Over the past few weeks, the chatter around artificial intelligence (AI) stocks on Wall Street has turned intense following the recent selloff that erased nearly a trillion dollars in valuation of tech stocks.

Michael Burry has reportedly turned his gaze towards AI, making investors edgy given the famous contrarian hedge fund manager’s claims of spotting the 2008 credit bubble.

The news from Europe is that Deutsche Bank is exploring short positions in AI-related equities and synthetic risk-transfer deals (derivative-based structures to cover potential losses without selling the underlying loans) to hedge against potential default in loans made to hyperscalers such as Alphabet, Microsoft and Amazon, which are pumping billions into AI infrastructure.

Meanwhile, in Tokyo, SoftBank Group this week revealed that it had raised $5.83 billion by offloading its entire stake (over 32 million shares) in Nvidia in October, sending the stock tumbling 10%.

For some, that was the loudest signal yet that the AI rally has peaked.

But the conclusion could be hasty though as the reality is that investors are rotating their bets on AI, not retreating.

SoftBank’s exit is strategic

The context of SoftBank’s exit from Nvidia has to be seen in the context of the fund’s larger capital allocation strategy. The group, led by Masayoshi Son, after years of volatility in its Vision Fund, is in the midst of a financial reset. The sale of Nvidia stock along with a $9.17-billion partial exit from T-Mobile forms part of what CFO Yoshimitsu Goto described over an investor call as “asset monetisation,” meant to strengthen liquidity and fund new AI bets. “We want to provide a lot of investment opportunities for investors while maintaining financial strength,” Goto told analysts.

In other words, SoftBank isn’t cashing out of AI—it’s doubling down on it. SoftBank is putting over $22 billion into OpenAI, chip startups, and its own internal artificial general intelligence platform. By liquidating a “mature” holding such as Nvidia, SoftBank is freeing capital to invest higher in the AI stack where the risks are higher but so are the returns.

That’s not a warning of a top. It’s a portfolio churn.

High valuations, but not bubble

At $4.85 trillion market capitalisation, Nvidia is now worth more than Amazon and Tesla combined. It’s a staggering reflection that portrays valuation excess, yet the underlying metrics indicate exuberance. The stock’s trailing P/E is 56.7 times, while forward earnings multiple is over 30 times. With a five-year price to earnings multiple (PEG) of 0.88 times, the valuation is not wildly overvalued.

Those ratios are high, but not irrational given the company’s revenue trajectory, which has ballooned from $26 billion in FY23 to $130 billion in FY25 with projected revenue of over $207 billion in FY26, powered by hyperscaler demand for its H100 and B200 chips.

As Howard Marks, co-founder and chairman of Oaktree Capital Management, mentioned in an interview with CNBC recently: “Valuations are high but not crazy. Expensive and going down tomorrow are not synonymous.”

Unlike companies in the dotcom era that were valued on “eyeballs” and “clicks,” Nvidia’s valuations are tethered to real profits, not speculative metrics.

An often-misread historical lesson comes from Markus Brunnermeier and Stefan Nagel of Princeton University, whose 2004 research paper titled “Hedge Funds and the Technology Bubble” examined how institutional investors behaved during the 1990s dot-com mania.

Hedge funds, they found, did not burst the bubble. They rode it. They participated heavily in tech stocks on the way up, and then, sensing the turn, quietly trimmed exposure before the collapse.

The implication is profound: even rational investors, aware of froth, often prefer to stay invested in rising markets rather than fight sentiment too early. “Limits to arbitrage” and career incentives make it rational to ride bubbles longer than logic alone would dictate.

SoftBank’s sale, and Deutsche’s hedging fit that pattern perfectly. They are adjusting the sails, not fleeing the ship.

The “Insider” story is far from “Short”

Nvidia’s insider activity reflects profit-taking, not pessimism.

Over the past year, executives and directors have sold roughly 19.9 million shares, against just 1.7 million purchased, leaving insiders with 4.3% ownership.

On the institutional front, ownership remains deep and stable. The four Wall Street giants, Vanguard Group: 2.22 billion shares (9.15%); BlackRock: 1.91 billion (7.86%); Fidelity: 998 million (4.11%); and State Street: 978 million (4.03%), hold roughly 70% of Nvidia’s float, mostly through passive index funds. Such entrenched ownership limits speculative churn and dampens volatility.

In true bubbles, short sellers swarm.

In Nvidia’s case, short interest is barely 1% of float, or about 238 million shares. By comparison, during the dot-com peak, short interest in leading tech names often ran between 8–12%, according to the research paper.

With more than 10 million shares available to borrow and a borrow fee of just about 0.4% a year, taking a bearish position is cheap on Wall Street’s top stock. A trader shorting $100 million worth of stock would pay roughly $400,000 a year in borrow costs, and even a $1 billion short would cost only about $4 million annually for a big hedge fund.

In other words, anyone who wants to bet against Nvidia can do so at minimal expense. The reason short interest remains low is not because shares are hard to find or borrowing is prohibitively expensive, but because few investors are confident of betting aggressively against AI.

Rod McNaughton, professor of entrepreneurship at the University of Auckland, in his inaugural lecture at Otago University in 2000 at the peak of the dot-com bubble, had warned that “many internet firms were naked” as their business models were visible and unsustainable.

“In 2000, the internet promised to revolutionise commerce, with success measured in ‘eyeballs’ and ‘clicks,’” McNaughton recalls in an article featured in The Conversation. “Today, those indicators have become ‘tokens processed’ and ‘model queries.’ The language might have changed, but the belief that scale automatically leads to profit hasn’t,” writes McNaughton.

But there’s one crucial distinction: this time, the core of the ecosystem is profitable. Nvidia, Microsoft, and TSMC are cash-rich giants with pricing power. The speculative froth exists on the fringes, that is, among startups burning through venture cash to chase foundation-model dreams in the AI race.

It’s not blow-phoria

True bubbles aren’t defined by numbers but by psychology. “The main ingredient is mania: a sense that prices can never be too high,” mentions Marks in the interview.

In the case of listed AI stocks, we’re not there yet.

The S&P 500 now trades at a forward PE ratio of 23 times, the tech heavy Nasdaq 100’s multiple is hovering near 33 times. Those numbers show investors are paying a significant premium for growth. In contrast, at its peak in 2000, the Nasdaq Composite index was trading at a multiple of over 70 times forward earnings, while S&P 500 was trading at a P/E of 30 times.

Importantly, in the recent tech selloff, the profitable, established online property portal, Rightmove, with a dominant market share in the U.K, plummeted 28%. The reason: Rightmove announced plans to invest £60 million over the next three years and expected its operating profit to rebound after 2028.

So, it isn’t a Pets.com-style collapse; in fact it’s a reminder that investors are questioning whether every solid business really needs an AI story.

Coming back to SoftBank, does the $5.83 billion Nvidia exit symbolise a red flag for AI? Hardly. It’s part of a broader strategic repositioning to monetise exuberance and reallocate capital towards promising opportunities.

Yes, valuations are rich, but as the Princeton study reiterated, rational investors aren’t bursting the bubble. They’re riding the wave with their eyes open.

And that means the AI story, for now, is far from over.