The argument over an AI bubble has shifted from whether one exists to how investors can profit before it pops. This week, some of the world's top investors clashed publicly on the issue, taking sharply opposing views while reaching a rare consensus on one point: this is the biggest bubble in our lifetime.

On October 8, Galaxy Digital founder Mike Novogratz said at the Greenwich Economic Forum that AI is in "the biggest bubble of our lifetimes," but he simultaneously advised investors to enter immediately. His logic: "Bubbles don't end like this. I know how bubbles end — they end spectacularly, and this isn't spectacular enough yet." In other words, the bubble is real, but it has not yet inflated to its extreme.

Meanwhile, Bridgewater Associates founder Ray Dalio issued a diametrically opposite warning from Singapore, calling AI a "classic bubble" that is nearing its breaking point. Nassim Taleb, author of The Black Swan, shifted his focus to the bond market, arguing that the real fault line lies there.

The market implications of this debate cannot be ignored. Goldman Sachs data shows that AI-related stocks now account for 42% of the S&P 500's total market capitalization, exceeding the 41% reached at the peak of the 1999 tech bubble and marking the second-highest market concentration in history. At the same time, the market breadth of S&P 500 constituents has fallen to its narrowest level since the peak of the dot-com bubble.

Goldman Sachs strategist Bobby Molavi, after reviewing all major capital expenditure boom cycles over the past two centuries, found that the mechanism triggering a collapse was almost never "spectacular" — instead, it began with a central bank policy pivot, and capital expenditure often continued climbing for as long as two years after stocks peaked.

Novogratz: The Bubble Isn't Big Enough, Get In Now

Novogratz's bullish thesis is built on valuations. He argues that AI-related stocks remain cheap on a price-to-earnings basis and, in near-aggressive language, urged those on the sidelines to act: "If you haven't invested in AI yet, you might as well go home and stick your head in an ice bucket."

On the policy front, he sees no political brake on AI. He noted that Washington "will find it very hard to slow AI down, even to the extent that the people building AI themselves would want."

Notably, this is not the first time Novogratz has made a "biggest bubble in our lifetime" call. In November 2017, he used the same phrase to describe cryptocurrency while maintaining a bullish stance. At that time, Bitcoin peaked the following month and subsequently fell more than 80%, with his company Galaxy Digital losing as much as $273 million in 2018.

To be fair, however, he was ultimately proven correct on Bitcoin in the long run — it just came after a brutal crash in between.

Dalio: A Classic Bubble, Close to Bursting

Dalio said in Singapore on Wednesday that AI is a "classic bubble" approaching its breaking point, driven by two forces: rising interest rates and the need to convert paper wealth into cash.

"We are on the verge of that point, but getting closer. I think we are very close."

Since June, Dalio has issued repeated warnings on this front. He told Bloomberg Television that "the process of pricking a bubble is the process of converting wealth into money." Since then, conditions have deteriorated further: long-term U.S. Treasury yields have risen to their highest level since 2002. Dalio also warned that Japan could reduce its holdings of U.S. Treasuries, because a relationship that is simultaneously creditor-debtor and adversarial "creates very difficult dynamics."

Dalio also noted that AI companies, which previously relied on equity financing, are now being forced to turn to debt financing — a shift that coincides with a sharp rise in the cost of capital, further exacerbating financial fragility.

Taleb: The Real Risk Is in the Bond Market

Taleb agreed with Dalio's diagnosis of a bubble at the Greenwich forum but differed on where the risk originates. According to Bloomberg, he said: "The bond market is very fragile."

His concern is not primarily directed at Tokyo's selling, but rather at the potential exit of U.S. pension funds and institutional investors from long-duration bonds. He believes that a withdrawal by this group would cause a far greater shock than selling by foreign central banks — and AI companies' debt financing needs are competing with U.S. Treasuries for the same pool of duration buyers.

Taleb issued a clear warning against the logic of "buy AI stocks because AI will change the world," calling it "a trap."

He cited history to point out that the biggest beneficiaries of technological change are often not the original pioneers. He also reminded that the S&P 500 is not equivalent to the U.S. economy, noting that its largest constituents are all global companies, and that the condition of middle-income households is a truer gauge of American economic health.

True to form, Taleb refused to predict when a market crash might occur — "thinking the market will fall is the wrong way to think... you should be thinking about rising risk" — and offered his signature advice: "Always hold tail hedges, even when you have no reason to hedge."

Inside Goldman Sachs: A Split Between Valuation Camp and History Camp

Goldman Sachs has not formed a unified position on this debate, with two distinctly different voices coexisting.

Strategist Bobby Molavi reviewed major capital expenditure boom cycles over the past two centuries — railways in 1845, utilities in 1929, Japan in 1989, the internet in 2000, shale oil in 2014 — and reached a pattern-based conclusion:

"The trigger is almost always the same — a central bank policy pivot. The sequence is consistent every time: stocks peak, capital expenditure continues to climb (for 6 to 24 months), then capital expenditure rolls over, and then stocks decline."

This means that when a bubble looks "spectacular enough," the top has often already passed. And the phase in which capital expenditure continues rising after stocks peak is precisely the moment when everyone points to growing spending and says, "See, there's no bubble at all."

Molavi's data is equally striking: AI-related stocks account for 42% of the S&P 500, surpassing the 41% at the peak of the 1999 tech bubble and trailing only the railway mania of the 1830s (63%) and Japan in 1985 (44%). Meanwhile, S&P 500 market breadth has fallen to its narrowest level since the peak of the internet bubble.

Bank of America strategist Michael Hartnett holds a similar view, calling AI "the biggest bubble since railways."

However, Goldman Sachs' Tony Pasquariello offers data on the other side: the median forward price-to-earnings ratio of AI infrastructure stocks has dropped from 32 times in April to 22 times. "These companies are delivering on earnings, and the market is treating them far more calmly than it did in 1999, 2007, or 2021." He also noted that AI infrastructure stocks are expected to contribute half of the S&P 500's third-quarter earnings growth, with NVIDIA and Micron together accounting for one-third of that.

The weakness of the "cheap P/E" argument is this: earnings increasingly need to be used to support capital expenditure, leaving very little free cash flow for shareholders. Currently, the free cash flow yield of the S&P 500's top ten constituents has fallen to its lowest level since the internet bubble.

Consensus and Divergence: Everyone Says Buy, but for Different Reasons

Taking all positions together, a paradoxical consensus emerges: whether it is Novogratz, who believes the bubble has not yet peaked, Dalio, who believes it is about to burst, or Bank of America and DBS Chief Investment Officer Hou Wey Fook, virtually everyone is advising investors to continue holding or buying AI-related assets — just with different hedging methods and levels of risk awareness.

Bank of America advises clients to hedge downside risk through equity derivatives while continuing to hold positions; DBS's Hou Wey Fook, citing NVIDIA's forward P/E of just over ten times, retorted: "How could this possibly be a bubble?"

The real disagreement lies in the trigger mechanism and timing. Dalio's interest rate and liquidity tightening mechanism and Taleb's bond market fragility are essentially two sides of the same coin, and the debt financing frenzy of AI companies sits right at the center of that coin. Goldman Sachs' own "monetary policy optimism" indicator has fallen to -7.98, approaching its historical low of -9.38, while the S&P 500 remains at historical highs over the same period — this divergence itself may be the most warning-worthy signal.

As Molavi's historical research reveals: the way a bubble bursts is often far more mundane than people expect — "spectacular" only becomes clear in hindsight.