Is AI a bubble? A value investor's view

Every few years a technology arrives that the market decides will change everything — and the share prices race ahead of the reality. AI may be transformative and its stocks overpriced at the same time. Those aren't contradictory. Here's how a value investor separates the two.

Two questions people confuse

"Is AI a bubble?" actually contains two separate questions that get tangled together. First: is artificial intelligence a genuinely important technology? Second: are AI-related stocks priced sensibly? The answer to the first can be an emphatic yes while the answer to the second is no. The internet was revolutionary and the dot-com stocks of 2000 were wildly overvalued. Both were true at once.

A value investor cares about the second question. The greatness of a technology tells you nothing about whether today's price is a good deal — because the price already reflects the market's expectations. You make money not when a company does well, but when it does better than the price implied.

What history rhymes with

Bubbles share a shape: a real innovation, genuine early winners, then a wave of capital that bids prices far beyond what the underlying cash flows can justify. Railways in the 1840s. The internet in the late 1990s. In each case the technology delivered — and many early investors still lost money, because they paid prices that assumed perfection.

The tell isn't whether the technology is real. It's whether valuations require near-flawless execution for a decade to make sense. When a stock is priced for everything to go right, anything less than perfect becomes a loss.

The capital-spending question

Here's where a value lens sees something the hype misses. The current AI boom is being funded by enormous capital expenditure — the big technology companies are spending tens of billions building data centres. That spending shows up as revenue for the chip and infrastructure suppliers, which makes their numbers look spectacular.

But capital spending cycles turn. If the returns on all that investment disappoint, the spending slows, and the suppliers whose valuations assumed it would continue forever get repriced fast. A value investor asks: is this revenue durable, or is it one company's capex boom that becomes another's revenue cliff when the cycle turns?

How a value investor actually responds

The other side: the growth investor's case

Intellectual honesty demands acknowledging where the value lens has been wrong. Many of the best-performing stocks of the last two decades — the ones that turned modest sums into fortunes — looked "overvalued" by traditional value metrics almost the entire way up. A strict Graham-style investor would have avoided most of them and missed the greatest wealth creation in market history.

The growth argument runs like this: for a genuinely exceptional business — one compounding at high rates with a widening moat and an enormous addressable market — today's high multiple can still be cheap relative to what the company becomes. If a company grows into and beyond its valuation, the entry price that looked expensive turns out to have been a bargain. Amazon traded on nosebleed multiples for years while building one of the most dominant businesses ever. Judged purely on price-to-earnings, it was perpetually "overvalued." Judged on what it became, it was cheap the whole time.

Applied to AI: if a handful of these companies are building genuinely durable, category-defining positions, paying up for them may prove entirely rational — even at prices that make a value investor wince. The metrics that flag them as overvalued are backward-looking; the value is in a future they may well deliver.

The honest answer

Is AI a bubble? Parts of it almost certainly are — not because AI isn't important, but because some prices assume a perfection that rarely arrives. Other AI-exposed companies may be genuinely reasonable. The point of a value approach isn't to call the whole sector a bubble or not. It's to look past the label, value each business on its own cash flows, and only pay a price that leaves room to be wrong. That discipline is what let careful investors survive every previous bubble — the technology was real each time, and the price was still what mattered.

Common questions