the report.

Michael Burry:

The reason they arrived at the “$1 billion” figure was by multiplying the number of shares corresponding to those options by the current stock price. In fact, my cost is less than 2% of the underlying market value, but the media calculated it as if “I held 100% of the underlying stock,” which is off by two orders of magnitude. **

The same situation can occur with index options. I use index options to hedge my portfolio, and then outsiders say, "Oh my gosh, he's shorting $1.5 billion of the S&P 500 index!" This creates such "sensational news." But that's just the notional amount, which does not represent the real risk exposure.

Interestingly, they don't calculate this way for others.

Host:

I previously asked you about the impact of "The Big Short" on your life, and this is one of them: compliance.

Michael Burry:

Since the financial crisis, the compliance environment has changed dramatically. We have a compliance officer in our company who constantly tells me: don't talk to anyone, don't respond to anyone. Since the movie was released, I have remained silent on external opinions, but inside, there has been a growing impulse—I really want to say something, but I can't.

During the COVID-19 pandemic, I had strong opinions on many issues, so I went on Twitter, but I was told I couldn't talk about stocks, only social issues unrelated to stocks. As you know, discussing social issues can "get you in trouble," so I left Twitter again.

I recently returned to Twitter because we canceled our investment advisor registration, and I no longer manage that external fund pool; I only plan to manage my own money.

Host:

So now basically only your own funds are in operation.

Michael Burry:

Yes, most of it is my own money.

Host:

Why did you make this decision?

Michael Burry:

I believe the current state of the stock market is not good, and the next few years could be very bad, possibly leading to a long-term bear market similar to 2000. But the market structure has changed.

Back then, it was mainly hedge funds, mutual funds, and separately managed accounts, with many managers actively managing funds and seriously researching stocks, investing from a long-term perspective. Although I didn't know I was on the autism spectrum at the time, I felt I had an "advantage" in such an environment: I could calmly look at issues outside the mainstream crowd's psychology, and it worked well.

Today, the market is dominated by passive funds, which account for a high proportion—over 50%, with numerous index funds. Some believe that the funds truly managed by active managers thinking long-term about individual stocks are less than 10%.

The problem is that in the U.S. market, if a decline occurs in the future, it won't be like 2000—when there were many overlooked stocks that could still perform well even if the Nasdaq collapsed. Now, I believe the entire market will decline together, making it difficult to protect oneself while holding long positions in the U.S.

This is why I decided to exit external management: Funds must structurally have a certain proportion of long positions, and I don't want to take investors through another cycle like this. Of course, after I closed the fund, I immediately rebuilt all these positions for myself.

Short Logic on Palantir

Host:

The same positions. In other words, you still hold these positions now.

I want to talk specifically about this portfolio. I noticed some things from a distance before, but you tell me: what key information am I missing? Someone—like CNBC—got your 13F. It shows that your position in Palantir is particularly large, looking very substantial, but that's actually because it's a put option, and it's significantly out of the money, with a strike price of $50.

In other words, you only disclosed the fund's holdings as required, and you didn't actively "promote" this position to the outside world, nor did you spread the word about Palantir. This is just a position that was made public due to regulatory requirements, right?

Michael Burry:

Correct.

Host:

Then I saw that Palantir's CEO Alex Karp began publicly criticizing you for buying put options on his company's stock. I recall that during the financial crisis, Morgan Stanley's CEO John Mack did something similar, blaming the company's stock price issues on short sellers, and regulators even briefly banned short selling on some stocks. In the U.S. market, once someone starts targeting short sellers, it often signals a very bad sign.

I thought to myself, "Wow, I definitely don't want to be in your position right now." You just made this trade, but it did pique my curiosity, and I want to understand your trade. Your bet is that within two years, Palantir's stock price will drop significantly. What do you think are the aspects of their business that the market hasn't understood?

Michael Burry:

My view is that this company originally had a set of application software with very high installation costs because customers, after buying the software, had to hire their consultants to complete the installation and learn how to use it. Palantir has a certain reputation in the government sector. Government contracts are difficult to manage, but I think they have figured out a way and secured some contracts.

Host:

How much of their revenue comes from government contracts now?

Michael Burry:

That portion has significantly decreased. It used to account for almost the majority, but now it's more balanced. Because in this wave of AI infrastructure, they have basically started pushing themselves towards enterprise clients—more accurately, enterprises are actively seeking them out. All the management teams of listed companies—board members and CEOs—feel they must "get some AI going," leading to a rush to hop on the AI bandwagon. Now they always emphasize that they are "the only company that can do this," but what IBM does is essentially similar. IBM's related business scale is actually larger than that of Palantir, and it does not primarily rely on government contracts. Government contracts typically do not have high profit margins. Internally, IBM is actually doing well in this business area, but it does not receive the same "premium" in valuation as Palantir, even though its growth is also rapid, roughly comparable to Palantir, or at least it used to be.

