Central Banks & AI Watch: Material Inflection Detected

“Europe may import the cost of AI before it imports the productivity.”

The most consequential new development is Christine Lagarde's 14 September 2026 speech in Vienna, reinforced by a new BIS assessment released the same day. Together they push the central-bank debate one step further.

The issue is no longer simply whether AI will eventually raise productivity and lower unit costs. The new question is: who finances the AI Supercycle — and who captures the productivity? That distinction goes directly to the heart of the Great Split.

Lagarde disclosed that euro-area firms are now expected to devote roughly 10% of total investment to AI in 2026, while AI-related borrowing accounted for around one-quarter of the growth in euro-area corporate credit in Q1. AI use by euro-area workers has already risen above 50%, yet U.S. digital investment has been growing roughly twice as fast, and American workers spend two to three times as much of their working week using AI as workers in the largest euro-area economies.

THE NEW GREAT SPLIT: COST OF CAPITAL VS PRODUCTIVITY CAPTURE

Lagarde's most important macro-financial point is even sharper. U.S. hyperscalers issued more than $100 billion of bonds last year, mostly long-dated, and now represent close to 10% of new euro-denominated corporate bond issuance by non-financial companies. Market participants already attribute part of the rise in U.S. long-term real yields to AI-related borrowing. Because euro-area long-term rates are correlated with U.S. yields, Europe is importing part of the financing cost of the American AI buildout. That produces a striking chain of causality: U.S. AI capital formation drives hyperscaler bond issuance, which lifts global demand for capital, pushing up long-term real yields and, in turn, European borrowing costs. Simultaneously, U.S. AI investment accelerates adoption, raising productivity and cash flows, and widening America's economic advantage. Europe risks participating in the first chain without capturing enough of the second.

The New Chains of Causality

how AI is entering the monetary machinery

The fact that AI-related borrowing generated roughly a quarter of euro-area corporate credit growth in Q1 is especially important. It means aggregate credit data may increasingly conceal a widening internal divergence: AI-linked sectors show rising investment, credit demand and capital formation, while traditional sectors face higher borrowing costs, weaker investment economics, slower capital formation and employment pressure.

That is the financial version of the Great Split. A central bank can therefore face the paradox of maintaining restrictive financial conditions while a specific technology-driven investment boom continues accelerating. The same interest rate can be restrictive for the Old Economy and insufficiently restrictive for the AI Economy.

Signal Monetary transmission becomes materially harder to interpret when one rate hits two economies differently.
Lagarde also quantified the upside. ECB estimates suggest that rapid AI adoption could raise the level of European productivity by as much as 4% over a decade. With Europe's workforce projected to shrink by more than one million people annually over the next 25 years, she framed productivity growth not as an optional benefit but as necessary to sustain Europe's economic and social model.

The ECB increasingly confronts two forms of inflationary pressure — nominal inflation from energy, wages and geopolitical shocks, and structural scarcity inflation from capital, electricity, data centres, GPUs and the infrastructure the AI buildout requires. Against both sits the potential deflationary force of AI raising productivity and lowering unit costs. The sequencing will matter enormously.

Signal Industrial AI deflation must offset demographic decline, defence, energy, digital infrastructure and rising debt-service costs.
The BIS's new 14 September assessment says the AI-driven global market rally is beginning to show "growing signs of vulnerability," with particular concern around leverage, opaque financing and the durability of expected AI profitability. Aggregate tech-company borrowing in private credit has risen from around $22 billion in 2010 to more than $1 trillion by 2025, according to the BIS analysis cited by Reuters.

This builds on Pablo Hernández de Cos's warning that the five largest global technology companies are expected to invest more than $1 trillion in AI across 2025–26, while industry estimates imply global AI investment could eventually reach as much as $4 trillion by 2030. The concern is that capital expenditure is beginning to outrun cash generation, forcing greater reliance on debt, private credit and financing structures that can be opaque and interconnected.

Signal The next phase runs capex up, free cash flow short, and leverage and interconnection higher — pulling central banks in.

The BIS Warning Is More Serious Than "AI Bubble" Talk

the concern is transmission, not valuation

Europe's New AI Trilemma

adoption, sovereignty and capital

Adoption

Europe must adopt AI quickly enough to capture productivity — yet U.S. digital investment is growing twice as fast and American workers use AI two to three times as intensively.

Sovereignty

Europe cannot become completely dependent on external models and infrastructure, even as it hosts only around 5% of global AI computing capacity against roughly three-quarters in the United States.

Capital

Europe must finance enough domestic AI infrastructure to avoid exporting both its savings and its future growth. Euro-area households already hold roughly €440 billion in U.S. technology companies, while Europe's data-centre shortfall is projected to expand more than sixfold within a decade.

What This Means For Neutral Rates

a moving equilibrium

The Great Split is no longer purely an equity-market phenomenon — it is moving into credit.

The next differentiation should increasingly occur between AI cash-flow generators and AI capital consumers. The first group can self-finance. The second becomes progressively more exposed to long-duration yields, refinancing risk, private-credit pricing, power costs, data-centre utilization and the speed at which AI revenues materialize. The AI Supercycle therefore moves from technology selection toward capital-efficiency selection — mattering for equities, investment-grade credit, private credit, infrastructure and eventually bank exposures. For Europe, the policy hierarchy is becoming clearer. The objective cannot simply be "spend more on AI." Europe must close the circuit: European savings → European capital markets → European AI infrastructure → European models and industrial AI → productivity → cash flow → reinvestment. Lagarde's explicit answer is the Savings and Investments Union, no longer merely a financial-market reform but part of Europe's AI industrial policy. Otherwise the adverse loop holds — European savings finance American AI, American AI raises global capital demand, European borrowing costs rise, European investment slows and productivity divergence widens.

Key takeaways

the debate has genuinely changed

The question has shifted from whether AI is inflationary or deflationary to who captures the productivity and who carries the cost of capital.
AI now shapes credit growth, long-term yields, capital allocation and monetary transmission — not just office productivity.
The BIS has moved AI into the domain of leverage, private credit and systemic financial stability, citing tech private-credit borrowing above $1 trillion by 2025.
Europe risks importing the financing cost of the American buildout before it imports the productivity.
The Savings and Investments Union is becoming part of Europe's AI industrial policy.

Francesco De Leo Kaufmann · The Velocity Edge