Why the AI Rally Could Trigger a Dot-Com-Style Correction

The European Central Bank’s warning is broader than a prediction that AI is overhyped. Its researchers argue that a correction in technology valuations is likely even if the underlying technology succeeds. The concern is that concentrated stock ownership, enormous infrastructure spending and a growing dependence on debt could make the adjustment economically significant.

That is why the debate is not simply whether AI is useful. It is whether investors, companies and policymakers are prepared for the financial consequences if expectations, risk premiums or funding conditions change.

The ECB’s core argument: success does not rule out a correction

The ECB’s analysis draws on earlier technological revolutions, including the dot-com era. Its conclusion is deliberately narrower than a forecast of an imminent crash: when a small group of companies becomes central to the economy, investors can eventually demand higher compensation for the risks that are difficult to diversify or insure.

That repricing can lower valuations even when the technology itself is genuinely transformative. In other words, AI can deliver real productivity gains and still produce disappointing returns for investors who paid too much for those gains in advance. The ECB describes a correction as a likely outcome under both an overly optimistic scenario and one in which current enthusiasm is fundamentally justified.

Why Europe is directly exposed

A U.S. technology sell-off would not stay confined to Wall Street. The ECB says euro-area households have about €440 billion of exposure to U.S. technology stocks, much of it held indirectly through investment funds and index trackers. Pension funds and insurers have a comparable level of exposure, bringing the combined figure to roughly €880 billion.

That concentration creates a direct wealth channel. A decline in the “Magnificent Seven”—Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla—would reduce the value of European households’ savings and institutional portfolios. It could also weaken confidence and encourage investors to reduce risk across other markets.

The ECB identifies a second channel: European equities could fall because the region’s own markets are vulnerable to a broader shift in risk appetite. A U.S. correction could therefore affect Europe both through existing holdings and through changes in global financing conditions.

The financing boom adds a second point of fragility

The AI buildout is also being financed on an unusually large scale. Goldman Sachs Research estimates nearly $500 billion of AI-related debt issuance in 2026, spanning investment-grade bonds, high-yield project finance and securitised credit.

The issue is not simply whether major technology companies can repay their borrowing. Many of the largest firms have strong businesses and cash flows. The more immediate concern is whether markets can absorb a continuing supply of long-dated debt without demanding materially higher yields.

Reuters reported that AI-related borrowing had reached close to 15% of investment-grade bond issuance by June, as banks developed new financing structures for data centres, chips and supply chains. Credit spreads for several large AI spenders also widened as investors sought greater compensation for the scale and uncertainty of the investment cycle.

That matters because a higher cost of debt can change the economics of the AI buildout before a company faces a conventional solvency problem. Projects may be delayed, capital budgets reduced and expected returns reassessed.

Why off-balance-sheet financing worries investors

Traditional corporate bonds are only part of the picture. The Bank for International Settlements says hyperscalers have increasingly used off-balance-sheet arrangements alongside conventional borrowing to finance infrastructure expansion. It also notes that credit-default-swap spreads rose, particularly for lower-rated hyperscalers, reflecting both the volume of new supply and uncertainty over project payoffs.

Other reporting has pointed to supplier arrangements, private-credit structures and related entities as additional ways to fund data centres and equipment. These arrangements do not automatically indicate misconduct or imminent distress. But they can make the overall network of obligations harder to evaluate, especially when several companies, lenders and infrastructure projects depend on the same assumptions about future AI demand.

The historical concern is familiar: risk can accumulate outside the most visible part of a company’s balance sheet, leaving investors with an incomplete view of leverage and interdependence.

Why this may not be another 2000-style collapse

The comparison with the dot-com bust has limits. Many leading AI companies today have substantial revenues, established products and significant cash flow, unlike numerous internet firms whose valuations depended on distant or unproven business models.

That difference supports a less extreme scenario: an extended deflation of valuations rather than a sudden systemic collapse. Prices could fall through lower earnings multiples, while earnings growth catches up with expectations. Companies could also reduce capital spending without failing.

Still, financially stronger companies are not immune to repricing. If expected growth slows or the return investors demand rises, even profitable businesses can lose considerable market value. The ECB’s warning is about that mechanism—not a claim that every major AI company is a dot-com-era shell.

Why policy support could be less powerful

The ECB also highlights a more limited policy cushion than investors had after the early-2000s technology bust. Interest rates entered the current cycle at higher levels, while governments face greater constraints on public finances. That could leave less room to respond to a confidence shock with rapid monetary easing or large fiscal support.

This does not mean policymakers would be unable to act. It means that the same tools could provide less immediate relief, particularly if a market correction occurred alongside inflation, geopolitical stress or concerns about government debt.

The feedback loop policymakers are watching

The most important risk is an interaction between equity markets and credit markets:

1. High AI valuations make aggressive infrastructure investment appear economically justified.
2. Companies borrow to expand data centres, computing capacity and related supply chains.
3. A disappointing earnings outlook, higher interest rates or rising risk premiums reduces stock valuations.
4. Lower equity prices and wider credit spreads increase the cost of financing.
5. Companies respond by cutting investment, which can further weaken growth expectations and validate the original sell-off.

The ECB’s concern is that this loop could move from a correction in a narrow group of technology stocks to weaker euro-area sentiment, tighter financing conditions, lower business investment and slower hiring.

What the warning does—and does not—say

The ECB blog does not provide a timetable, a predicted percentage decline or a claim that a crash is inevitable. Its conclusion is that a valuation correction should be expected at some point, even under a positive long-term view of AI.

The financing evidence points to a related but distinct risk. Nearly $500 billion in AI-related debt issuance, rising bond supply and the use of less transparent funding structures may increase the sensitivity of the buildout to changes in market appetite.

Taken together, the message is not that AI is a worthless bubble. It is that technological success, stretched expectations and heavy financing can coexist—and that a market adjustment could still spread well beyond the companies at its centre.