Warnings of a potential AI bubble have not exactly been thin on the ground lately. All the same, a blog post on August 17 by four senior economists at the European Central Bank (ECB) garnered extra attention, thanks to its source and prognosis.

“Extremely optimistic valuations raise questions: do today’s stock market prices reflect a rational bet on the transformative technology?” the blog post notes, adding that a replay of the turn-of-the-century dotcom bubble and bust is far more likely.

Concentration risk, always a danger regardless of the health of the concentrated sector, has reached alarming levels. As far back as October 2025, the so-called “Magnificent Seven” tech giants already accounted for over 35% of the total market cap of the S&P 500. Obviously, any specific upset for tech would impact the entire S&P 500.

The ECB economists conclude that the Magnificent Seven’s dominance of global indices is a worldwide structural danger — especially for households investing in index-tracking exchange-traded funds (ETFs). They estimate that European household investors in ETFs have some US$525 billion of exposure to US tech stocks. “A Magnificent Seven correction is a question of financial stability for the euro area,” they say.

Then there is the simple question of the inevitability of a correction. Remember the recent warning by Nicolai Tangen, CEO of Norges Bank Investment Management (NBIM)? He said “it is fairly likely” that the entire value of Norway’s $2 trillion Government Pension Fund Global (GPFG) could be lost under current conditions.

Even outside the AI bubble, valuations are looking very peaky. The ECB blog shows that the Cyclically Adjusted Price-to-Earnings (CAPE) ratio valuations on the S&P 500 Index, currently at some 41.5–42, are close to their all-time high of 44.2 — last seen in 1999 during the dotcom bubble.

Tangen also spoke broadly on the historical context of losses of value — that what goes up must come down. Associated risk analysis concluded that an AI industry correction could cost the GPFG 35% of its value.

Do professional investors seriously think that AI stocks will continue rising forever, up and up from their already stratospheric highs? I don’t believe most of them do. Rather, I suspect that they’re hanging on for as long as they can until the bubble bursts.

Prudent institutions, meanwhile, might do well to diversify out of AI-related stocks, do some profit-taking, assess and perhaps rebalance their actual exposure, and/or take out insurance by hedging against an AI bust. Other investors locked into passive index-tracking funds, like the retail investors cited by the ECB economists, perhaps should consider profit-taking too — if they can.

Read more @asiaasset