Is an AI Crash on the Horizon? The ECB Compares AI Euphoria to the Internet Bubble

The Next Internet Bubble?

In the United States, stock valuations are nearing their historical record. Measured by the CAPE ratio, which compares stock prices to corporate earnings averaged over 10 years, they have reached a level last seen during the Internet bubble of the 2000s.

According to the authors, history often repeats itself. Railroads in the 19th century, electricity in the 1920s, the Internet in the 1990s: with each technological revolution, the stock prices of the involved companies soar before plummeting sharply.

From this parallel, they draw a straightforward conclusion: a correction of current stock valuations is likely. This is true whether today's prices are rational or not.

Why a Correction is Likely, Even if AI Delivers on Its Promises

The most counterintuitive aspect of the reasoning from the ECB economists is that the stock market could fall even if AI genuinely transforms the economy and boosts profits.

The explanation lies in the very nature of risk. As long as AI remains a limited experiment, a failure only affects a handful of companies, and the risk is diluted. But as the technology spreads everywhere, it becomes a risk for the entire economy and can no longer be diversified.

To accept this risk, investors demand a higher return, thus a lower purchase price. Mechanically, valuations then decline. Additionally, there is a psychological spring: when euphoria fades, the fall often goes further than fundamentals would justify.

The ECB also acknowledges a significant unknown (obviously): the exact timing of a potential bursting of the AI bubble is impossible to predict in advance.

€440 Billion: The Hidden Exposure of Europeans

Beyond mere stock commentary, households in the eurozone hold about €440 billion in American tech stocks, often without being aware of the concentration risk.

This exposure mainly comes through low-cost ETFs and global index funds like MSCI World, where the Magnificent Seven (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, Tesla) carry significant weight. Insurers and pension funds are also heavily exposed.

Unlike the Internet bubble, the authors believe that the room for maneuver today is narrower: less room to lower rates, less budgetary latitude to cushion the shock.

Even if European stock valuations remain more reasonable, the continent's markets historically closely follow the performance of Wall Street. As the authors conclude, an American accident would not remain an American problem.

For long-term investors, the challenge is not to guess the peak of the AI sector, but to avoid betting everything on a handful of tech stocks and to check what their ETFs actually contain. In the event of a sudden crash, some simple reflexes are also necessary, such as placing stop-loss orders on their positions.

Investing solely in the AI sector also carries a risk: even without a bubble, the sector could go through tough periods, which would weigh heavily on portfolios exclusively exposed to this theme.