When the AI bubble bursts, who’ll be left standing?

As Sir Isaac Newton discovered, the core scientific law of gravity is that what goes up must come down. The principle applies in many areas, which is why markets are jittery about the near-unchecked, three-year growth of stock prices fueled by the strength of the generative-AI revolution.

The market is on a tear, with a large gap growing even wider between public market valuations and the significantly higher private-market valuations of AI-exposed companies. The top five tech companies in the U.S. are, collectively, valued at more than the combined size of the Euro Stoxx 50, the U.K., India, Japan, and Canada—and account for around 16% of the entire global public equity market, according to Goldman Sachs.

It’s not just AI model makers and the firms that provide their infrastructure: It’s the associated industries that help serve the AI market. Earlier this year, Harvard economist Jason Furman estimated that U.S. GDP growth in the first half of 2025 was almost entirely due to investment in data centers.

Investors in companies like Nvidia are seeing blockbuster returns, as the firm’s value has risen more than 1,200% in the past five years, thanks to being one of the few companies able to provide the computer chips required for the AI revolution. Even so, some are worried that Nvidia is providing financing to customers looking to buy its chips—a supposedly circular chain that short sellers have quibbled with. (Nvidia, for its part, has issued responses to market analysts to refute those claims.)

It all adds up to a tetchy time, with nervousness and debate about an AI bubble. Not helping matters are the public comments about the current moment by some of the industry’s biggest names. 

OpenAI CEO Sam Altman has said that we’re currently in an AI bubble where “investors as a whole are overexcited about AI.” Microsoft founder Bill Gates has called it a “frenzy.” Meta CEO Mark Zuckerberg said on a podcast in September that an AI bubble, and its potential burst, was “definitely a possibility.” Comparisons have been drawn to the 2000s-era dot-com bubble.

Weathering the storm

So if we are in an AI bubble and it does burst, then who’ll be left standing at the end of it?

The idea that entire economies might be hit by the bursting of any bubble is unlikely to happen, reckons Christopher Tucci, professor of digital strategy and innovation at Imperial College Business School in London.

“The internet bubble, for example, wiped out many companies and investors, but the technology itself only grew in importance afterwards,” he says. Tucci sees AI in a similar way, noting, “Even if the investment bubble bursts, the underlying technology will remain critical and will continue to advance.”

And while the bubble continues to inflate, Tucci believes that’s good news for smaller companies. “At the moment, large amounts of money are flowing into AI startups,” he says. “This lowers startup costs, increases the number of competing companies, and creates vulnerabilities, mainly for investors.”

But if and when that bubble bursts, those smaller companies are more likely to be exposed, while larger companies will be insulated from more significant risks. 

“Survivors will be the ones that own distribution,” says Sergey Toporov, partner at early-stage VC firm Leta Capital. Toporov is blunt about the lack of a moat for smaller companies, saying, “Nobody cares about your ‘best-in-class AI startup’ unless people actually know it exists.”

In that view, companies like the big four AI firms—Google, OpenAI, Anthropic, and Meta—are likely to weather any storm, but smaller competitors could struggle. “The rest will consolidate or become specialized model shops,” Toporov says.

Smaller companies that have what Toporov calls “defensible advantages” like proprietary data or deep integration into business workflows could withstand an AI-caused market correction. He says the same is true for firms with “strong distribution, recurring demand, and a deep technical moat.”

Companies that piggyback on existing technology, including AI wrapper services that use their larger competitors’ AI models in order to provide answers to their customers, sometimes in specific specialties, may face a tough road ahead.

Big unknowns

However, not everyone agrees with that vision of the future. “AI apps with high valuations look the riskiest at the moment,” says Sampsa Samila, professor of strategic management at IESE Business School. “They don’t have easy moats against improving foundation models or other apps.” 

Samila believes even those that operate foundation models, like OpenAI, could be in a difficult position. “Foundation labs burning billions are also looking shaky,” he says. “It’s not at all easy to see how OpenAI will manage, unless it develops winner-take-all superintelligence.”

In part, that’s down to what Samila sees as “circular financing deals,” including those supported by Nvidia’s funding in order to obtain Nvidia chips to power their models.

While OpenAI could struggle because of its cash burn, Samila contends that bigger, more established names in the space are better placed to weather the problems. “Google is interesting because they control TPUs [tensor processing units], have proprietary data from Search, YouTube, and Gmail, and are already monetizing AI through Cloud,” he explains. 

But the big unknown for Google is whether its rollout of AI-native ads can replace its search revenue. Another area of concern for Google, given competition from the likes of Microsoft, is that its tech stack doesn’t always integrate well with the existing IT systems being run by organizations.

“Amongst the AI apps, deep embedding into customer workflows is going to be key to survival,” Samila says. Many companies tend to use Microsoft’s products rather than Google’s in large part because it’s what they’ve always done.

Whatever happens, most people believe there are fundamental differences between a possible imminent burst of the AI bubble and the dot-com stock market crash. The Magnificent Seven tech firms have a 24-month forward price-to-earnings ratio that is 25 times their collective valuation—high, but half the level it was in the dot-com era. Price-to-earnings growth is also around half the level it was a quarter century ago.

And many of the biggest names in the space are well-capitalized tech firms with cash reserves that can pay for any financial hiccups ahead in a way that the dot-com era’s biggest names couldn’t.

Regardless, those in and around the AI sector need to be aware of what’s ahead.

“When a correction comes, venture capital will dry up, potentially for several years,” Tucci predicts. “In the long run, however, AI as a technology will continue to grow in importance, regardless of short-term investment cycles.”

View the full article