In other words: Palantir has produced about five billionaires, all of whom became wealthy by holding Palantir stock. At that time, the company's revenue was about $4 billion, which means that the ratio of "number of billionaires / company revenue (in billions)" is actually greater than 1. I have never seen such a situation before.

Host:

Is this the initial reason that attracted your attention to Palantir?

Michael Burry:

That was just an interesting detail. I thought to myself, "Wow, how did they create five billionaires at this scale?" In a company with $4 billion in revenue, stock-based compensation is almost equal to all their profits. They had to use a large amount of stock to pay the consulting employees, so they kept using equity incentives. Then the company would buy back those stocks. The company certainly hopes you give them "extra points" in your analysis, and Wall Street's usual practice is to look at earnings per share first, then add back stock-based compensation because it is considered a "non-cash expense," thus adding it back to profits.

But I believe that GAAP accounting's treatment of equity incentives actually underestimates their true cost. You can see how much a company spends to buy back stock to offset dilution and then deduct that amount from cash flow. If you look at Palantir this way, it has historically made almost no real money. My basic conclusion is: your company's market value is so high, but in reality, it hasn't made much—or only a little profit—yet it has already created a bunch of billionaires.

Comparison of Current AI Investment Frenzy and Historical Bubbles

Host:

In 2008, you had your own timing judgment on when the subprime mortgage bond market would start to collapse. Do you have a similar timing judgment regarding Palantir this time?

Michael Burry:

This is related to AI consulting. Palantir and NVIDIA are the two luckiest companies on the planet; neither of them initially produced products specifically designed for AI.

Host:

I know. But now they have become the two major "brands" of AI.

Michael Burry:

Yes. NVIDIA was originally a decent graphics chip company. I actually knew its CFO at the time; we talked in 2015 or 2016. I was bullish on the stock and even told her, "You are doing great, I really appreciate your buyback strategy." Her child was also on the same basketball team as my child About a year or two after I bought this stock, the price rose from 20 to 90, which, after accounting for the subsequent stock split, is equivalent to the current 0.4 USD. NVIDIA was very lucky. The first time was due to the demand for GPUs for cryptocurrency mining. The GPUs were not "customized" for mining; they just happened to be usable. Later, AI emerged, and the situation was almost identical.

About a year and a half to two years ago, Palantir was not yet an AI company. Basically, after ChatGPT came out, they put an "AI shell" on their existing applications, continued to sell related consulting services, and called it all AI. In fact, every company is doing this now.

Host:

What about the "timing" of AI itself? What is your judgment?

Michael Burry:

This brings up a question: what will this round of the bubble look like? It is very similar to the "internet bubble" of the past, but that round was not entirely an "internet bubble"; it was more like a "data transmission bubble." At that time, there was a large-scale deployment of fiber optics, which required routers, and routers needed more fiber optics, creating a mutual push until everything collapsed. The market peaked on March 10, 2000, but Cisco's revenue still grew by 55% that year and by 17% in 2001 because capital expenditures continued. This company peaked in the stock market, but it took about another year for the business side to truly peak.

You can look at "net investment," which is capital expenditure minus depreciation. By comparing this metric with nominal GDP, you can make comparisons across different periods. The result will show "mountains of investment frenzy." In every previous cycle like this, when the relevant sector's stock market peaked, capital expenditures often hadn't even completed half of their course. In most cases, capital expenditures themselves hadn't even peaked.

We are currently in a phase of accelerating capital expenditures. It has reached a point where if you announce an increase of 1 USD in capital expenditure on AI, your market value can increase by 3 USD as a result. We saw this with Oracle—a giant company that saw its stock price soar by 40% after announcing a massive investment plan worth hundreds of billions of dollars. They announced "booked orders," but to fulfill these orders, they actually have to spend the money and build things, and that process is still ongoing.

Host:

So where do you think we are now?

Michael Burry:

I can't give a precise answer because this round isn't over yet. Our current level of capital expenditure is comparable to several previous peaks: for example, it is similar to the level during the shale oil revolution relative to GDP and close to the level when the Nasdaq peaked during the internet era.

Host:

So you think a two-year put option is enough? Michael Burry:

I think two years should be enough. Yes, I believe two years is sufficient. If you really have to buy something now, I would say consider buying healthcare stocks, which are currently very out of favor. If you hold some stocks that have already risen a lot, performed very well, and are now almost on a "straight-up" trajectory, and you feel they are a bit overvalued, then I think those can be considered for selling.

Google's Search Business Faces AI Challenges

Host:

I would like to ask one more question about the stock market. You hold Berkshire Hathaway, and Berkshire recently announced—or it was disclosed—that it bought a large amount of Google stock. Were you disappointed by this news, or do you understand their logic?

