440 Billion Euros, No Votes

Ask whether it is a bubble and you have taken the bait. The question has an answer, the answer is worth having, and it keeps the argument on when to sell rather than on who decides. Five economists at the European Central Bank published the number that started this round: Malin Andersson, Johannes Breckenfelder, Stefano Corradin, Kalin Nikolov and Maria Antonietta Viola, writing on the ECB blog on 17 August 2026, put euro area household exposure to US technology equities at around €440 billion. It is a signed blog post rather than an official position of the Bank, and the distinction will be gone within a week.

Index construction picked those companies, and the household picked the index. The MSCI World Index held 1,282 companies from 23 countries on 31 July 2026, with the United States at 72.03% and the top ten positions at 26.41% of the whole index. Eight of those ten are the US AI complex: Nvidia at 5.18%, Apple at 5.07%, Microsoft at 3.66%, Amazon at 2.94%, Alphabet at 2.32% and 1.84% across its two share classes, Broadcom at 1.96%, Meta at 1.37%. A fund that calls itself global puts one euro in four into eight US companies. On the S&P 500 the same eight sit at 35.8%.

An index fund looks like diversification the way a monoculture looks like a forest. A single-species plantation is planted for yield and for ease of management, never against risk, and it keeps the shape of a forest right up to the season one pest takes it. Nothing here is concealed: the concentration is in the prospectus, in the monthly factsheet, and in the rebalancing rule that sends fresh money toward whatever has just grown.

A share is not a say. Meta's filings set out the mechanism without euphemism: Class A carries one vote, Class B carries ten, and the founder holds roughly 14% of the economic interest against roughly 61% of the voting power. The tallies show what that does. Two shareholder proposals in 2025 carried 82.4% and 65.8% among Class A holders alone and were defeated, and four proposals in 2024 stood in the same position and were defeated. A savings plan opened in Hamburg or in Bologna was inside those votes, and it had lost before it voted.

The vote leaves the saver one step earlier than that. European fund units reach the annual meeting in the hands of a small number of asset managers, most of them US-domiciled, voting under stewardship policies that are published, contestable, and not the saver's. The European household bought the return and the risk. The one part of a share that resembles power left its hands twice, and nobody asked either time.

European savings finance US capacity and buy no voice inside it. Around €300 billion a year of European savings has been leaving for foreign markets, most of it for US ones, out of a private pool on the order of €33 trillion: the figure is the one Letta, Draghi and Macron used to argue for the Savings and Investments Union. The €440 billion is a withdrawal from a flow with a settled direction. Money that stayed would run into the other half of the problem, because capital markets supply 14% of European non-financial corporate funding against 36% in the United States. Europe supplies the capital, the United States keeps the command, and the second was never on sale with the first.

This is a sovereignty question wearing financial clothes, and the costume earns its keep: it starts arguments about the exit and closes them before they reach the ownership.

The honest summary of the risk is short. Announced 2026 capital spending by the large compute operators has passed $700 billion, Amazon's trailing free cash flow fell 95% to $1.2 billion over the same stretch, and US valuations measured on a decade of inflation-adjusted earnings sit near their historical peak. The ECB authors argue that a correction is likely even if today's prices are perfectly rational: as uncertainty about AI productivity spreads across the whole economy it stops being diversifiable, and risk that cannot be diversified is paid for in lower prices. If the enthusiasm is not rational, the correction is sharper. Redemptions force funds to sell, selling lowers prices, lower prices bring further redemptions, and central banks and treasuries have far less room now than they had twenty-five years ago. Nobody knows the date. Anyone selling you the date is selling you the product.

A technology that genuinely reorganises an economy ends as infrastructure, which means rules, duties toward the people who depend on it, and an owner who answers when it fails. Grant the premise and it argues for rules and for ownership. It does not argue for sitting still.

Which brings back the question the ECB authors have already answered. Euro area households hold that €440 billion "without necessarily being aware of the associated concentration risk". Read that as a description of a product working correctly rather than as a complaint about savers: it collects €125 on the first of the month, places an eighth of it in eight companies on another continent, and was never obliged to tell you in those words, because in other words it did tell you.

A share is not a say, and the day one of those systems decides something on someone's behalf, your unit does not sign. Someone else signs, in another jurisdiction, holding a class of share you are not allowed to buy. Who signs? Not you. And so far nobody has put the question to you.