AI Bubble Trouble: Are We Reliving the Dot-Com Era?

Jensen Huang, the face of Nvidia's meteoric rise (a 300% value spike in two years), is understandably eager to quell talk of an AI bubble. He's not alone. White House AI czar David Sacks calls it an "investment super-cycle," and JPMorgan Chase's Mary Callahan Erdoes sees a "major revolution." But let's pump the brakes for a minute and look at the actual data.

The Speculative Stampede

The core issue? A massive influx of capital chasing returns that, so far, are largely theoretical. OpenAI, the poster child of this boom, projects spending $1.4 trillion (yes, trillion) on data centers over the next eight years. That's predicated on perpetually skyrocketing sales, a bet that increasingly looks shaky.

Because here's the rub: studies suggest most businesses aren't seeing AI impact their bottom lines, and only a tiny fraction of consumers—about 3%, according to one analysis—are actually paying for AI services. MIT economist Daron Acemoglu, a Nobel laureate no less, calls much of the industry hype "exaggeration."

And yet, Amazon, Google, Meta, and Microsoft are collectively poised to drop $400 billion on AI this year, primarily on data centers. Think about that: that's like every iPhone user on the planet shelling out $250. (Spoiler alert: it's not going to happen.)

To avoid completely draining their coffers, companies like Meta and Oracle are increasingly turning to private equity and debt to fuel this data center build-out. That's where things get really interesting, and potentially precarious.

Risky Business: Debt-Fueled Dreams

Goldman Sachs analysts point out that hyperscalers have taken on $121 billion in debt in the past year, a 300%+ jump from their typical borrowing levels. And some of this debt is being cleverly (or not-so-cleverly) structured to stay off balance sheets, using "special purpose vehicles."

Here's how it works: a tech firm invests a small chunk in a data center. Outside investors put up most of the cash. Then, the SPV borrows money to buy the chips. The tech company gets the computing power, but the debt remains hidden. A recent deal between Blue Owl Capital and Meta for a Louisiana data center illustrates this perfectly: Blue Owl took out a $27 billion loan, backed by Meta's lease payments, but that debt never appears on Meta's balance sheet.

Analyst Gil Luria of D.A. Davidson draws a direct line from these SPVs back to Enron. While he concedes that companies aren't exactly hiding these arrangements, the underlying risk remains. If the AI bubble bursts, Meta could be on the hook for billions to Blue Owl.

This whole house of cards rests on the assumption that AI revenues will explode and cover all this debt. But Morgan Stanley estimates that Big Tech will need to spend $3 trillion on AI infrastructure by 2028, and their own cash flow will only cover half of that. What happens if AI growth plateaus? Luria warns of overbuilt capacity, worthless debt, and a potential financial crisis.

And this is the part of the report that I find genuinely puzzling: the sheer scale of the bet being made on perpetually accelerating AI adoption rates. It's not just a matter of "if we build it, they will come," but rather, "if we build it at a cost of trillions, they absolutely must come, and keep coming forever."

Circular Logic and the Specter of 2000

Then there are the circular deals, like the rumored $100 billion agreement between Nvidia and OpenAI. Nvidia funds OpenAI's data centers. OpenAI fills those centers with Nvidia's chips. Critics (rightly) argue that this artificially inflates demand for AI. It's like Nvidia subsidizing its own sales.

As MIT's Acemoglu puts it: "The danger is that these kinds of deals eventually reveal a house of cards."

And what about the big players selling off their positions? Peter Thiel recently dumped his entire $100 million stake in Nvidia. SoftBank unloaded nearly $6 billion worth. And Michael Burry (of "The Big Short" fame) is now betting against Nvidia, calling out the industry's "fancy accounting tricks" and circular deals. (Burry's track record since 2008 has been mixed, I'll grant you that—more of a Cassandra than a consistent market oracle.)

Even AI's biggest cheerleaders are starting to show signs of unease. OpenAI CEO Sam Altman admits that investors are "overexcited." Google's Sundar Pichai acknowledges "elements of irrationality."

The question isn't whether AI is important—it is. The question is whether the current frenzy is sustainable, or if we're building another fiber-optic cable to nowhere, fueled by debt and wishful thinking.

A Repeat of History?

The dot-com bubble burst twenty-five years ago after massive debt financing built out fiber-optic cables for a future that hadn't arrived yet. Tech companies are not worried about repeating that. If hundreds of billions of dollars are spent on data centers that aren't needed in a few years, then it's another financial crisis.

This Smells Awfully Familiar