Why Wall Street Is Ignoring Big Tech's Debt

**TL;DR:** Big tech does indeed carry enormous off-balance-sheet AI-related commitments, but comparing this to "Enron-style accounting fraud" isn't quite right. What's actually worth watching out for is circular financing, aggressive accounting practices, and the "big market illusion." ## Starting with the "Enron" Accusation A few weeks ago, *Nikkei Asia* reported that America's five largest tech companies were carrying $1.65 trillion in debt that didn't appear on their balance sheets at all. Not the debt you could see — but a second, much larger pile hidden behind it. A few days later, the *Financial Times* discovered that Nvidia had signed a $50 billion lease in Texas for a data center stuffed with its own chips — a commitment nobody had known about before. And the number keeps shifting. When *Nikkei* did its tally, most of the companies hadn't even reported earnings yet. By the time they did a few days ago, just three of them signed nearly $900 billion in new AI commitments in a single quarter. So "$1.65 trillion" is already an undercount. This is why, if you spend time on finance YouTube, you'll see plenty of commentators calling this "Enron." People with large social media followings but no obvious accounting background look at these numbers and conclude: this is another Enron. The real question under discussion here is: how much of what they're saying is true? Is this fraud? Are the real numbers being hidden from investors, Enron-style, until the whole edifice comes crashing down? Or is this actually something far more boring — and far more interesting? To answer that, you first have to understand what Enron actually was. Enron was an American energy giant that turned out to be a scam. It was hiding massive debt and losses inside a secret network of off-balance-sheet entities. The accounts investors could see were essentially fiction. When the truth came out, the company collapsed within weeks. It took down Arthur Andersen — one of the Big Five accounting firms — and wiped out the retirement savings of thousands of its own employees. So when someone points at big tech and says "Enron," they're not complaining about messy bookkeeping. They're alleging deliberate fraud on a criminal scale. ## Off-Balance-Sheet Debt: Accounting Rules or Deliberate Concealment? When you see a headline claiming tech giants have hidden over a trillion dollars in debt, you naturally assume someone committed a crime. But when you dig a little deeper, you find that a large portion of this debt consists of long-term purchase agreements for graphics cards and leases on data centers that haven't been built yet. Under standard accounting rules, if goods haven't been delivered yet, or a building isn't operational yet, you don't need to record it as a liability on your balance sheet. You disclose it in the footnotes. A good analogy is a phone contract. When you sign a two-year phone contract, you commit to paying the carrier roughly $50 a month for the next 24 months. That's a real obligation. You can't just stop. If you add it up, you're actually on the hook for more than $1,200. But you don't sit down on the day you sign and record a $1,200 liability on your personal balance sheet. You pay it month by month as you use the service. What tech companies are doing is exactly the same thing, just with a few more zeros attached. Except instead of phone contracts, it's a 15-year lease on a data center in Ohio. Of course, those leases that haven't started yet will hit the balance sheet as real liabilities once the data centers come online. That part is simply a matter of timing. Just one company — Meta — signed $233 billion in new commitments last quarter, of which $96 billion was leases that will flow onto the balance sheet as the data centers go live. The purchase commitments will mostly turn into chips and buildings the company actually owns. What *won't* be cleaned up so neatly are the cleverer things — the joint ventures and off-balance-sheet vehicles deliberately designed to stay off the books. But even that isn't hiding in the Enron sense. The details are all written in the accounts. You just have to go looking for them. So this isn't Enron-style fraud. It's a disguise, and it only works on people who aren't paying close attention. Of course, that doesn't mean big tech is beyond reproach. They're extremely aggressive. They're just doing it out in the open, under the heading of "adjusted earnings" on the front page of their earnings reports. ## Financing Methods Are Sending Signals to the Market For decades, financial commentators complained that tech companies hoarded cash, using their cash flow to buy back stock instead of investing in new things. Now these companies are issuing securities to invest in new infrastructure, and the same commentators have found a new reason to complain. Investors pay close attention to how a company pays for things, because it sends a clear signal to the market. The intuition goes like this: if you genuinely believed you'd built a machine that turns $1 into $5, you wouldn't sell half the machine to strangers to raise money to build it. You'd find a way to borrow money, build