European Central Bank president Christine Lagarde used her final address to the Frankfurt European Banking Congress on Friday to argue that public money should absorb the first losses on chip factories and data centers, so that private investors will fund the rest. She steps down as head of the ECB in February.

"Where we agree that a challenge requires a European solution, we should strengthen the European institutions that can provide it," she said. In technology, she added, projects of that size need "public-private partnerships at European level, with public money, including from the EU budget, taking the first losses and drawing in private capital."

The proposal is a specific answer to a problem she has raised before. Lagarde has warned that Europe lacks the data center capacity to meet its own demand, and that plans for a network of up to seven gigafactories fall short of what the bloc needs. Her estimate, cited in the speech, is that lifting Europe's share of global AI computing capacity from about 5% to 15% by 2030 would take around 1.3 trillion euros of investment.

Why public money has to go first

A first-loss guarantee is the oldest tool in the development bank kit, and it exists because private capital will not price certain risks on its own. A chip plant or hyperscale campus carries construction risk, technology risk and, in Europe, permitting risk that can run for years before a single rack is powered. If the public balance sheet takes the first slice of any loss, an investor's downside becomes calculable.

The politics have moved in the opposite direction. Opposition to data centers was a major factor for more than half of voters in the Nov. 3 midterm elections, according to CBS exit polling, and projects tied to the AI buildout have become harder to approve from Asia to the United States. European projects have already been delayed or shelved, which makes the case for public cover stronger and the case for spending public money on it weaker.

Lagarde acknowledged the tension directly. "As long as we can buy computing power from foreign providers and frontier AI models remain available to us, the case for paying that price is hard to make," she said. The argument for investment, in other words, weakens exactly when the investment is easiest to defer.

The capital markets problem underneath

The speech returned repeatedly to the unfinished capital markets union, the decades-long effort to make cross-border investment in Europe work like it does inside a single country. Lagarde said every year without an integrated market and supervisory system means "more of our savings are invested abroad and more of our growing firms look elsewhere for funding."

That is the reason a first-loss guarantee has to come from the EU budget rather than from private markets. Europe has the savings. It does not have the machinery to route them into a domestic chip plant at a price both sides accept, and it does not have a single supervisor to reassure a pension fund in one country about a construction site in another.

She pointed to the 800 billion euro pandemic recovery fund, worth about $877 billion, and to the bloc's support for Ukraine as moments when Europe acted at scale. Both had something the AI buildout lacks: a deadline and a visible emergency.

What happens after February

Lagarde leaves with the policy recommendation on the table and no authority to carry it out. The decision sits with the Commission, which is negotiating the next multi-year budget, and with member states that have spent the year arguing about joint borrowing.

Her warning was about timing. "By the time everyone can see the need to act, the best time to act has usually passed," she said. "We can see the problem early, but we can act only once it has become impossible to ignore. And by then, almost by definition, it is too late."

For investors, that is a forecast about European chip and data center equities rather than a spending plan. If a first-loss facility appears in the next budget, the projects that were waiting on it get a floor under their cost of capital. The alternative has been on display all year: demand grows, capacity does not, and the gap gets filled by imports.