AI stocks will correct even if the boom is real
Five European Central Bank economists just argued that history does not care whether today's AI valuations are a bubble or a fair bet — either way, a pullback is the base case.
A Monday ECB Blog post concludes that a correction in current stock-market valuations is likely, even if prices fairly reflect how much AI will remake profits. The authors — Malin Andersson, Johannes Breckenfelder, Stefano Corradin, Kalin Nikolov, and Maria Antonietta Viola — are writing in their own names, not as a Governing Council forecast. That does not make the mechanism any softer. US cyclically adjusted price-to-earnings ratios sit near their historical peak. Euro-area multiples have risen too, just less. Nvidia's share price is up twenty-fold since 2022. The question they pose is not "is AI real?" It is what happens to the price of a real thing once the whole economy is riding it.
They walk through two stories that end in the same place. The rational one says high multiples can be justified by option value: if the technology works, the upside is hard to bound, so early winners trade rich. Then adoption spreads. Uncertainty stops being a single-sector bet you can diversify away and becomes an economy-wide risk. Investors demand a higher risk premium. Historically, that premium tends to beat even rising profits, and prices fall. The behavioral story is simpler: overconfident buyers push prices past fundamentals, then lose their nerve. Railways, 1920s electricity and radio, and the 1990s internet all followed some version of this arc. The authors are explicit that they cannot time the turn, and that a correction is not a ceiling — if AI is transformative enough, prices can still be much higher later. They are also explicit that you should prepare anyway.
The euro-area reason to care is not a home-grown Nvidia. It is the index fund. Households in the currency bloc now hold about €440 billion of US technology equities, mostly through cheap ETFs and mutual funds, often without seeing the Magnificent Seven concentration underneath — Alphabet, Amazon, Apple, Tesla, Meta, Microsoft, and Nvidia. Insurers and pension funds sit on the same stack. A sharp drop can force funds to sell liquid holdings to meet redemptions, then distressed ones, which is how a US tech drawdown becomes a euro-area financial-stability problem. Unlike the dot-com crash, they note, there is markedly less room to cut rates or spend fiscally to cushion the fall. Local markets look less frothy — more old-economy, lower multiples, a digital-investment boom that has still outgrown GDP by more than three times over the past decade — but US and euro-area equities have historically moved together.
This is the price-side twin of the lock-in we already traced on the cash-flow side — The $3T isn't Enron. It's a no-exit clause. The contracts keep the build-out from stopping cleanly. The risk premium is how markets eventually charge for that. Those clocks do not have to ring on the same day.
What to watch: whether euro-area supervisors start treating Magnificent Seven weight in household ETFs as a stability issue, not just a portfolio-choice issue.
If AI really does remake the economy, should a correction still be the base case — or is this just the last cycle's playbook taped onto a new one? Tell us in the comments.
Sources: ECB Blog · CNBC · Pástor and Veronesi, American Economic Review
