Transcription
Somewhere in Louisiana, there exists a company called Biget Investor LLC, a pretty name that smells of powdered sugar and café au lait. Been has no employees, no activity, no address worthy of the name, just a line in a Delaware registry and a tax number. Yet, last October, it took out the largest private loan in the history of modern finance. 27 billion dollars, a very real loan. Guaranteed by concrete walls and servers equipped with Nvidia. This money was used to build a data center somewhere in the Bayou that Bennet quietly rents back to the one who really ordered it, MTA. And Benet is just one among dozens of similar structures all adorned with such innocuous names. Put together, they have moved over 120 billion dollars in debt off the official balance sheets of the world's largest tech companies. At the end of the chain, these billions are yours. Your life insurance, your retirement savings, the euro fund that your advisor sold you as the wisest investment ever created. While debt is hidden behind rounded logos and screen company colors, Wall Street's most seasoned investors are silently leaving the table. Peter Seal has liquidated his entire position in Nvidia. Soft Bank sold nearly 6 billion dollars worth just below the peaks. Michael Burry, the very same one who bet against subprimes in 2008, closed his fund after betting against Nvidia and repeats that we are reliving the last months of 1999. Finally, Warren Buffett is letting nearly 400 billion dollars in liquidity sit idle. An absolute record in the entire history of Berkshire Hathaway. 26 years after the dot-com bubble, the same warning lights are flashing on the same dashboard. Except this time, the debt is well hidden. And now, you're probably thinking, "Very well, but what do I do?" Because it's one thing to see Seal, Delio, and Buffet cover themselves all at the same time, it's another to know how to do the same. They think on the scale of major economic cycles. This way of investing is Global Macro, a multi-asset and multi-geographic approach that the wealthiest have been using for decades to navigate exactly the moments we are living through right now. And it shows. +4.7% in 2008 when the market went up in smoke, barely -1.6% in 2020, +8.98% in 2022 while global stock markets plummeted. The problem is that this method was reserved for a very exclusive club. A seven-figure entry ticket, analysis terminals at €24,000 per year, not for you or me. It is this barrier that Jordan Alouche, 15 years at the heart of the financial elite, is breaking down with Macro Insider. All the macro analysis translated into three turnkey ETF portfolios that you can simply draw inspiration from. No stock picking, no screens to watch all day, one hour per month, and your wealth withstands crises instead of suffering them. So, a small warning, we are closing registrations on July 12th. Until then, -35% off the annual plan and a 15-day money-back guarantee. The link is in the description. This year, Amazon, Microsoft, Google, and Meta will collectively spend about 725 billion dollars, largely on data centers. The Bank for International Settlements has calculated that the five largest AI players will consume over 1 trillion dollars between 2025 and 2026 alone. To give you an idea of the scale, IT investment in the United States now accounts for around 5% of gross domestic product, meaning everything the country produced in a year. This is already more than the peak of the dot-com bubble in 2000. No treasury can keep up with such a pace. According to Goldman Sachs, these expenses consume almost all of the cloud giants' operating cash flow. They are putting almost everything they earn into AI. Usually, it's the end customer who pays for the entire chain. In AI, this customer doesn't pay yet; the cloud advances the money. So, when cash is no longer enough, they borrow. And the strangest thing is that not a single dollar of this debt appears on their official accounts. So, how do they do it? The answer is an old Wall Street trick: the shell company. The technical term is special purpose vehicle, an empty shell with a single mission. When a tech giant wants a 30 billion dollar data center, putting this loan on its books would cause the stock to plummet. So, they set up a shell. Wall Street lends to the shell. The shell builds the data center, then leases it to the tech giant. The company gets its computing power. The lenders collect their rent, and since it's the shell's name on the loan, the debt never touches the company's accounts. Investors don't see a 30 billion dollar loan; they see a small, clean monthly rent. None of this is illegal. Everything is declared and validated by auditors. The debate is about transparency and risk, not fraud. This is precisely what makes it dangerous because a risk buried in a footnote is a risk that almost no one assesses until it's too late. The BIS has a name for this: phantom borrowing. And it added a sentence that should be displayed in all trading rooms: Leverage does not disappear because you hide it. Which brings us back to Beignet. The private credit fund Blue owns 80% of the shell. Meta retains 20%. And the debt came from Pimco. Someone registered the largest loan in history under the name of a fried pastry. And this ridiculous name is a gift because once you know it, you can track the debt yourself. Oracle played the same game on a larger scale, around 38 billion for its AI sites. Musk raised about 20 billion through a shell that buys Nvidia chips and then leases them to XAI. And by the way, Nvidia itself invests in this shell that is used to buy its own chips. Between mid-2024 and the end of 2025, over a hundred billion dollars in expenses have left balance sheets to be housed in these structures. And even then, 120 billion is just the debt already moved. In February, the agency Moody's released a staggering figure. The five giants have 662 billion dollars in lease