저자: CurranFrancis · 외국인
요약 보기
WSJ 보도를 인용해 AI 데이터센터 관련 부외부채(약 3조달러 추정)와 차입 기반 투자 구조의 위험을 지적. 마이클 버리가 SOXX·QQQ 등에 숏 포지션을 보유하며 AI 조정을 예상한다고 소개. 금리 인상 전망과 마진론 급증으로 2027년 전후 압박이 커질 수 있다고 진단. 개인투자자에게는 공매도 대신 마진·부채 축소, 자산배분 점검, 패닉 매도 회피를 권고.
다듬은 글
So, this report just came out from the Wall Street Journal last month, showing AI spending on chips and data centers is far higher than investors have been told. It points to $3 trillion in hidden spending being fueled by unsustainable debt, which calls Nvidia's $5 trillion rise in recent years, or Palunteer's $2500% rise in the last 3 years into question.
If these valuations have been built on faulty data that doesn't tell the whole story, what happens when these bills come due? But investor Michael Bur has gone one step further, claiming AI is a bubble that could bring down the rest of the economy when it implodes. And Bur isn't just talking here. He is actively positioning his portfolio for a major market downturn in AI.
Explaining his trades in a recent post, saying, "If this looks like preparation for a larger fall in the market, that is because it is." So, Michael Bur became famous for predicting the housing bubble before the whole market collapsed in 2008, profiting over $700 million for his investors and over $100 million personally off that one trade.
He's also made big profits shorting other large tech runs, including a decent profit shorting Tesla recently. But he's also made several negative predictions that never actually came true. This chart lists a bunch of them. But after stepping away from the public markets, deleting his Twitter completely, Bur has now come back with a private Substack where he is showing everything that he is buying and selling and sharing some charts that show us just how deep the AI debt worries may run because even if you don't trust Bur's conclusions, the data that he is showing is worth understanding for any investor.
So, let's just start by what is he actually betting against. So in August, Bur shared that he had added to his short positions against the big AI players, Micron, Nbius, Nvidia, Oracle, and Palanteer. But his biggest bet right now is against SOXX, a semiconductor ETF. So instead of betting against a single chip company, he is betting against most of the chip industry all at once, an industry that is up over 100% in the past year.
On top of this, he holds a large negative bet against QQQ, an ETF of the 100 biggest companies in the NASDAQ, which today is dominated by big tech and AI companies. Now, we don't know the exact size of those trades. Bur no longer reports to the SEC after he closed his fund at the end of last year. So, these estimates aren't perfect, but they do show us that Bur is betting against several steps in the AI supply chain.
Now, in addition to these shorts, Bur does also hold several long positions, which I covered in a separate video where I looked at what some of the biggest investors were putting money into right now. But for this video, I want to laser focus on why Bur is betting so heavily against AI and specifically semiconductor stocks.
In recent months, many people have called Bur a broken clock who is right twice a day. He claims he's usually just early, but I want to look at the data and the three biggest reasons behind his bets so we don't just have to trust Bur we can decide for ourselves. And we'll also look at what investors can do to not just survive, but actually profit from what is happening right now in the AI market.
So, this all kicked off last month when the Wall Street Journal published this deep dive into the massive amount of spending that AI companies have been pouring into hardware and data centers. Now, we already knew that around $600 billion will be spent in 2026 alone. But this investigation found that there may be a lot more money committed to these data centers than meets the eye.
Money that isn't factored into these stocks earnings and thus their stock prices. And the key here is what's known as offbalance sheet liabilities, where companies have promised to pay money in the future and they may even have a contract, but they haven't actually spent the money or listed it as debt on their balance sheet.
Now, to be clear, this isn't some huge secret. I mean, it's listed right here. This is a footnote from Google's latest 10Q report. So, it's there, but they don't make it obvious. Or take Meta for example. They're now building their Hyperion data center in Louisiana. But neither Hyperion nor the $27 billion in debt shows up on Meta's balance sheet.
Instead, that debt is owned by Blue Owl Capital, which Meta partners with, slash is also a tenant. But even that is not a true secret. It's not like they're fraudulently hiding these loans, but it also does not show up on third party websites like Google Finance or in most investing apps. So, if you look up their leases or their debt ratios, these numbers are not factored in, which is a big deal because some of these are like 25-year lease commitments, which even if there's no revenue coming in for AI in the future, they are guaranteed to have to make those payments going forward.
And Bur has a very strong opinion on all this, saying, "The fish have gorged themselves on off-balance sheet liabilities, backstocks, unccommenced leases, purchase agreements, all ways to spend future money without having to list it as a liability. And it's $3 trillion in future promises. So more than is actually listed on the balance sheets.
And all that spending does not show up in the balance sheet or income statements for these companies. Now, to be fair, this isn't money that's immediately leaving the company. Alphabet had $811 billion in future promises, but that includes things like 25-year leases on data centers, which obviously isn't paid all at once.
