Is there an AI bubble in the markets? When will it burst?

Yes, there is a bubble, and it is one of the biggest we have ever seen. The top ten names in the S&P 500 are now more stretched against earnings than they were at the peak of the dot com mania. AI is the story being sold to the public, but the real driver is cheap money, share buybacks and a handful of mega caps holding the whole index up. Strip those out and the rest of the market is already weak. When does it burst? Bubbles do not pop to a diary date. They pop when the credit behind them cracks. The warning signs are already flashing. Record private credit defaults, debt-fuelled AI capex, and circular deals between the same handful of companies. No, it is not scaremongering. Scaremongering is telling people this time is different. The maths, the charts and the history all say the same thing, and the smart money is already rotating into hard assets.

The AI bubble debate is understandable, but too simplistic. There are warning signs: market concentration, stretched valuations, enormous capital spending and increasingly circular relationships between chipmakers, cloud providers and AI developers. That should make investors cautious.

But caution is not the same as panic. Unlike parts of the dotcom era, many of today's AI leaders are highly profitable, cash generative and funding investment from genuine balance sheet strength. AI is already being monetised through cloud demand, software tools & productivity gains.

Markets could certainly correct if expectations run ahead of reality. More speculative AI stocks could fall sharply. But that is different from saying the entire market is fictitious. AI may be overhyped in places, but it is not imaginary. The real risk is not AI itself, but investors paying today for too much of tomorrow.

The answer is not to abandon markets, but to be increasingly selective about what you own and why.

Markets have been expensive for a while, even before ChatGPT stormed onto the scene. Most of the AI companies are private, and seem to throw numbers around like it's monopoly money. It will be interesting to see if Anthropic and OpenAI do look to float on the market, as is rumoured. Listed companies like Nvidia, are showing strong revenue and profit growth, but they have been big beneficiaries of AI companies hyperscaling. There is a lot riding on the continued success of this technology and so a setback of a confidence wobble could create big waves in the market. Being diversified and not following the crowd on this may well prove to be a sensible strategy.

From an HR point of view, an “AI bubble” matters less for headlines and more for hiring decisions. If the hype cools, the first impact is usually jobs: hiring freezes, “efficiency” restructures, and AI teams being trimmed back to the roles that actually deliver value. Training budgets can get cut too, and people get nervous about their roles, which hits morale and retention. If it doesn’t burst, you still get pressure, because leaders will expect AI to boost productivity quickly, and that can mean role redesign, new KPIs, and tougher performance conversations. Either way, the smart move for SMEs is steady and practical: don’t hire based on hype, upskill the team, be clear on what AI will and won’t replace, and keep your comms human. The risk isn’t AI, it’s rushing change without a plan.

The framing of an “AI bubble” mistakes the symptom for the system. Valuations are stretched because markets are pricing in future enterprise adoption that has not fully materialised yet. A small group of companies are simultaneously funding, supplying and consuming the same AI ecosystem while capex is being committed years ahead of measurable returns. The assumption is that enterprise demand will catch up. In many cases, the operational controls underneath AI deployment are still immature. Data lineage, oversight and model assurance remain unresolved at scale. Pilots clear demo conditions. They do not clear audit. The real risk is not AI disappearing, but markets reassessing how quickly AI can produce reliable commercial returns.

Nearly 90% of companies now use AI, yet only 6% report meaningful profit gains from it, according to McKinsey research, which is exactly the kind of expectation gap that turns market excitement into bubble territory. The dotcom crash proved a technology can survive even after investors wildly overprice its short-term impact, and AI may follow the same path: a painful repricing without the technology disappearing. What markets are really betting on is not today’s productivity, but tomorrow’s transformation arriving far faster than most businesses can realistically absorb it. That is why the real risk is not AI failing, but investors mistaking infrastructure spending and experimentation for immediate economic revolution. If the correction comes, it is more likely to punish exaggerated timelines than destroy AI itself.”