Michael Burry:

First of all, we don't know for sure that this was personally bought by Buffett. Secondly, among that type of company, Google has always been a favorite of value investors.

Everyone says it is cheaper than several others and is relatively a "undervalued stock," and after all, it is Google. But since I started using ChatGPT and Claude, I basically don't use Google anymore. The magic of Google Search lies in its extremely low cost. Most searches cannot be monetized. About 85% of searches are not for buying things and are unrelated to products, such as, "What did Columbus actually do?" Such searches cannot be monetized at all, so they must ensure that they "don't lose too much money" on these queries. AI has changed that. AI is expensive. I often run some queries, and I know that each query might cost tens of dollars. Google used to be able to reduce the cost of a search to a fraction of a cent. That search business is their "golden goose," and it is basically the source of all their cash flow.

Regarding large models, there is one more point: looking back at the internet bubble, it was an incredible communication revolution. If you experienced the 1980s and then the early 2000s, those were two completely different worlds, and the internet greatly changed everyone's lives. But AOL disconnected its last dial-up user just last year or early this year. This means that in the U.S., the spread of the internet was actually very slow; in "single city states" like Singapore or high-density cities like Seoul, the spread was rapid, but in the U.S., by the time of the financial crisis, there were still many people who were not online or were still using dial-up.

As internet connectivity gradually improved, people had a lot of things they wanted to accomplish online: they wanted to shop online (which led to the growth of Amazon), wanted to socialize online, and so on. And now, for large language models, most people can already get what they want through the free tier, and the penetration rate is already very high. How much additional value can they bring to the average person? Not much. The real money will be in the developer ecosystem, where there is indeed a lot of money. But the problem is that the proportion of people willing to pay for large models is very small, and they may never need to pay because these services will be highly commoditized and priced very low

Unwilling to Bet that "America Will Definitely Find No Way Out" in the Short Term

Host:

Alright, I'll let you go in a moment, but I want to ask a few more questions and hear your thoughts. What could trigger a debt crisis in the United States? Are you particularly concerned about the fiscal situation of the U.S. government right now? How do you view these issues, and where do you think they will lead in the future?

Michael Burry:

Predicting such things is inherently difficult. I often use a metaphor: "Waiting for Castro to die is not a strategy." People can live a long time. The power of a nation is also strong, capable of doing many things. The U.S. has the status of a global reserve currency. Clearly, Trump is currently "running wild" around the world, but the U.S. is still an extremely important country. I am unwilling to bet that "America will definitely find no way out" in the short term.

However, I do find the current situation absurd. We collect about $4.5 trillion in personal taxes and about $400 billion in corporate taxes. Even if you double the corporate tax, it would only add another $400 billion; what difference would that make? Our annual interest expenditure is about $1 trillion. The interest burden is already high, plus all the welfare programs. We do not have a thick social safety net like many developed countries, but even so, we actually cannot afford to do more.

Host:

So, you believe that debt will continue to rise and rise, but you won't bet on "when it will collapse"?

Michael Burry:

No, because this is America.

Host:

What do you think about the independence of the Federal Reserve? Do you care?

Michael Burry:

I might have a somewhat "pathological" view on this matter: If Trump starts to directly manipulate the Federal Reserve, it could actually lead to the end of the Federal Reserve, because by then, not just me, everyone will hate it. We'll see. I believe that the Federal Reserve has done a lot of harmful things since its establishment in 1914; I think we do not need the Federal Reserve. We do not need it. Unless the Federal Reserve can clearly explain: why do they want to cut interest rates? There is no reason to cut rates now. Inflation is slightly rebounding, and the economy is barely "dragging along"; our neutral interest rate is not 1%, not 0%, and certainly not at the level Trump wants; our neutral interest rate is around 4%, or close to the current level.

You have to understand that once interest rates are cut, all groups relying on fixed income savings will be "wiped out." They have suffered for too long, and are just now starting to regain their rhythm of life. Cutting rates comes at a cost. When you say you want to cut rates, you better think carefully about what you are really asking for. You might find that by cutting rates, due to the terrible debt situation, the yield curve could become steeper.

Host:

Did you just say you want to abolish the Federal Reserve?

Michael Burry:

Yes. I believe the Federal Reserve has not done anything truly beneficial; it might be the easiest job in the world. Host:

So what do you want to replace it with?

Michael Burry:

I think the U.S. Treasury could definitely set up a department specifically to make these decisions. After all, the Federal Reserve is already monetizing the Treasury's debt; they are almost like one department now.

Host:

With your kind of "institutional pessimism," where do you hide your assets? Do you have gold?

Michael Burry:

I have held gold since 2005.