But there are other cases that look more concerning. Nvidia just committed $15 billion to back financing for Open AI. Kind of like a parent co-signing a kid's first car loan. Only this backing is worth three times more than Nvidia's entire publicly disclosed debt levels today. So, you better hope it's apparent that OpenAI doesn't crash their car because Nvidia's investors are going to be on the hook for that loan either way.
And this is where we get this famous chart where it feels like AI is a circular economy that's just propping itself up with loans and borrowing and deals. But that takes us to Bur's second big worry about the AI market. And this is more useful to investors because it reveals the areas of the market that are the most vulnerable to what is happening.
the market is not all equal. And after that, we'll look at why Bur is worried about this spreading into a broader market downturn and what we can do about it as investors. One thing you'll notice about Michael Bur is he focuses on using his money effectively. For me, that usually means finding smart ways to make my income go further, like using tax advantaged accounts like an FSA, a flexible spending account for tax-free healthcare dollars.
And this video sponsor, FSA Store®, helps you get the most out of your funds. FSA Store® is a curated store to buy products that qualify for FSA spending, over 2500 of them. So, if your income tax is 30%, that's like saving 30% on eligible expenses. And we're getting into the end of the year, so if your FSA dollars might expire, you might as well use them instead of letting those pre-tax dollars go to waste.
But what's great is FSA Store® isn't limited to medications. They've got sunscreen, allergy meds, even pain relieving foot massagers. FSA Store® sent me this Caring Mill™ by Aura Migraine Relief Mask. It's pretty great after staring at a screen for hours or during the afternoon slump. It's like a 15-minute powernet plus massage, heat, and vibration.
It even plays music. But the best part, it's FSA eligible. So, you can buy this and everything else on the site using your FSA card. No submitting receipts. Just go to fsastore.com/fintech or scan the QR code. So, if you have FSA dollars, they're taxfree and they could expire at the end of the year.
So, it's worth checking out their collection of products. Visit fsastore.com/fintech or scan the QR code to start shopping. So, thanks to FSA Store®, where everything is eligible for sponsoring this video. Smarter saving is smarter investing. And now, let's look at the second reason for Michael Bur's bet against the AI market.
So, you have all of these offbalance sheet liabilities, but those by themselves do not cause an AI meltdown, but they do introduce fewer points of failure in the market. So Bur's other big fear is how much actual debt is being fed into a system that is already less stable than it should be. Because it's one thing if you have $3 trillion promised on future profits, but today much of big tech's cash flow has already disappeared into hardware costs they can't afford to spend anymore.
Meaning that $3 trillion time bomb has to be funded by debt. In August 2026, Nvidia announced a partnership with six investment and private equity firms with the goal of raising more than $500 billion for AI buildings and equipment. That is a target, not money sitting in a bank account today. But the goal is to get outside investors to start paying and writing loans directly for this buildout.
So this now adds an extra layer to that chart. Investors and lenders can now give money to a data center project. that project can then buy Nvidia systems. Then Open AAI or another AI company can promise to rent the computer power. So now Nvidia can record a chip sale today even though the final money to pay for the whole system must come from years of future rent.
And we can see how this whole situation looks a little dangerous just in the latest Nvidia OpenAI deal as an example. So let's break this deal down. We have SB Energy which is building a data center campus in Ohio. Open AI has agreed to rent space and power there. Pretty normal business things. So, lenders are willing to give money to SB Energy to fund the construction of the buildings because they expect those rent payments to come in.
But Nvidia has now had to step in and promise to cover some of OpenAI's rent and power payments if OpenAI doesn't pay up to that $ 105 billion number across the nine different sites they're building. And in return, OpenAI promises to only use Nvidia's chips in those data centers. Then on top of all of that, every company in this deal seems to own a piece of every other company with the investment firm Soft Bank owning both SB Energy and a big chunk of Open AI.
That is the circle that Bur is worried about. Money flowing through a system is good. That's just how an economy works. But in this case, it's a series of contracts all built onto each other, like stacking blocks higher and higher. And ultimately, the only money flowing into this system is going to have to come from OpenAI or the end AI users.
But right now, OpenAI doesn't make a profit, and they won't anytime soon. So, they need investors and lenders to pay for all this. And that's basically the bet to keep this whole thing building higher and higher long enough that they can become profitable because if they don't, there's nothing holding it up.
And we've seen even some traditional banks make some interesting moves here. The banks that loaned money for an Oracle data center, for example, are actually selling off those loans to private credit, an area of the market that already looks vulnerable to any outside shocks. And Michael Bur summarizes all this by saying that the increasing capital that all these companies are spending is coming from debt.
And this puts the AI boom on a clock with a countdown because every loan has interest and every loan must be paid back eventually. It's not that these companies are putting profits into this hardware, they're borrowing for it. And Bur thinks this countdown timer will hit sometime in 2027 and certainly before 2028, but recent moves from the Federal Reserve may move that clock even earlier.
