“If we delivered a bad quarter, it is evidence there’s an AI bubble. If we delivered a great quarter, we are fueling the AI bubble,” said Nvidia CEO Jensen Huang in the aftermath of the chipmaking giant’s Q3 results.
On 19 November 2025, investors waited with bated breath as the markets closed and Nvidia was set to release its results. AI stocks had taken a battering in the days leading up to it, partly led by warnings sounded by the investor Michael Burry. Better-than-expected earnings did nothing to placate fears, and the AI-infused S&P 500 continued its week-long nosedive before attempting a recovery in the days that followed.
But the party continues. Three months on and Nvidia reported a strong Q4, with revenue up 20% from Q3 and 73% year on year. Although the market has been incredibly choppy in the intervening three months, the ‘AI bubble’, it seems, has yet to pop.
“Nvidia’s Q4 earnings confirm the AI revolution is firing on all cylinders, which somewhat puts the AI bubble fears to rest, at least for now,” says Kate Leaman, chief market analyst at AvaTrade. “The results provide hard evidence that Big Tech’s AI bet is paying off big time, with Nvidia as the undisputed market leader. The main story here is that AI infrastructure demand is structural, not cyclical, and Nvidia owns the road ahead.”
Most of the concerns centre around the deals struck between different tech giants, AI developers, cloud companies and chipmakers. But how this circular financing adds up is confusing. To make matters more confusing, experts in witnessing these deals play out in real-time cannot agree on when the bubble will pop, what that might mean, and whether there’s a bubble in the first place. Assuming we take the word of Burry and many others, including the diligent author Ed Zitron, at face value, what knock-on effects might we expect when it eventually goes pop?
In the long run, fears remain that an AI bubble is continuing to be fueled by the likes of Nvidia, OpenAI, Oracle, Microsoft, et al. Huang even went so far as to say “the whole world would have fallen apart” if Nvidia’s Q3 earnings was off by even “just a hair”.
AI bubble or just ‘froth’? Spotting the symptoms
We’re six months on from the infamous MIT report that found that 95% of generative AI pilots have struggled to deliver financial return – but the waters are muddy, with MIT also finding through its Project Iceberg index that AI can already replace 11.7% of jobs in the US. Both these findings are contributing to an ongoing debate – and you can’t avoid having a conversation about AI nowadays without some kind of bubble talk coming up.
This anecdotal factor is a key indicator that a bubble is forming, Arie Brish, professor of business at St Edwards University in Austin, Texas, tells ITPro. It’s especially the case given that people who aren’t “normal investors” are talking about it. But what about hardcore metrics that support this view?
Price-to-earnings (P/E) and price-to-sales (P/S) ratios, he says, are drifting significantly out of range. When higher than normal, it means that investors have astronomical expectations for future growth, which they are pricing in today. Elsewhere, we can see “irrational exuberance” – when the fear of missing out (FOMO) overrides strategic due diligence, says Bob Hutchins, CEO at Human Voice Media.
“We see this when valuations detach completely from revenue realities, and when companies slap an “AI” label on legacy products just to bump their stock price without adding genuine utility,” Hutchins, whose doctoral dissertation is focused on the effects of generative AI on human identity and creative authenticity. Pedro Varela, head of AI at Slalom, agrees, telling ITPro that valuations are “surging far ahead of real revenues or proven economic impact”.
“AI hype is spilling well beyond the tech sector into areas that are only very loosely connected to the technology,” he adds, citing company research that 90% of businesses plan to increase AI investments next year. “That kind of enthusiasm leaking across unrelated sectors is a classic symptom of past manias, from the dot coms to the Teslas.”
Bad bubble or good bubble – which camp does AI fit into?
Echoes of the dot-com bubble are not lost on experts of all stripes. The difference is previous bubbles, including the railways and even bicycles, saw some misallocation of resources that are still used to this very day, explains Varela. There are some differences between the AI bubble and the dot-com bubble, of course, but there are also some key similarities, John Bates, CEO at document management firm SER Group, adds.
“Back then, we all got very excited about the potential of the Web and online, but there was a huge crash as just too much VC money had flowed in too quickly into unproven models and no one could see how to cash out,” Bates explains.
“AI today has similar soaring valuations that, as yet, don’t seem fully backed by clear evidence of business value. The key point, though, is that, just like the dot-com crash, that doesn’t necessarily mean the underlying technology isn’t actually incredibly powerful. I think the balanced view is that we need to be a bit more patient.”