the machine, keep all $5 of profit for yourself, and pay off the loan. In a high-conviction investment like that, you'd want to keep all the equity for yourself. It's only when you're not so sure the business will make big money that you're willing to sell off part of the company. So debt isn't always a bad signal. Borrowing to expand often signals that management believes the return on investment is worth preserving for existing shareholders. Here's where it gets interesting: these companies are also issuing stock at the same time. In June, Alphabet completed the largest equity raise in its history — nearly $85 billion, with Berkshire Hathaway putting in $10 billion as an anchor investor. Berkshire is perhaps the last name you'd expect to see on that list. This is a company that generally thinks buying back its own stock is smarter than funding someone else's moonshot, and it's famous for its notoriously tough negotiating style. According to reports, it bought Alphabet stock at a discount of about 6% to market price, because of course it did. This is not an institution that pays a premium for a story. But it still decided that $10 billion into AI at scale was worth it. When you see big tech raising money through every available channel simultaneously — record debt, record equity, convertibles, you name it — the signal isn't which one they chose. It's the sheer scale of the capital raise itself. You don't go out and raise capital like a struggling company unless you think there's something on the other side of the road worth fighting for. ## "Bullshit Earnings" on the Income Statement and Stock-Based Compensation If you're spending tens of billions of dollars on data centers and chips, those assets depreciate and need replacing. But many of these companies would rather talk to you about EBITDA — earnings before interest, taxes, depreciation, and amortization. Charlie Munger once suggested that every time you read the word EBITDA, you should mentally replace it with "bullshit earnings." His point about depreciation is that it's kind of like a reverse float. You pay cash for equipment upfront, and the expense shows up later as the equipment wears out. Ignoring it is essentially assuming physical objects last forever. It's a beautiful idea, but it's almost never true. You can see the strain in the actual cash numbers. When the latest earnings landed, the four largest hyperscalers combined posted their lowest free cash flow in a decade — just $7 billion — while Alphabet turned cash flow negative for the first time since going public. Then there's stock-based compensation. Tech companies love paying employees in stock, then adding that expense straight back to the earnings they show investors, on the grounds that it's "non-cash." Aswath Damodaran of NYU has called adding back stock-based compensation "one of the worst abuses of modern financial reporting." His point is that it isn't a non-cash expense like depreciation — it's a barter transaction. If a company sold shares on the market and then paid employees in cash, everyone would call that a cash expense. But handing the stock directly to employees instead of selling it and paying cash doesn't make the cost disappear. Warren Buffett has been asking the same question for years: if options aren't a form of compensation, then what are they? If compensation isn't an expense, then what is it? And if expenses shouldn't count toward the calculation of earnings, then where exactly are they supposed to go? To stop all those shares issued to employees from diluting the share count, these companies buy back their own stock with real money and tell investors it's returning capital. In reality, they're running as hard as they can on an expensive treadmill just to stay in place. Here's the trap: buybacks only truly benefit remaining shareholders when the stock is bought cheaply. But a company trying to absorb its own stock-based compensation can't afford to wait for a good price. It has to keep buying on schedule, no matter what the stock price is that quarter — and prices lately have been anything but cheap. ## Circular Financing: Who's Paying Whom? Nvidia is currently pushing forward a total of more than $750 billion in AI deals. It's in negotiations to provide $250 billion in "back-to-back" support to OpenAI, helping it lease computing capacity, and to finance another $350 billion in chip purchases for OpenAI. It has also invested $5 billion in a secretive new startup founded by Ilya Sutskever, OpenAI's former chief scientist. Google has agreed to guarantee Anthropic's lease payments, effectively giving it a $35 billion loan. SoftBank has committed $65 billion to OpenAI and taken out a $40 billion bridge loan to finance it. If you draw a map of who owns whom, the companies at the center of the AI boom are, for the most part, all investing in each other. If you're a car salesman trying to hit your monthly quota, you might think: lending customers the money to buy cars from you, then booking it as sales, is a very effective way to clear inventory — at least until the customers stop paying. The fear in the market is that AI has become a giant version of this pattern: a vast web of