Barkclay's Bank now expects the Fed to raise interest rates for the next two quarters, meaning all these companies are going to be paying even more money in interest, and by 2027, we're going to see increasing pressure on the hundreds of billions of dollars in debt that is funding this buildout. Now, this pressure won't hit everywhere all at once.
The AI market is not one uniform thing. So, let's look at the parts of the market Bur is most worried about, as well as the specific warning signs that investors can watch out for. So, Bur's biggest bets are focused on the hardware players in the AI market, especially the NeoClouds like Coreweave and Nebius, each of which are already paying hundreds of millions of dollars in interest for the debt that they've taken on to grow, which has led Bur to say about Nbius that this is what the top of a boom looks like.
Now, to be fair, these are also just aggressively growing companies, and taking on more debt can help you grow faster. They also have real revenue and real paying customers. But if there is an AI slowdown or an AI credit crunch, who's going to have more leverage to negotiate contracts and loans, Google and Nvidia or these smaller Neocloud providers?
The issue here is that these smaller players need two things to be true. One, they need their customers to continue signing more and more contracts. and two, they need their lenders to continue offering more and more debt to fund it. So far, that hasn't been an issue. Both of these pillars look pretty strong, but they do point to three warning signs that we can look for.
So, even if we don't trust Bur's total market crash prediction, we have smaller pieces of data that we can look at and act on. So, number one, if customers don't use these companies compute, if they have excess capacity, then the growth story here might be over. Second warning sign, if they can't find new lenders for their loans.
And third, if we see a rapid rise in interest rates that makes their loans too expensive for these companies to cover. Any of these could be a reason to sell out of these stocks because they have to balance supply and demand with the rates from lenders and keeping those in balance is a tightroppe walk.
But there's a third reason Bur is worried about the market and it goes beyond semiconductor stocks. It's the risk that any major fall in AI could spread to the rest of the market and cause chaos to use Bur's word. And after that, we'll look at three specific things that investors can do to protect ourselves if Bur's prediction comes true.
Right now, AI stocks make up nearly onethird of the S&P 500, the 500 biggest companies in the US. So, a major AI crash is already going to hit retirement counts of most long-term buy and hold investors who are holding index funds with Bur even warning back in 2022 that most of the market's money is just passively invested.
Meaning, as all the money flows into the biggest stocks in the market, that props up any bubbles that appear. But there's another concern which is how much of the market is propped up on investors borrowing margin debt where investors borrow money specifically to invest in a market that seemingly keeps going up.
This is a chart of all the customer margin debt in the United States. It shows debt reaching a record $1.5 trillion in June 2026 and currently sitting around 38% higher than last year. And for retail investors like you and me, that rate of debt growth is way higher. Robin Hood reported its margin loans crossed $21 billion in their second quarter, up 127% year-over-year.
Now, we don't know what stocks all that debt flowed into, but it's a reasonable guess that some of it has flowed into the AI market. So, if AI were to dip substantially, that is more investors who could be forced to sell into the dip to cover their margin loans, potentially driving the market even lower in a feedback loop.
And as I covered in my video on what causes stock market crashes, those feedback loops downward are what defines a true crash. But this channel is not about panic, it's about preparation. So, what can we do to prepare ourselves if Bur is right about what's happening? So, first off, I personally would never short a stock because even if you correctly guess that a stock will decline if your timing isn't perfect, you can still lose money.
Bur himself almost went bankrupt when it took 3 years from when he predicted the housing crash in 2005 to when it actually collapsed in 2008 with his investors revoling and trying to pull their money out early. So, instead of shorting the market, I would do three things. First off, if you have a large amount of margin in your investing account or even high debt levels like a credit card while also trading stocks, now could be a good time to pay that down.
It protects your downside if there's a major dip before 2028 like Michael Bur predicts. Number two is to check your allocations. If you own an index fund and you invest in individual stocks, let's say you own Google, you might be overexposed to some of the big tech companies because these big companies are so overweighted in these indexes.
buying an international index or even a small cap value index could help even things out. And these allocations really matter because we know there are still deals out there. Even Michael Bur is buying stocks. But certain areas like the NEO clouds, we now know what warning signs to watch for. And the most important is number three, which is to avoid panic selling.
Even if you're close to retirement, there are calculators for how much you should have in stocks versus bonds to weather a market downturn. For everyone else, study after study shows that staying invested through the downturn tends to outperform trying to time the market. And over a 20-year period, the market has never had a negative return.
But the most important thing is to have a plan ahead of time and stick with it. Stay invested. I'm Curran Francis, and if you found anything valuable, just consider subscribing to the channel. And here's a video of what five of the biggest investors in the market are buying right now, including a deeper look into what Michael Bur is actually buying
Comments 0