Double, double, are we in an AI bubble?

Economics has a fairly straightforward approach to identifying bubbles. Principally, three criteria must be met for a financial market boom to be classified as a bubble. One, investors must chase a new asset class where there is no precedent for valuation. Two, there must be consensus that the object of investor pursuit is the next big thing. Three, returns on investment become unanchored from earnings. The market frenzy over AI satisfies all three criteria. Unsurprisingly, murmurs of a bust are getting louder. But this is where economics fails to provide any further guidance. It has no deterministic criterion to predict when a bubble will pop. The science, though, becomes more sanguine after a financial crash. It's been the same story since the tulip mania, possibly even earlier.

AI differs from all productivity-enhancing technologies that have preceded it in its economic impact. If it delivers, it will replace not just a section but an overwhelming majority of workers. Success here has severe economic consequences. Incomes would diverge uncontrollably between the AI-enabled and AI-deprived. Signals from the broader market are flashing orange as politics drives sound economics off the cliff. AI has concentrated the world's investment appetite on an unprecedented scale. Its failure, of course, would bring about the biggest wealth destruction in history.

There is enough economic literature on stages of a bubble, and investors will have to navigate the current one with a set of incomplete tools. The promise of AI itself is indeterminate, with lawmakers yet to decide how much of a free run the technology will be allowed. If they manage to tame the beast - by averting widespread job losses - the returns on current investment will be watered down. On its part, if AI has, indeed, been oversold, it could create the infrastructure for productivity enhancement through less disruptive technologies down the line. Investors must heed the Cassandras. But doom may not be nigh.