Hutchins echoes this view, suggesting that a bubble might not necessarily be a bad thing, describing it as the “chaotic “installation phase” of a technological revolution”, adding that it “funds the infrastructure that eventually changes the world”.
A rocky road ahead for the AI industry
There are a handful of traps and inflection points on the long road ahead for the AI industry and for businesses invested in the infrastructure that will propel it forward. Firstly, it’s unfathomably expensive – and the industry is soon to face an energy bottleneck that might mean sluggishness unless more power can be unlocked.
At the same time, much of its everyday usage (such as students using it for homework) delivers “limited real business value”, says Bates. “We may soon hit the Gartner ‘trough of disillusionment’, and that could be a pretty hard stop.”
Brish says it’s ultimately hard to predict at the moment, but some scenarios include leading companies beginning to post underwhelming financial results under analyst expectations or some unrelated external event, such as a political crisis or major world conflict. In situations like these, investors become more cautious. Hutchins agrees with the former point, telling ITPro that the market might one day realize the “astronomical” cost of compute is not being met by efficiency gains or revenue growth.
“Specifically,” he ponders, “if we see a few quarterly earnings reports where major enterprises admit that their massive AI spend hasn’t resulted in the predicted productivity boom, the sentiment will shift overnight. It won’t be a technology failure; it will be an ROI [return on investment] reality check.”
How will the bubble bursting affect businesses?
The experts ITPro spoke with agree that an AI bubble bursting would affect different kinds of businesses in different ways, with some feeling the effects in more pronounced ways than others. But there is also nuance in the way that different levels of exposure to AI, and in different areas, may render a different outcome.
Tech Giants: The “magnificent seven” will be fine, says Hutchins, despite sharp valuation reductions, as Varela and Brish predict. This is in light of their massive cash reserves. They will likely pivot and move to acquire the best IP from smaller startups that have crashed and burned. With time, innovation will continue, but with more caution. They would redirect attention to “safer” applications of AI that support existing systems, Varela adds, rather than attempting to weave “grand narratives” about mass transformation.
Wider Tech Industry: Hutchins is less optimistic about the impact on the wider tech landscape, describing a bursting bubble as “a Darwinian extinction event”. That’s because thousands of startups built on top of large language models (LLMs) will vanish, and capital will dry up for anything that isn’t solving a specific problem. Bates concurs, adding that a slowdown in finding, innovation, and hiring will ensue thanks to higher caution. On the flipside, he adds that “killer applications” of AI will emerge and be successful, while the best ideas and IPs will likely get hoovered up by the bigger players through M&A activity.
Businesses that use AI: Whether there will be any material effect depends on whether a business has adopted AI for show or for substance, Valera says. If an organization used it to weave a narrative, then they would find it harder to sustain their story. But organizations that used AI in a meaningful way to experiment with operational changes – like in coding or customer service – might be better placed. Bates adds: “Even at the depths of the dot-com crash and the credit crunch, software continued to be bought and used, albeit at lower volumes, if it delivered even a slight incremental value. The same will be true for AI – the tools that actually work and move the financial needle will continue to be adopted and used.”
Should businesses worry, and how can they prepare?
Businesses, both in the tech industry and that are tech-adjacent, have faced several crises in the last few years, including the aftermath of the financial crash, COVID-19, and war in Europe. Now, an AI bubble – if you believe it is there – threatens to burst and bring down the economy with it. But businesses need not worry so long as they take adequate measures to protect themselves and their operations now.
“Don’t buy the hype, buy the solution,” says Hutchins, adding that the mitigation is simple. “As an organizational psychologist, I tell clients to focus on the human problem first. If AI solves a friction point for your employees or customers, the macro “bubble” status is irrelevant to you.”
One key step is to avoid long-term vendor lock-ins, especially with unproven startups, and to make sure that humans are always kept in the loop – not just deploy AI to reduce headcount nakedly and without a long-term strategic vision. Bates agrees, adding that companies involved in this space must double down on their focus to deliver tangible business value through clear and practical use cases. Simply put, as Varela puts it, if and when the bubble bursts, the technology will still be there – it’s just valuations, the availability of cheap capital, and the tolerance for unproven claims that all dry up.
“The main risk for firms is building their AI strategy on fragile assumptions: dependence on a single vendor that may stumble, business cases that rely on heroic productivity projections, or public commitments that outpace what the tools can currently do,” Varela explains. “To mitigate this, companies should focus on strengthening the parts of their AI journey that matter in any market conditions. That means investing in data quality, process redesign and targeted, verifiable use cases rather than grandiose promises.”
