The rise of artificial intelligence has driven valuations in much of the US tech sector to levels reminiscent of the dot-com era. Five economists linked to the European Central Bank (ECB) believe that a correction in current stock market valuations is likely even if AI fulfills the huge expectations placed upon it. This warning is especially relevant for Europe: eurozone households have around €440 billion of exposure to US technology companies.
The key points of the AI stock market boom in 20 seconds
- US market valuations are near all-time highs according to the CAPE ratio.
- The authors consider a correction probable, even if AI growth justifies current optimism.
- Nvidia exemplifies the boom: its stock price has increased about 20-fold since 2022.
- Households in the eurozone have approximately €440 billion exposed to US tech companies.
- While valuations in Europe are lower, it would be hard to remain insulated from a US downturn.
The analysis, published on August 17, 2026, on the ECB blog, is authored by Malin Andersson, Johannes Breckenfelder, Stefano Corradin, Kalin Nikolov, and Maria Antonietta Viola. A caveat: their opinions do not necessarily represent the official position of the ECB or the Eurosystem.
Their question is particularly interesting for the tech sector. AI could become one of the most important technologies in the coming decades and, at the same time, trigger a market correction among the companies leading its development.
These scenarios are not mutually exclusive.
Nvidia shows how high market expectations for AI have risen
Artificial intelligence has transformed the structure of the tech industry in just a few years. GPUs, HBM memory, data centers, high-speed networks, and electrical capacity have become strategic resources as major cloud providers allocate increasing capital to infrastructure.
Nvidia exemplifies this transformation better than any other company.
Economists note that its stock price has multiplied roughly 20 times since 2022, coinciding with the commercial explosion of generative AI and the rising demand for accelerators for training and inference.
The question isn’t whether Nvidia is actually selling more chips or if AI is creating demand—both are observable. The financial concern is how much future growth is already priced into the stock.
They use the CAPE ratio for this, which compares the stock price to an average of ten years of inflation-adjusted real earnings.
The indicator for the US market is near its all-time highs and close to levels seen during the dot-com bubble. In Europe, it has also increased but remains significantly lower.
Caution is needed when comparing to Internet era valuations. High valuations in two periods don’t necessarily mean they will end similarly.
In fact, the analysis suggests something more interesting: a bubble isn’t even necessary for a correction to occur.
AI can meet expectations and stocks still fall
Major technological revolutions often have complicated relationships with financial markets.
The railway in the 19th century, electricity and radio in the early 20th century, and Internet in the 1990s all generated enormous growth expectations. These were real technologies that eventually deeply transformed the economy.
This didn’t prevent cycles of sharp rises followed by falls.
Economists offer two explanations.
The first is rooted in rational markets. When a new technology emerges, it’s extremely difficult to estimate its economic impact initially. Early investors are buying, in a sense, a call option on a potentially huge future.
A company may fail, but another could become the next tech giant. This possibility justifies high valuations.
The situation changes once the technology becomes widespread across the economy.
Uncertainty no longer concentrates in a few firms but becomes a broader risk. Investors may then demand a higher risk premium and be willing to pay less for future profits.
Stock prices could fall even if profits continue to rise.
The second explanation aligns more with the traditional bubble concept: investors may overestimate a technology’s potential, pushing stocks higher than justified by actual business results.
In both scenarios, a correction is possible.
That’s why the authors avoid categorizing the AI market as rational or irrational. They believe past technological revolutions make it reasonable to prepare for a correction at some point, though they can’t predict exactly when it will happen.
The Magnificent Seven hold much of the risk
AI development isn’t evenly distributed across all tech companies.
A significant portion of capital is concentrated among the so-called Magnificent Seven: Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla.
Several of these hold key positions within AI infrastructure.
Microsoft, Alphabet, and Amazon operate some of the largest cloud platforms worldwide and are investing heavily in data centers. Meta is developing its own models and building extensive training infrastructures. Nvidia supplies much of the accelerators used by these companies.
This concentration also reflects in stock indices.
When investors buy certain global funds or index ETFs, a considerable part of their assets are indirectly invested in these giants due to their large market caps.
This creates a link with Europe.
Europe has €440 billion exposed to US tech stocks
Based on the analysis, households in the eurozone have approximately €440 billion of exposure to US tech stocks.
Much of this isn’t through direct purchases.
Investment funds are the main direct owners of shares in the Magnificent Seven among the European investors analyzed. When examining fund holdings, households appear as the largest group with indirect exposure.
There are also relevant holdings by insurers and pension funds.
The rise of cheap ETFs and index-replicating products has allowed millions of investors to gain exposure to US tech growth. At the same time, the enormous size of a few companies can create a concentration that isn’t always apparent to investors buying global or diversified products.
A decline in these firms would have effects beyond individual losses.
If a correction triggers significant redemptions, funds may be forced to sell assets to meet withdrawals. These sales could exert downward pressure on prices and lead to more redemptions.
The authors consider that a severe drop in the Magnificent Seven could pose a financial stability risk for the eurozone.
Europe isn’t as enthusiastic about AI in the stock markets
The European tech sector’s situation is different.
Valuations in the eurozone have increased but remain well below US levels. European markets are still largely dominated by traditional sectors.
Economists also don’t see a behavior in the European ICT sector akin to the dot-com bubble anymore.
This doesn’t mean Europe is on the sidelines of AI development.
Business adoption is rising, and digital investment over the last decade has grown more than three times faster than the eurozone’s GDP, based on the data used in this analysis.
While Europe’s AI market may be experiencing a less intense boom in stocks, it’s no shield from Wall Street’s movements.
US and European markets have historically shown high correlation. A significant correction in US tech stocks could spill over into European markets, affecting confidence, financing conditions, and corporate hiring decisions.
The real risk may lie in everything built around AI
The ECB analysis also comes at a particularly interesting time for the industry.
The competition for AI isn’t just about developing better models. Major tech firms are building data centers, buying accelerators, contracting electrical capacity, and deploying networks to support the expected continued growth in computing demand.
Market valuations of these companies depend, among other things, on whether these investments generate enough future revenue and profit.
That will be one of the main tests for the current AI cycle.
A slowdown in growth wouldn’t necessarily mean the technology has failed. It might just mean that profits take longer than expected to justify investments
Internet provides a useful analogy.
The dot-com bubble destroyed enormous amounts of stock market value, but the internet didn’t disappear. Instead, it became the infrastructure underpinning much of today’s digital economy.
Similarly, AI might follow a paradoxical path: the technology could prove even more important than anticipated, even as companies linked to it experience a significant correction.
The five economists don’t know when or from what level the correction might start. They also don’t assert that current valuations are irrational.
Their warning is more uncomfortable for an industry used to debating if there’s a bubble or not: the correction could happen regardless of that question’s answer.
FAQs
Do ECB economists believe AI is a bubble?
Not necessarily. Their analysis considers both the possibility that valuations are rational and that there’s excessive optimism. In both cases, they consider a correction likely at some point.
Why is Nvidia included in the analysis?
The authors cite Nvidia as an example of the sharp rise in expectations around AI. They note that its stock has roughly multiplied by 20 since 2022.
How much European money is exposed to US tech companies?
Eurozone households hold about €440 billion of exposure, mainly through investment funds and ETFs.
Would a market correction mean AI has failed?
No. The analysis reminds that previous technological revolutions experienced boom and bust cycles but ultimately transformed the economy.
Sources: Portal Financiero and European Central Bank