companies funding each other's revenue. Nvidia's CEO Jensen Huang has called any suggestion that this is a circular economy "absurd." That's a strong word, especially when you're simultaneously guaranteeing purchases of your own products worth up to a quarter of a trillion dollars. But to be fair, he does have a point. As the *Financial Times* has noted, this is really just old-fashioned vendor financing. Telecom equipment makers and aircraft manufacturers have been writing checks to help customers buy their products for decades. Nvidia's rationale is equally defensible: the AI boom is moving so fast that companies like OpenAI simply can't raise the compute they think they need through ordinary debt or equity. So by stepping in, Nvidia locks in a customer, ensures its chips actually get used, and if the bet pays off, it ends up owning a piece of a pie that could be enormously valuable. The problem with vendor financing is what happens when things go wrong. That's when it becomes a double whammy: you don't just lose the customer — you lose the money you lent the customer to be a customer in the first place. Credit guarantees make it worse. If equity goes to zero, that's just wasted money — annoying, but survivable. But when you've promised to repay the customer's debts if they get into trouble, that can turn a valuation problem into a solvency problem. Right now, Nvidia generates around $20 billion in cash per year. So if one or two of these startups goes wrong, it can absorb the hit. The question is what happens when the total of guarantees climbs into the hundreds of billions, and a company that never used to take on debt suddenly finds itself standing behind everyone else's promises. The most obvious signal is in Nvidia's own credit market. When these deals were announced, the cost of insuring its debt against default posted its largest single-day jump on record. Translation: the people whose entire job is to price the risk of "Nvidia not being able to pay its debts" looked at all of this and became noticeably less relaxed. The real risk has never just been that the AI market turns out smaller than expected. It's that the people buying the chips and the people making the chips are increasingly the same people. All this circular financing is happening because every participant is convinced that the AI market will be so astronomically large that everything spent today will look like a rounding error tomorrow. ## The Real Problem: The Big Market Illusion Aswath Damodaran has a name for what happens next. He and his co-author Bradford Cornell call it the "big market illusion." Here's how it works: a new technology appears, with a huge potential market attached. A wave of companies emerges to serve that market, and investors price every single one of them as if it will be the winner. It's not the companies doing the self-promoting — it's the people buying the stock doing the pricing. Each group of investors looks at the company they've chosen and thinks it's obviously the future giant. The problem is, they can't all be right. If you add all these companies up and look at what the market expects each one to earn, you get a number bigger than the market itself. Everyone is being priced as if they'll come in first in a race that only has one winner — which is why the entire market can be priced for a future that's mathematically impossible. The story is doing all the work, and nobody's looking at the numbers. So how big is the story here, exactly? *The Economist* estimates that AI construction is on track to become the largest investment wave in history. In just this year alone, about $900 billion is being spent on chips, data centers, and power — of which more than $400 billion is borrowed. Then it calculated what would be needed to pay for all of this: the industry would need to earn roughly $2.5 trillion per year from AI revenue. That's more than the entire global tech sector currently earns in a year across all its business lines. The actual numbers are nowhere close. Adoption is real: about one in five American companies reports using AI in some form, but a lot of them are using the free versions. According to a Bank of England study, the average American executive spends about 100 minutes a week on AI. That's not a typo. The largest capital investment in human history is being supported by about an hour and a half of usage per week. ## Conclusion Directly equating Big Tech's off-balance-sheet AI commitments with Enron is oversimplifying the issue. Most of these commitments are written in the footnotes and will gradually move onto the balance sheet as data centers come online. The scale of the financing itself is even seen as a signal of high management conviction, not concealment. But the real concern isn't the accounting question of "whether the debt is hidden off-balance-sheet." It's that the entire industry is built on circular financing and a shared belief in astronomical future revenue. If that belief wavers, the guarantee chain, the credit markets, and valuation levels could all come under pressure at the same time. Source: Why Wall Street Is Ignoring Big Tech's Debt – YouTube (https://www.youtube.com/watch?v=NufJ7g63KSY)