ECB Warns AI Stock Valuations Depend on a Profits Boom That Has Yet to Arrive

The European Central Bank’s latest analysis presents the AI stock boom as a financial-stability risk to monitor, not a definitive bubble diagnosis. U.S. equity valuations measured by the cyclically adjusted price-to-earnings ratio are close to historical peaks, and the case for current prices increasingly depends on large future gains in revenue, margins and productivity.

That distinction matters. A market can be supported by real technological progress and still become vulnerable if investor expectations move faster than realised profits. The ECB’s concern is less that AI companies have no value than that too much of today’s valuation may already depend on an unusually optimistic outcome.

The ECB’s central warning: elevated valuations meet high expectations

The ECB’s Financial Stability Review materials show that U.S. equity valuations remain elevated relative to their longer-run history. One measure, the CAPE ratio, places the U.S. market close to its historical peak; euro-area valuations have also risen, although less sharply.

The ECB has also highlighted the role of the “Magnificent Seven”: Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. These companies have significantly outperformed the rest of the market, helping drive the broader U.S. equity rally.

This creates a concentration problem. When a small number of very large companies account for a substantial share of index performance, a change in expectations about those companies can affect portfolios that investors may view as diversified. The risk is not limited to specialist AI funds; the same firms are widely represented across major equity benchmarks.

Why the Magnificent Seven have amplified the rally

The seven companies are not identical businesses, but they sit at important points in the AI investment chain: cloud infrastructure, chips, software, advertising, platforms and consumer technology. Their scale means that strong earnings news can lift both their own share prices and the indices in which they carry large weights.

ECB data show how pronounced that leadership became. In one period covered by its November 2025 review, the Magnificent Seven rose 58% from 8 April, compared with 24% for the rest of the S&P 500.

That outperformance can reinforce itself. Rising prices increase the market value and index weight of the leaders, while strong results encourage investors to raise their forecasts. But the same mechanism works in reverse: if earnings expectations are cut, the effect can spread through index funds, institutional portfolios and other assets linked to U.S. equities.

How XLK can rally while its forward P/E falls

The apparent contradiction in the Technology Select Sector SPDR Fund, or XLK, is explained by the denominator in the valuation calculation.

Forward P/E = share price ÷ expected future earnings

XLK gained roughly 42% over the year in the supplied market comparison, while its forward price-to-earnings multiple fell by about 30% at a reported low. The fund also recorded its strongest 45-day surge in the available history cited by that analysis.

That can happen when analysts raise expected earnings faster than the share price rises. The comparison attributed the lower multiple to an approximately 80% increase in projected earnings.

The figures should not be treated as a single-period accounting identity. A 42% price increase alongside an 80% earnings increase would imply a smaller P/E decline—roughly 21%—if both measurements used exactly the same dates, constituents and estimates. The reported 30% decline therefore appears to reflect different measurement windows or portfolio composition. The underlying lesson remains valid: a falling forward multiple can reflect rapidly rising forecasts rather than a genuinely inexpensive stock market.

What must go right for these stocks to look “cheap”

A valuation based on forward earnings assumes that the forecasts will be achieved. For AI-linked companies, that means investors are implicitly counting on some combination of:

• sustained demand for chips, cloud services and AI software;
• large revenues from new AI products and services;
• durable profit margins despite intense competition;
• continued corporate spending on data-centre capacity; and
• attractive returns on the capital being invested in AI infrastructure.

Goldman Sachs estimates that AI-related companies have added roughly $27 trillion in market value since late 2022, equivalent to about 36% of the current U.S. equity market. It says the market value can be reconciled with future profits, but doing so requires more optimistic assumptions.

That is the key caveat behind the word “cheap.” A lower forward P/E does not establish that companies are inexpensive on current, realised earnings. It may instead mean that the market is pricing in a very large future earnings expansion.

How a correction can begin without an immediate price fall

Suppose the share price stays unchanged but analysts reduce their earnings forecasts. The forward P/E rises mechanically because the denominator has become smaller. Investors may then decide that the stock is too expensive at the revised outlook, leading to a price decline.

Several developments could produce that change in expectations:

1. AI capacity becomes oversupplied. Companies may build more computing infrastructure than customers ultimately need, weakening pricing power and returns on investment.
2. Customer spending slows. Businesses may delay or reduce AI projects if the productivity gains take longer to appear.
3. Competition compresses margins. New models, cheaper computing or aggressive rivals could make it harder for leading firms to convert demand into durable profits.
4. A wider economic shock hits technology spending. A slowdown, financial-market stress or weaker business confidence could reduce capital expenditure and risk appetite at the same time.

The ECB has previously pointed to the sensitivity of AI-linked valuations to changes in expectations. The market reaction following the release of DeepSeek-R1 was cited as an example of how quickly assumptions about AI capabilities and investment needs can affect valuations.

How the ECB warning compares with the Bank of England

The Bank of England has used more direct language in comparing some U.S. valuation measures with the dot-com peak. Its analysis said some U.S. index multiples were close to levels seen at the height of the dot-com bubble and estimated that AI stocks accounted for roughly 44% of S&P 500 market capitalisation in October 2025, up from about 26% in late 2022.

The BoE’s July 2026 Financial Stability Report also said AI-company valuations had grown faster than relevant broad equity indices and warned that high, rising concentration could magnify the effects of a revaluation.

The difference is mainly one of emphasis. The BoE stresses the potential size and transmission of a correction. The ECB’s framing focuses on whether current prices can be justified by future earnings and what happens if confidence in that earnings story weakens. Both point to the same vulnerability: a concentrated market can make an expectation shock larger than the initial disappointment suggests.

Why this remains a risk to monitor—not a confirmed bubble

The available evidence does not show that every AI-linked stock is in a bubble, or that a crash is inevitable. Many of the leading companies are established businesses with substantial revenues and profits, unlike the large number of unprofitable ventures that characterised parts of the dot-com boom.

But profitable companies can still be overvalued. The ECB’s warning is conditional: if share prices continue to outrun realised profits, or if the expected AI earnings boom fails to materialise, valuations could fall sharply. Because U.S. technology companies are widely held around the world, a repricing could affect U.S. indices, European markets and broader financial conditions.

The practical takeaway is to look beyond the headline multiple. Investors and analysts need to distinguish between earnings that have already been delivered and earnings that depend on years of uninterrupted AI adoption, infrastructure spending and margin strength. The rally may be rational in part—but its resilience depends on whether the profits eventually catch up with the expectations.