commitments that haven't even started yet and are entirely off-balance sheet. This is more than their total adjusted official debt. Off-balance sheet debt has ended up weighing more than the balance sheet itself. The entire system relies on two conditions: that the tenants pay their rent and that the chips retain their value. In mid-December, the world's most sophisticated lenders saw the first one begin to crack. So, what did they see? Oracle released its results, and everything seemed normal until analysts opened the books. The company had significantly exceeded its announced quarterly spending budget. Its data center bills were outpacing its cash flow. The credit market reacted within days. The cost to insure Oracle's debt, what's called a credit default swap, in plain terms, an insurance contract against default, soared to levels not seen since the 2008 crisis. The stock market, however, shrugged. Retail investors continued to buy Oracle near its highs, while bond trading floors were already pricing in a catastrophe. On paper, the agencies still rate Oracle as investment grade. In the market, these bonds were trading like junk debt. For anyone who lived through the last crisis, the pattern is familiar, and it's not just anonymous accounts saying so. The Federal Reserve is polling markets on what worries them. In spring 2025, 9% cited a reversal in AI. In the fall, it was 30%. In spring 2026, half. The ECB has warned that parallels with the early 2000s fuel fears of a bubble. The IMF has calculated that the concentration of the US market exceeds that of the dot-com bubble. In January, senators led by Elizabeth Warren wrote to the Treasury demanding an investigation into what they call an AI debt bubble, with over a trillion dollars heading towards these infrastructures. When the BIS, the Fed, the ECB, the IMF, and the Senate point to the same risk at the same time, it generally means something is wrong. AI money is going in circles. Nvidia has announced its intention to invest up to 100 billion dollars in OpenAI. OpenAI uses the money it raises to rent computing power from Microsoft, Oracle, and CoreWeave, who in turn buy Nvidia chips. And remember, Nvidia also invests directly in the shells that buy its chips. The same dollars go around the table, counted as revenue at each stop. The IMF has finally stated it: these companies are simultaneously customers, investors, and financiers of each other, and they can inflate their revenues disconnected from fundamentals. The difference with the dot-com bubble is that Nvidia's customers have real, rapidly growing revenues. This circularity doesn't necessarily create fake revenue. It concentrates risk in the same place. And at the center of the wheel is OpenAI. About 13 billion dollars in revenue last year and losses far exceeding its earnings. It burns cash with every dollar that comes in and has already signed for something like 1,400 billion dollars in long-term computing commitments. Will this gap close or remain wide open? That's the question for every credit analyst on this file. And then there's the collateral, what the lender seizes when you stop paying. Here, it's Nvidia chips. No one agrees on their lifespan. Some say 6 years, others two. And in the same quarter, Amazon shortened the lifespan it attributes to its chips, while Meta lengthened its own for the same hardware. This game of lifespan has a very concrete reason. The day these server farms are fully depreciated, profits will take a hit. Goldman already expects the profit growth of these giants to slow next year to its weakest pace since 2022. The cost of data centers is starting to show up in the results. In the rental market, the price of a high-end chip has fallen by about 70% since 2023. These chips don't turn into pebbles overnight. But if a borrower defaults and the lender repossesses the machines in a hurry, they are worth a fraction of what they cost. And in MTA's footnotes, there's a small promise. If the data center is worth less than expected at the end of the lease, MTA will cover the difference up to 28 billion dollars. 28 billion in exposure signed and not recorded anywhere on the main balance sheet. So, the question is no longer whether some of this will unravel, but who will pay when it does? The peak was on May 14th. Nvidia hit its all-time record, over 5 trillion dollars in market capitalization. On June 4th, the market wakes up. Broadcom releases its results and does not raise its AI chip forecasts. Nothing catastrophic on the surface, except that the market no longer forgives the slightest hesitation. The stock falls 13% and drags the entire sector down. The Nasdaq loses over 4%. Its worst session in over a year. And in the aftermath, a stronger-than-expected US jobs report sends rates higher. The market that rewarded AI spending is starting to demand proof of return, and it's not getting it. On June 12th, the Federal Reserve, now led by Kevin WH, takes a harder turn, hinting that an interest rate hike is back on the table and revising inflation upwards, partly due to oil and the conflict with Iran. For a market valued for perfection, every tenth of a point in interest rates is less oxygen. And yet, euphoria has a final surge, and it comes from Korea. On June 18th, the Seoul index crosses 9,000 points for the first time, driven by SK Hynix and its AI memory chips. On June 22nd, SK Hynix even surpasses Samsung, ending the Korean giant's 26-year reign. Together, Samsung and SK Hynix represent nearly half of the Seoul stock market. When you buy the Korean index, you are buying a concentrated bet on two chip manufacturers. It's the same trap as in the US market, where the 10 largest companies now account for around 40% of the S&P 500, compared to barely 25% at the peak of the dot-com bubble. On June 23rd, the trap closes. The simmering fear erupts in Seoul, amplified by the tougher Fed and by forced deleveraging on the two chip manufacturers. The