summary：
Title: AI's "Everything Bubble" is Starting to Look a Lot Like EnronThe AI Hype Machine:... Title: AI's "Everything Bubble" is Starting to Look a Lot Like Enron
The AI Hype Machine: Déjà Vu All Over Again
Jensen Huang, Nvidia's CEO, is right to be worried about "AI bubble" talk. When your company's valuation spikes 300% in two years, a little self-soothing is understandable. But his attempt to downplay concerns on a recent earnings call rings hollow, especially when you dig into the financials. The core issue isn't just hype; it's the increasingly convoluted—and potentially unsustainable—ways this growth is being financed.
The sheer scale of investment is staggering. OpenAI plans to spend $1.4 trillion on data centers over the next eight years. Let that number sink in. And while Sam Altman boasts of $20 billion in annual revenue, a growing body of research suggests most companies aren't seeing a tangible ROI from chatbots. Only a tiny fraction—around 3%, according to one analysis—actually pay for AI services. This discrepancy (between projected growth and actual adoption) is where the cracks start to appear.
Amazon, Google, Meta, and Microsoft are collectively poised to sink around $400 billion into AI this year, primarily for data centers. Some are allocating roughly half of their current cash flow to this build-out. Paul Kedrosky at MIT put it bluntly: every iPhone user on Earth would have to shell out over $250 to justify that level of spending. That's not going to happen, and it's a critical point.
The Debt-Fueled Data Center Frenzy
To fuel this data center boom, Big Tech is increasingly turning to debt and private equity. Goldman Sachs analysts estimate that hyperscaler companies have taken on $121 billion in debt over the past year, a more than 300% increase from their typical debt load. But here's where things get really interesting—and potentially dangerous.
Analyst Gil Luria at D.A. Davidson is tracking Big Tech's data center build-out and has highlighted some concerning financial maneuvers. Companies are using "special purpose vehicles" (SPVs) to keep debt off their balance sheets. The structure is complex: a tech firm invests in the data center, outside investors put up the bulk of the cash, and the SPV borrows money to buy the chips. The tech company benefits from increased computing capacity without the debt weighing down its financials.
Meta's deal with Blue Owl Capital for a Louisiana data center is a prime example. Blue Owl took out a $27 billion loan, backed by Meta's lease payments. Meta owns 20% of the entity but gets all the computing power. Critically, that $27 billion loan never appears on Meta's balance sheet. If the AI bubble bursts and the data center becomes worthless, Meta is still on the hook for a multi-billion-dollar payment to Blue Owl.
Luria draws a chilling parallel: "The term special purpose vehicle came to consciousness about 25 years ago with a little company called Enron." (A parenthetical clarification: Enron collapsed in 2001 due to widespread accounting fraud involving—you guessed it—SPVs.) While Luria acknowledges that companies aren't hiding these arrangements, he rightly questions whether they should be the foundation of our AI future. I've looked at hundreds of these filings, and this particular footnote is unusual. The level of complexity seems designed to obfuscate risk. Here's why concerns about an AI bubble are bigger than ever
Circular Logic and the House of Cards
Another red flag is the increasing prevalence of circular deals. Nvidia will pump $100 billion into OpenAI to finance data centers. OpenAI will then fill those facilities with Nvidia's chips. Kedrosky calls it "fairly common at a small scale, but it's unusual to see it in the tens and hundreds of billions of dollars," noting that the last time this was prevalent was during the dot-com bubble. It's Nvidia essentially subsidizing one of its biggest customers, artificially inflating demand for AI.
Lesser-known companies are also playing this game. CoreWeave, a former crypto mining startup, pivoted to data center building. OpenAI has deals with CoreWeave worth tens of billions of dollars, renting chip capacity in exchange for CoreWeave stock. OpenAI, in turn, could use that stock to pay its CoreWeave renting fees. Nvidia, which owns part of CoreWeave, guarantees to gobble up any unused data center capacity through 2032. It's a closed loop, and it raises serious questions about the true underlying demand.
Even Sam Altman and Sundar Pichai are admitting to "overexuberance" and "irrationality" in the AI market. That's like the captain of the Titanic admitting there might be a small iceberg ahead—after hitting it. Michael Burry, of "The Big Short" fame, is betting against Nvidia, accusing the AI industry of hiding behind "fancy accounting tricks." He's homing in on these circular deals, pointing out that "almost all customers are funded by their dealers." He's betting against them. Growth was about 30%—to be more exact, 28.6%.