The ECB Incorporates AI into Its Credit Outlook

The Precedent: 1999, the Fed, and Productivity as Foundation

In 1999, Alan Greenspan's Federal Reserve incorporated productivity growth driven by information technology into its monetary policy framework. The mechanism was clear: private capex as a structural premise.

The central bank maintained accommodative conditions because the boom in network and semiconductor investments promised sustained growth. The market dictated the investment cycle, and monetary policy ratified it.

Then came 2001. Telecommunications capex collapsed, recession followed, and the productivity anchor revealed itself as a market hypothesis dressed up as monetary fundamentals. Today's context differs; the structure remains identical.

The Current Pattern: The ECB Places AI in the Baseline

ECB staff projections from September 2026 list investments related to artificial intelligence alongside defense and infrastructure as explicit drivers of medium-term growth. The ECB and AI thus enter the same planning framework.

The document sets real GDP growth at 0.9% in 2026, 1.4% in 2027, and 1.5% in 2028, with AI spending supporting the later years of the horizon, while inflation peaks at 3.6% in the fourth quarter of 2026, according to official projections^[1].

This is the signal. A G7 central bank formalizes AI capex within its medium-term anchor.

The Mechanism: Whoever Finances Data Centers Dictates the Cycle

Here is the point the market has yet to grasp. The ECB recognizes AI as a structural engine of credit demand for the next 36 months, while the supervisory tools to govern it remain beyond its reach.

AI capex is a debt loop. Approximately one trillion dollars in assets are engaged between chips, laboratories, and new cloud infrastructure, and roughly $115 billion in private credit price software revenues that AI has yet to generate.

The central bank thus inherits an investment cycle designed by private markets. It ratifies demand while ignoring financing structure. The 1999 precedent shows where this asymmetry leads.

This Transcends the Cycle. It Is a Regime Change

The classical function of a central bank involved guiding the credit cycle through the reference rate. The global financial system ceded that control to private markets, and the ECB now puts it on record.

Three precedents are sufficient to call it a pattern: the Fed in 1999, the Bank of Japan in 1989, the ECB in 2026. In each case the central bank embraced a private capex boom as fundamental and lost the tools to restrain it.

The BIS Quarterly Review has documented for quarters the convergence between private credit and technology capex. The IMF signals the same risk. The danger lies with whoever finances the hardware, distant from equity valuations.

From Geopolitical to Systemic

For years AI risk lived on the geopolitical map: export controls, semiconductor sovereignty, the chokepoint of TSMC and Samsung factories. Finance remained the gap in American strategy, with Hong Kong financing laboratories under sanction.

That gap now migrates into the balance sheets of European central banks. The credit demand that fuels AI capex becomes a variable of monetary policy, and its origin remains outside the ECB's supervisory perimeter.

The dollar's share of reserves fell from 71% in 2001 to 58% in 2024, with 244 tonnes of gold purchased in a single quarter. Capital is repositioning around structures, and central banks chase the cycle they surrendered.

The Central Bank That Forecasts with AI

There is a second layer. The ECB admits that R-star, the natural rate, escapes its models, while LLM agents with web search match the Fed's nowcast on sixteen indicators.

Whoever controls compute will write the rules of supervision. A central bank that adopts AI in forecasting before resolving the regime change cedes twice: on the credit cycle and on informational advantage.

My Position, and What Would Refute It

My thesis is clear. By incorporating AI capex into its medium-term anchor, the ECB has transformed a geopolitical risk into a systemic one, and lacks the supervisory apparatus to contain it.

What would change my reading? An ECB supervisory framework that treats AI credit exposure as a distinct risk category, with dedicated capital requirements, by 2027. That step would demonstrate the institution closed the gap between recognition and instrument.

Until then I read the September move as passive ratification. The central bank describes an engine that others control.

Three Implications for Capital

For family offices and sovereign wealth funds: sensible reallocation over the next 36 months reduces exposure to private credit financing data centers and favors real assets in the bottleneck sectors, energy and semiconductor fabrication. Horizon: 36 months.

For the chief risk officer: the scenario of AI credit repricing is missing from VAR models calibrated on historical variance. It must be inserted as a structural shock, with an 18-month horizon, because the correlation between technology debt and the European cycle is rising.

For CFOs and investor relations: the narrative of European growth driven by AI, which you carry to investors today, risks appearing fragile within 18 months. The European AI goods sector remains small, and the ECB itself admits it.

The Forecast

ECB staff projections from June 2027 will confirm AI capex as an explicit driver of medium-term growth, keeping it within the 2027-2029 baseline. Confidence: 70%. Horizon: June 2027.

The signal that would refute this reading: June 2027 projections remove AI from the list of structural growth drivers.

What to Watch

Three indicators will guide verification in the months ahead.

• AI capex spending within ECB quarterly projections.
• The volume of private credit allocated to European data centers.
• Any ECB supervisory framework on AI credit risk.

The divergence between recognizing the driver and the instrument to govern it always resolves. The question remains how.