I recently attended an MBA module at Warwick University on behavioural finance. A group of around 20 of us explored how biases, incentives, agendas, and emotions influence decision-making — both our own and that of others.

A recurring theme was how these signals fuel herd mentality. That, in turn, can drive overinflated stock valuations, euphoria, and sustained bullish markets.

A lot of parallels were drawn between what we’re seeing now with AI and the dot-com bubble. Not as a perfect analogy, but as a useful reference point.

There’s already plenty of speculation about AI in the media. But working through behavioural biases in detail made the discussion feel more grounded — and more uncomfortable.

It reinforced the idea that we may be in a period where AI’s impact is being priced aggressively.

That led to a lot of debate in the room:

  • Is the AI bubble similar to dot-com’s boom-and-bust?
  • If we’re following a “bubble-like” trajectory, what would cause it to burst?
  • How do investors and firms reduce downside risk?

The technology is real — the question is the valuation

There’s no doubt that AI is transformative. Like the internet, it represents a genuine step-change in how data and decisions are used.

Investors can see that promise. Today, a significant share of venture funding is flowing into AI enablers, platforms, and adopters. Even a private company such as OpenAI — arguably a late-stage start-up — is roughly valued at $500bn, with annualised revenue of around $20bn. That implies a valuation multiple of roughly 25× annual revenue.

This is where perception and reality begin to diverge.

Reading the price-to-sales signal

The divergence is most visible when looking at price-to-sales ratios — a rough signal of how much future growth is already being priced in.

By early 2026, the S&P 500 was trading at a trailing price-to-sales multiple of roughly 3.4–3.5, close to historical highs.

Against that backdrop, companies at the centre of the AI narrative stand out. NVIDIA, with a market capitalisation of roughly $4.4–$4.6 trillion and around $130bn in annual revenue, is effectively being valued at about $35 of market value for every $1 of current annual revenue. That expectation implies extraordinary future growth.

Even Apple, trading at a price-to-sales multiple well above the index average, reflects similarly optimistic expectations.

None of this proves valuations are wrong. But it does highlight how much of AI’s anticipated impact is already embedded in prices, and how sensitive those expectations may be to shifts in sentiment or realised returns.

Concentration risk

When many of the top 10 companies depend on the same resources — semiconductors — the diversification benefit of the index is weaker than it appears.

NVIDIA now accounts for roughly 7.4% of the S&P 500, concentrating that dependency further.

The gap between expectation and demonstrated impact is where things get more interesting.

The correction question

As more companies exploit AI to increase productivity and differentiate their offerings, strategies increasingly cater to market trends, with investors following these signals.

When the dot-com bubble burst, the market corrected and a more cautious approach followed. In the long run, the technology still transformed the world.

However, if firms don’t deliver on expectations, and if investors begin to substitute market signals with a more rational stance, is it only a matter of time before we see a similar correction?

That remains to be seen.

What’s clear is that the same behavioural forces we studied in that seminar room — overconfidence, herd mentality, narrative bias — are visible in the market right now. Recognising them doesn’t tell you when the correction comes, or whether it does at all. But it does change how you think about the decisions you’re making in the meantime.

Data sources: Multpl.com. Price-to-sales figures reflect early 2026 market data. Valuations are approximate and subject to change.