ECB economists warns AI boom could still end in a correction
Written by Rebecca Uffindell Wed 19 Aug 2026
The current AI-driven stock market boom could still end in a correction, even if investor enthusiasm proves rational and the technology delivers on expectations, according to analysis published by five European Central Bank economists.
US stock market valuations are now close to historical peaks, while enthusiasm around artificial intelligence (AI) has contributed to a sharp rally in technology stocks. The authors argued that previous technological revolutions showed transformative technologies could succeed while the valuations attached to them still underwent significant corrections.
The analysis was published on the ECB Blog by economists Malin Andersson, Johannes Breckenfelder, Stefano Corradin, Kalin Nikolov, and Maria Antonietta Viola. The ECB said that views expressed in its blog belong to the authors and do not necessarily represent the ECB or Eurosystem.
AI Valuations Approached Historical Extremes
The economists said US stock market valuations, measured using the cyclically adjusted price-to-earnings ratio, were close to their historical peak.
“We argue that economic research on past technological revolutions points to a worrisome conclusion: a correction of current stock market valuations is likely,” said the authors.
That conclusion did not depend on current AI valuations being irrational.
Instead, the economists pointed to previous technological movements including railways, electricity, radio, and the internet, where genuinely transformative technologies attracted investment and pushed company valuations higher before markets subsequently corrected.
A Correction Would Not Require AI to Fail
The distinction between technological success and investment returns was central to the analysis.
Under what the authors described as the rational explanation, investors can legitimately assign high valuations to companies exposed to a new technology because its future productivity benefits are both uncertain and potentially very large.
As adoption spreads across the economy, however, the nature of that risk changes. Initially, failure can be concentrated within individual companies or sectors and diversified across an investment portfolio. Once a technology becomes more deeply embedded across the economy, the associated uncertainty becomes harder to diversify. Investors may consequently demand a higher risk premium, putting downward pressure on valuations even if underlying profits continue to grow.
The behavioural explanation reaches a similar destination through a different route: excessive optimism pushes prices beyond fundamentals before confidence eventually weakens.
Both scenarios, the authors argued, point towards a correction at some stage. They also stressed that its timing is impossible to predict and that current valuations could rise considerably further.
“This does not mean that today’s prices represent a ceiling,” they added.
The argument was therefore more nuanced than a warning that AI is simply another dot-com bubble. AI could prove highly transformative while financial markets still experience a substantial repricing along the way.
Europe Has £376bn of Exposure to US Technology Equities
The potential consequences matter for Europe because euro area investors have significant exposure to the US technology sector.
The economists estimated that euro area households held around £376.4 billion (€440 billion) in exposure to US technology equities, with much of that exposure coming indirectly through mutual funds and exchange-traded funds rather than direct share ownership.
Insurance companies and pension funds also held significant exposure to the Magnificent Seven: Alphabet, Amazon, Apple, Tesla, Meta Platforms, Microsoft, and NVIDIA.
The structure of that exposure could itself become a transmission mechanism during a downturn.
The authors said a sharp correction could prompt investors to redeem money from funds, forcing those funds to sell assets. If the downturn persisted, further sales could push valuations lower and trigger additional redemptions.
That is why they characterised a potential Magnificent Seven correction as a question of financial stability for the euro area, rather than simply an investment loss for individual shareholders.
Europe Faced a Different Valuation Picture
The economists did not identify the same degree of exuberance within European equity markets.
Euro area valuations had increased, but remained considerably below US levels. The authors also found the region’s information and communications technology sector more resilient than during the dot-com period, pointing to rising productivity and markups, increased AI adoption, and substantial growth in digital investment.
They described Europe’s AI transformation as proceeding at a “steady if unspectacular pace.”
That reduced the likelihood of a domestically generated correction on the scale potentially facing US technology stocks. It did not, however, insulate Europe from events across the Atlantic. Euro area and US equity markets have historically been highly correlated, while European investors’ direct and indirect exposure provides another channel through which a correction could spread.
“A US AI fallout would not remain a US problem,” the authors warned.
They said the consequences could extend beyond investment losses into sentiment, financing conditions and hiring across the euro area.
Anthropic Showed How Far AI Valuations Are Reaching
The analysis arrived as investors were already being asked to look further into the future when assessing some of the largest AI companies.
This week it was reported that bankers and prospective investors considering Anthropic’s IPO valuation were using a 2028 revenue forecast of roughly £139 billion ($190 billion) to £146.9 billion ($200 billion) as part of the pricing process.
Anthropic’s revenue run rate had crossed £34.5 billion ($47 billion) by May, meaning prospective investors were considering a future revenue level several times larger than its current pace of business.
That does not make Anthropic evidence of an AI bubble. It does, however, illustrate the valuation challenge identified by the ECB economists: expectations about AI’s future economic impact are requiring investors to place considerable weight on growth that has yet to materialise.
AI’s Valuation Test
The ECB economists’ argument separated two questions that are often treated as one.
AI may prove genuinely transformative. Companies deploying it may generate substantial productivity gains and profits. Neither outcome necessarily means today’s valuations will continue rising without interruption.
Historical technological revolutions suggest markets can correctly identify a technology capable of reshaping the economy while still struggling to price the timing, distribution, and scale of the returns it eventually produces.
For Europe, that uncertainty is no longer confined to Wall Street. With households, insurers, and pension funds exposed to US technology equities, a significant repricing of the AI boom could travel through European portfolios and into the wider economy.
The issue is not just whether enthusiasm for AI is justified, but how financial markets adapt as expectations about the technology align with economic reality.
Written by Rebecca Uffindell Wed 19 Aug 2026