Few figures embody the AI frenzy more than Jensen Huang, the CEO of chip giant Nvidia, whose market value has surged by 300% over the past two years.
Amid this feverish momentum, Huang sought in his first remarks during the latest earnings call to calm concerns over an inflating bubble.
“There’s a lot of talk about an AI bubble… but from our perspective, we see something completely different,” he told shareholders.
As the debate over an AI bubble deepens, it’s clear that those with the most to gain from continued AI spending are the ones dismissing fears of excess and speculative overreach.
David Sacks, investor and head of the White House AI Office, said on the All-In podcast: “I don’t think we’re at the start of a collapse cycle… we’re in a boom, in a super-cycle of investment.”
Prominent investor Ben Horowitz said: “The idea that we’ll face a demand problem in five years sounds absurd to me… if you look at demand, supply, and valuations relative to growth, this doesn’t look like a bubble at all.”
And in an interview on CNBC, JPMorgan’s Mary Callahan Erdoes called the characterization of massive AI inflows a “crazy idea,” adding: “We are on the cusp of a major revolution that will transform how companies operate.”
But a closer look reveals fragile foundations
Still, some observers argue that what’s happening in the AI industry today is genuinely worrying.
Investor and MIT Digital Economy researcher Paul Kedrosky says the vast sums flooding into this “revolution” are still fundamentally speculative.
“The technology is extremely useful, but the pace of improvement has slowed dramatically… so the belief that the revolution will continue at the same momentum over the next five years is unfortunately mistaken,” he said.
Massive inflows… and questionable growth
The scale of current spending is staggering even for financial analysts.
OpenAI — the creator of ChatGPT, which ignited the AI race in late 2022 — says it generates $20 billion in annual revenue and plans to spend $1.4 trillion on data centers over the next eight years. This level of investment requires continuous growth in demand for its services.
But doubts are rising: research increasingly shows that most companies are not seeing meaningful financial benefits from chatbots, and that only about 3% of people pay for AI tools.
MIT economist and 2024 Nobel laureate Daron Acemoglu said: “I have no doubt that real value-adding AI technologies will emerge over the next ten years, but much of what we are hearing from the industry right now is exaggerated.”
Even so, Amazon, Google, Meta, and Microsoft together are expected to inject nearly $400 billion into AI this year, mostly to fund data centers. Some of these firms plan to allocate up to 50% of their cash flow to building these facilities.
As Kedrosky put it: “For this level of spending to make sense, every iPhone user in the world would have to pay more than $250… and that’s not going to happen.”
To avoid draining liquidity, companies like Meta and Oracle have started turning to debt and private financing to support the data-center boom.
Risky financing… and the return of special-purpose vehicles
Goldman Sachs analysts found that hyperscalers — companies with massive cloud and computing capacity — added $121 billion in new debt over the past year, an increase of more than 300% versus the sector average.
Analyst Gil Luria at D.A. Davidson says tech giants are using special-purpose vehicles (SPVs) to obscure debt exposure on their balance sheets.
One example: a data center in Louisiana financed by Blue Owl Capital in partnership with Meta. Blue Owl took out a $27 billion loan, while Meta enjoys the capacity without showing the debt. But if demand weakens and the center stalls, Meta will face multibillion-dollar obligations.
Luria said: “The term SPV surfaced 25 years ago with a small company called Enron… today companies are not hiding it, but that doesn’t mean it’s a sustainable model for the future.”
Huge spending built on expectations that may be illusions
Companies are projecting massive AI revenues in the coming years. But Morgan Stanley estimates that big tech will spend around $3 trillion on AI infrastructure by 2028 — and cash flow will cover only half of it.
“If market growth slows even slightly, we’ll be left with excess capacity and overbuilt infrastructure, debts that become worthless, and significant losses for financial institutions,” Luria warned.
The previous bubble in the early 2000s also burst after debt accumulated to build capabilities the market wasn’t ready to use.
Giant circular deals add to the anxiety
Analysts point to circular deal structures that artificially inflate demand.
One example: a $100 billion deal between Nvidia and OpenAI, in which Nvidia funds data centers that will later be filled with Nvidia chips. Kedrosky described the logic: “I want OpenAI to buy more of my chips, so I give them the money to do it.”
CoreWeave — originally a crypto-mining platform — leases data-center capacity to OpenAI in exchange for shares, while Nvidia guarantees the purchase of any unused capacity through 2032.
Acemoglu said: “The risk is that these deals ultimately expose a fragile financial structure like a house of cards.”
Signs of fear that the bubble may burst
Some high-profile investors have already shown anxiety.
Billionaire Peter Thiel sold his entire Nvidia stake worth $100 million, while SoftBank exited a roughly $6 billion position.
Skeptics are increasingly focused on Michael Burry — famous for predicting the 2008 crash — who recently bet against Nvidia and criticized “accounting tricks” and circular financing.
Burry wrote on X: “Real demand is laughably small… almost every customer is funded by the vendors.”
He added: “OpenAI is the lynchpin here… can anyone name its auditor?”
Even top executives acknowledge the hype
OpenAI CEO Sam Altman said last August: “Are we in a phase of investor overexuberance? In my view, yes. Is AI the most important thing happening in a long time? Also yes.”
Google CEO Sundar Pichai told the BBC that “there are elements of irrationality” in today’s AI markets, adding that any bubble burst would hit everyone: “No company will be immune — including us.”