Korean stock market collapses by nearly 10% in one session. The circuit breaker is triggered. Samsung and SK Hynix fall 12% each. Foreign investors dump nearly 4 billion dollars in the session. The fear of the bubble, long simmering in bond trading rooms, has just burst into the open. Let's return to those who left the table before everyone else. Thiel sold his Nvidia position to shift to Microsoft and Apple. Softbank sold all its Nvidia. But to reinvest the money in OpenAI, which says a lot about the circularity. Neither of them is truly betting against AI. They just refuse to hold the leader at a premium price. Michael Burry goes further. He closed his fund after revealing 1.1 billion dollars in notional value of bearish bets against Nvidia and Palantir. Note that this notional value is not his actual stake, which is likely in the millions. But the message is clear. He believes we are reliving the last months of 1999. And when the world's most patient investor, Warren Buffett, lets nearly 400 billion dollars sit idle rather than invest it, that speaks volumes. But those building the bubble admit it. Sam Altman acknowledged as early as summer 2025 that valuations were insane and that some would get burned. Jeff Bezos spoke of an industrial bubble. Ray Dalio in January saw the very early stages of a bubble. When the firefighters themselves say it's going to burn, you don't wait to see the flames. The easy reading is that all of this remains Wall Street's problem. Hedge funds absorb the losses, and too bad for the suits and ties of Manhattan. That was true before, much less so today because the money financing all of this passes through the same banks, insurers, and pension funds. The private credit companies behind AI, Blue Apollo, BlackRock, Pimco, don't pull money out of a hat. It comes from regulated banks, life insurers, pension plans, and from 2026 onwards, from simple retirement savings accounts after the US administration opened that door. So, a retired teacher in Ohio is unknowingly lending to Mark Zuckerberg's data center in Louisiana. You don't even need to have chosen AI to be exposed to it. If you hold a simple S&P 500 index fund, you already own about 8% of Nvidia and nearly a quarter of semiconductors without having decided to. This money doesn't stop at the US border. The ECB has warned that eurozone investors hold large positions in dollar-denominated AI assets and US private credit, and that some of these funds are already facing redemption requests that are testing their limits. Europe is not just importing risk; it is building its own machine. In March, Mistral borrowed 830 million dollars to buy nearly 14,000 Nvidia chips near Paris, financed by the biggest names in French finance: BNP Paribas, Crédit Agricole, BPI France, and La Banque Postale. If Mistral's revenue catches up, that's how Europe builds a champion. Otherwise, these loans end up on the balance sheets of banks that also hold French savings. It's becoming even more direct. This spring, banks like JP Morgan are looking to sell parts of Oracle's 38 billion dollar loan to non-bank buyers, including European insurers and asset managers. US debt is being repackaged and shipped across the Atlantic in real-time. The plumbing looks different depending on the continent, but the pipes originate in the same place: your savings. This is the mechanism that exploded in 2008 when US subprime debt, sliced into securities and sold as safe, eventually brought down German and Swiss banks. European savers paid for an American crisis because the debt had migrated eastward before anyone understood how far it had spread. So, should we conclude that everything will collapse tomorrow? This is where many doomsday videos stop a bit too quickly. Unlike the startups without revenue in the year 2000, today's giants make enormous amounts of money. Their margins are comfortable, their balance sheets solid. If AI revenue catches up with expenses by the end of the decade, this construction becomes the foundation for the next 30 years of growth. Like the fiber optics laid during the dot-com bubble became the backbone of the web. The long-term valuation ratio for US stocks is currently around 40. Just below the record of 44 reached at the end of 1999. We are very high without having yet surpassed the madness of the year 2000. Both stories are on the table, and anyone who swears they know the ending is lying. The private credit market financing all of this is exceeding 2 trillion dollars. The wiring is now in place for this to become everyone's problem, not just Wall Street's. If you want to see where the cracks appear first, the bond market remains the best warning system ever built. Keep an eye on Oracle's credit default swaps, on the guarantees hidden in footnotes, and on the rental price of chips. And remember the dates of June 2026: the 4th, the 17th, the 23rd. Because if in a year, we're saying the bubble started to deflate that summer, these are the sessions we'll be talking about. The day someone asks you what the canary in the mine looked like, you'll answer a beignet. 26 years after the last time, the market is replaying the same play with bigger actors and better-hidden debt. It remains to be seen if this is the final act or just the intermission. We're dissecting all of this in the "Mon Radar" newsletter, the link is in the description. And you, would you entrust your money to a setup whose debt is hidden behind the name of a pastry? And if this entire video has made you a little uneasy about what you truly own without knowing it, that's exactly what Macro Insider is for. While debt is being hidden and pros are leaving the table, you have the three portfolios and the macro analysis that goes with it, one hour per month to avoid ending up as someone else's exit liquidity. Registrations close on July 12th with -35% off the annual plan and a 15-day money-back guarantee. The link is in the description.