The Man With a Track Record
Jeremy Grantham has been wrong before. He is also the person who called the dot-com crash, the 2008 housing collapse, and the 2021 speculative frenzy before each of them unwound. When he speaks about asset bubbles, markets have learned, sometimes too late, to at least hear him out.
In January 2026, Grantham appeared on the Merryn Talks Money podcast and said what he had been hinting at for months. "I think it's obviously a bubble, and I think it's quite a simple story."
That phrase, "quite a simple story," is the tell. Grantham is not presenting a scenario with uncertainty bands. He is not hedging toward a soft landing. He is stating a conclusion, and he has staked his professional credibility on similar conclusions before, and been right.
The co-founder of GMO, a Boston-based asset manager with more than a hundred billion dollars under management, Grantham has spent decades building a framework for spotting the point where narrative detaches from math. He believes that point has arrived again.
Every Condition Met
Grantham's framework for identifying bubbles is pattern-based. He looks for a specific combination: extraordinary price appreciation that outpaces any reasonable fundamental justification, a compelling story about why the old rules no longer apply, and a state of near-universal belief that the gains will continue indefinitely.
His verdict on AI stocks is unambiguous. "The probabilities that AI will not bust are slim to none. It meets every condition of the railroads and the Internet."
Those two historical analogies are precise for a reason Grantham returns to often. The railroad boom of the 19th century produced real, lasting infrastructure that changed how economies functioned. The technology was not fraudulent. The companies mostly were, in the sense that the valuations assigned to them during the boom bore no relationship to the returns investors would eventually receive. Most railroad investors lost money. The railroads stayed.
The Internet repeated the structure almost exactly. Between 1995 and 2000, the Nasdaq rose roughly 400%. Between 2000 and 2002, it fell 78%. The Internet did not disappear. Amazon and Google did not disappear. But the investors who bought into the NASDAQ at the peak of enthusiasm waited over a decade to see their money back in real terms.
Grantham's point is not that AI will fail as a technology. He is explicit that it will not. His point is that the market has already priced in a version of success so extreme that even a genuinely good outcome would likely disappoint shareholders who bought in at today's levels.
The Nvidia Prediction
Grantham does not stop at the general case. He makes a specific call: "My guess is Nvidia will lead it down, and all the others will follow for a while."
As of mid-2026, Nvidia's market capitalization sits at approximately $4.67 trillion. That is one of the highest valuations in the recorded history of publicly traded equities. For context, it exceeds the entire GDP of Japan and Germany combined. Nvidia's Q4 revenue grew 73% year over year, a number that sounds like validation until you try to construct a financial model that justifies the current market cap from that base.
Capital Economics analyst John Higgins, writing in March 2026, offered a useful distinction. One phase of the AI stock frenzy, the early speculative wave driven largely by narrative and momentum, has already cooled. But a second, rarer bubble may still be building, one driven by actual earnings that are nonetheless priced at levels no realistic growth trajectory can sustain. Higgins called the revenue trajectory "unsustainable at this pace."
Grantham's logic is clean. When a single company becomes the most visible proxy for an entire investment theme, and when that theme has attracted enough capital to push a company to $4.67 trillion, the company becomes the most vulnerable point when sentiment turns. Whatever cracks first defines the mood for everything downstream. Grantham thinks Nvidia is that crack.
He could be wrong about the timing. He does not claim certainty on when, only on whether.
Concentration and Consequence
The Magnificent Seven, shorthand for Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla, now represent roughly 36% of the S&P 500 by market weight. That degree of concentration is historically anomalous. It means that passive index investors, including the vast majority of 401(k) participants who have never made an active decision about AI stocks, are deeply exposed to a narrow cluster of AI-adjacent bets simply by owning the index.
The S&P 500 as a whole has risen approximately 80% over the past five years, even as the warnings have mounted. That fact cuts in both directions. It is evidence either that the critics have been wrong, or that the bubble has grown larger than even the most cautious observers anticipated. Markets can stay irrational longer than individual analysts expect them to.
Paul Tudor Jones has framed the moment with characteristic precision. "All the ingredients are in place for some kind of a blow off... now is so much more potentially explosive than 1999." Jones is not a perennial bear. He has made money in bull and bear markets alike. His specific concern is not whether AI matters as a technology. It is about the mechanics of what happens when the flow of capital into a theme exceeds the ability of the market to absorb new buyers. At some point, there are only sellers left.
Jamie Dimon has attached a number to his unease. He assigns roughly 30% probability to a serious market dislocation, compared to the 10% or so implied by current market pricing. Whether that gap reflects genuine insight on Dimon's part, or simply the asymmetry of a banker who sees worst-case scenarios for a living, is a judgment call. But the gap itself is notable.
The Bull Case Is Not Weak
Reading this as a one-sided story would be a mistake. The case for current valuations is not built on nothing.
Kevin O'Leary has pushed back on bubble framing with a practical argument: "You actually can see the productivity and measure it on a dollar-by-dollar basis." That is meaningfully different from the dot-com era, when companies with no revenue and no clear path to profit were trading at multiples that required a suspension of financial logic. The major AI infrastructure players have revenue. Real revenue, growing fast.
Nvidia's 73% year-over-year revenue growth is not a projection or a forecast. It happened. Companies with the scale to deploy AI infrastructure, hyperscalers like Microsoft, Google, and Amazon, are spending hundreds of billions of dollars on the hardware that Nvidia predominantly supplies. That spending is not sentiment. It is contracted capital expenditure.
Sam Altman, whose financial interest in AI success is real and should be weighed accordingly, has acknowledged publicly that some investors will lose "a phenomenal amount of money" if the current valuation levels prove unjustified. He does not dispute the possibility. His argument is that AI will still deliver "gigantic benefits" regardless of what happens to stock prices. The technology is real. The value creation is real. What remains uncertain is whether any particular stock price correctly reflects either of those things.
That distinction matters. Technology can reshape civilization while still being overpriced at the equity level. Those two things coexisted for railroads. They coexisted for the Internet. The question is not whether AI will matter in thirty years. It almost certainly will. The question is whether you are paying a fair price for the returns it will generate between now and then.
What Grantham Is Actually Saying
Grantham has been clear about the boundaries of his argument. He is not predicting that AI fails. He is not predicting a depression or a financial crisis on the scale of 2008. He is predicting that the current pricing of AI-adjacent equities has become disconnected from what even optimistic fundamental outcomes would justify.
His railroad and Internet comparisons are chosen carefully. Both technologies rewired the world in ways that made the optimists look prophetic over the long arc of history. Both still produced catastrophic losses for investors who bought at the height of speculative enthusiasm. Being right about the technology and being right about the stock are two entirely different bets.
The investors most at risk may not be sophisticated traders watching Nvidia's forward P/E. They are ordinary people who own index funds and have absorbed, without choosing it, a 36% concentration in seven companies. When Grantham says Nvidia will lead it down and the others will follow, he is describing a scenario that would hit retirement accounts, not just hedge funds.
Grantham has been early before. He has made this call when the market kept climbing for years afterward, and the waiting cost his clients real money in foregone returns. That is a real criticism of his track record, and anyone evaluating his current warning should weigh it honestly.
He has also been right.