Is AI a Bubble? How a Researcher Would Answer

AI has carried the market for three years, and every sharp sell-off in semiconductor stocks restarts the same argument: bubble or breakthrough? Most takes are vibes. Here is how a researcher would actually approach the question, and why it makes a great student research project.

First, define the claim

“Bubble” has a meaning: prices driven far above any defensible estimate of fundamental value, sustained by the expectation of selling to someone else at a higher price. That is a testable claim, not an insult. The dot-com era qualified in 2000. Whether AI qualifies now depends on numbers, not headlines.

The evidence on each side

The case for froth: valuations price in years of perfect execution, capital spending on data centers is enormous and circular (chip makers’ customers are also their investors), and history says infrastructure booms usually overshoot demand. The case against: unlike 2000, today’s AI leaders generate real earnings, and recent results have repeatedly beaten expectations. Both things can be true at once: a real technology and an overpriced one.

How a researcher would test it

Pick a measurable angle. Compare today’s price-to-earnings dispersion in AI-linked stocks against the 1999 tech sector. Track what fraction of AI capex is funded by profits versus debt and equity issuance. Study what happened to infrastructure builders versus infrastructure users after past booms: railways, telecoms, fiber. Each is a defensible paper with public data, and each forces you past the headline into the mechanism.

Why this matters for students

Questions like this are where finance research gets real: high stakes, live disagreement among serious people, and evidence that does not fit neatly on either side. If you are a high school student who wants to do that work properly, with a mentor who has done it for a living, that is exactly what GRF’s research desks exist for.