AI News Today: Concerns Mount Over a Potential Bubble

Generated Title: AI's $1 Trillion Hangover: When the Data Centers Go Dark

Alright, let's talk about this AI "boom" everyone's so hyped about. Nvidia's stock is up 300% in two years? Cool. But let’s dig into the numbers, because something smells off.

The Data Center Delusion

Jensen Huang, Nvidia's CEO, is out there trying to calm the bubble fears. White House AI czars and Silicon Valley investors are all chanting the "investment super-cycle" mantra. JPMorgan Chase is calling it a "major revolution." They would, wouldn't they?

Paul Kedrosky at MIT isn’t buying it. He sees a ton of capital pouring into something that's still mostly speculative. And he's right to be skeptical. OpenAI (the ChatGPT folks) are supposedly pulling in $20 billion a year and plan to drop $1.4 trillion on data centers over the next eight years. Trillion. With a "T." That's predicated on everyone buying into their AI services.

Here's the problem: studies are showing most companies aren't seeing a boost to their bottom lines from these chatbots, and only a tiny fraction (3%, according to one analysis) are actually paying for AI. MIT economist Daron Acemoglu, a Nobel laureate, calls it "exaggeration."

Amazon, Google, Meta, and Microsoft are collectively planning to throw around $400 billion at AI this year, mostly on data centers. Some are even dedicating half their current cash flow to building these things. Kedrosky puts it bluntly: that would require every iPhone user on earth to cough up over $250. Not gonna happen.

To avoid burning through all their cash, companies like Meta and Oracle are getting creative (read: risky) with private equity and debt to finance this data center spree.

The Enron Redux?

Goldman Sachs analysts found that hyperscaler companies (the big cloud and computing players) have racked up $121 billion in debt over the past year—a 300%+ jump from their usual debt load. Gil Luria at D.A. Davidson is tracking this Big Tech data center boom and sees some shady financial maneuvers, specifically "special purpose vehicles."

Here's how it works: A tech firm invests in a data center, outside investors put up most of the cash, and then the special purpose vehicle borrows money to buy the chips inside the data centers. The tech company gets the computing power without the debt showing up on their balance sheet. It’s like having a mortgage on a house you don't technically own.

Meta and Blue Owl Capital recently did this in Louisiana. 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. The $27 billion loan? Doesn't show up on Meta's books. If the AI bubble bursts, Meta's on the hook for a multi-billion-dollar payment to Blue Owl.

Luria points out that "special purpose vehicles" should ring a bell. "The term came to consciousness about 25 years ago with a little company called Enron," he said. (That's a parenthetical clarification for those who weren't around for the early 2000s). Sure, companies aren't hiding it this time, but that doesn't make it any less risky.

Silicon Valley is betting that massive new AI revenues will cover all this debt. But Morgan Stanley analysts estimate Big Tech will spend about $3 trillion on AI infrastructure through 2028, and their cash flows will only cover half of that.

Luria warns that if the AI market even plateaus, we'll quickly have too much capacity, and the debt will become worthless. He sees parallels to the dot-com bubble, where debt-financed fiber-optic cables were built for a future that never fully materialized.

Then there's the circular nature of these AI investments. Nvidia is pumping $100 billion into OpenAI to fund data centers. OpenAI then buys Nvidia's chips to fill those facilities. It's like Nvidia is subsidizing its own customer to artificially inflate demand. Kedrosky calls it "unusual to see it in the tens and hundreds of billions of dollars." The last time he saw this kind of thing, it was during the dot-com bubble.

Lesser-known players are getting involved, too. CoreWeave, a former crypto mining startup, pivoted to data center building to ride the AI boom. OpenAI is renting CoreWeave's chip capacity in exchange for stock in CoreWeave, which OpenAI could then use to pay its renting fees. Nvidia, which also owns part of CoreWeave, has a deal guaranteeing they'll gobble up any unused data center capacity through 2032.

Acemoglu from MIT calls these kinds of deals a potential "house of cards."

And this is the part of the report that I find genuinely puzzling. Why are investors so willing to ignore the obvious warning signs? Are they blinded by the potential for massive returns, or are they simply caught up in the hype?

Some high-profile investors are starting to get nervous. Peter Thiel dumped his entire stake in Nvidia worth around $100 million earlier this month. SoftBank sold a nearly $6 billion stake in Nvidia. Michael Burry (of "The Big Short" fame) is betting against Nvidia, accusing the AI industry of hiding behind accounting tricks. He's focusing on the circular deals.

Burry wrote on X, "True end demand is ridiculously small. Almost all customers are funded by their dealers." He later added, "OpenAI is the linchpin here. Can anyone name their auditor?"

Even OpenAI CEO Sam Altman admitted that investors are "overexcited about AI." Google CEO Sundar Pichai told the BBC that "there are elements of irrationality" in the AI market right now.

Asked how Google would fare if the bubble bursts, Pichai responded: "I think no company is going to be immune, including us." Here's why concerns about an AI bubble are bigger than ever

When the Music Stops, the Chips Will Be Down

The AI boom feels less like a revolution and more like a financial engineering scheme propped up by hype and debt. This isn't about innovation; it's about a rush to build infrastructure based on speculative returns. And if those returns don't materialize, we're looking at a lot of very expensive, very dark data centers.