The Capital That Leaves, the Risks That Return

“The next Great Split will be decided not only by who controls the intelligence, but by who controls the capital.”

Three consequential central-bank developments on November 2 reveal how the AI Supercycle is transforming the international financial system.

ECB Vice-President Boris Vujčić has identified elevated AI equity valuations, sovereign bond-market stress and potential non-bank losses as three major threats to financial stability. In a Dublin speech he showed that European capital markets are integrating with the United States faster than with one another — European savings increasingly finance American companies while Europe's fragmented markets remain less able to absorb shocks. And the Hong Kong Monetary Authority has launched Project SAGE to govern autonomous AI agents in finance.

The emerging contradiction is fundamental: Europe is increasingly exposed to the financial risks of the global AI investment cycle without an equally integrated financial system capable of financing its own productivity transformation. This is more than a competitiveness problem. It is becoming a structural financial-stability risk.

1. 01 AI Valuations — A Formal Financial-Stability Concern

   The ECB identifies elevated AI equity valuations alongside sovereign bond-market stress and potential non-bank losses among the principal current vulnerabilities. signal: AI valuations are now a named systemic risk.

2. 02 25 Versus 12 Percentage Points

   Since 2014, estimated cross-border equity-investment frictions between the euro area and the United States have declined by ~25 percentage points, versus ~12 points within the euro area. signal: Europe is integrating with the US faster than with itself.

3. 03 20% Greater Investment Contraction

   European companies with predominantly domestic shareholders reduced investment ~20% more than internationally financed peers during the COVID-19 shock. signal: Fragmented financing amplifies every shock.

4. 04 92% — Bank of England Tightening Probability

   Markets price approximately a 92% probability of a 25-basis-point increase on November 5, per LSEG data reported by Reuters. signal: Energy inflation is still forcing tightening.

5. 05 Project SAGE — Agentic Financial Supervision

   Hong Kong's monetary authority has launched a governance initiative for autonomous AI agents, accompanied by a formal industry discussion paper. signal: Supervision shifts to the autonomous systems themselves.

Speaking at the Central Bank of Ireland's Financial System Conference, Vujčić identified three interconnected concerns: elevated AI-related equity valuations, vulnerabilities in sovereign bond markets and potential losses among non-bank intermediaries. The ECB's next Financial Stability Review is scheduled for November 18.

The significance lies in the interaction. A correction in AI valuations could affect more than technology shareholders, because the financing architecture of the AI Supercycle connects technology companies, infrastructure investors, credit funds, banks, insurers and other institutional investors. The transmission runs: AI expectations repricing → equity losses → credit-spread widening → non-bank deleveraging → liquidity pressure → tighter financing conditions.

Sovereign debt adds a second channel: if government bond markets reprice simultaneously, institutions may face losses on both risky and traditionally defensive assets, restricting lending even where direct AI exposure is modest. This is a plausible stress scenario, not a prediction of imminent crisis.

The central concern is that a valuation adjustment could become a capital-allocation shock — institutions facing declining asset values may cut lending and investment across unrelated sectors, impairing financing for companies that never benefited from the boom. For investors, the priority is to evaluate exposures across equity, credit, private funds and sovereign debt simultaneously, not as independent risks.

Vujčić's speech, Integrated, Open and Resilient: Europe's Strengths in a Fragmenting World, provides the strongest structural evidence. ECB estimates show barriers to cross-border equity investment have declined far faster between Europe and the United States than among euro-area countries themselves, enabling rising international investment by European savers.

The result is an extraordinary asymmetry: European savings increasingly participate in American capital formation while European businesses remain dependent on domestically fragmented markets. The gap is sharpest in venture capital — US venture funds invest approximately six times as much as EU-based funds.

Financial integration also determines shock absorption. ECB research cited today suggests ~70% of state-level output shocks in the US are absorbed through markets and fiscal transfers; the euro-area figure is ~30%, with capital markets alone absorbing only 5–10%. When financing is primarily domestic, a shock disproportionately reduces investable capital.

Vujčić adds that European firms tend to cut investment when oil prices rise, whereas comparable US evidence shows no such systematic reduction — making Europe's investment cycle more vulnerable to energy shocks precisely when technological transformation demands sustained capex.

A separate ECB Research Bulletin published today, Monetary Policy in Times of High Public Debt by Stéphane Dupraz and Anna Rogantini Picco, examines how elevated public debt affects monetary transmission and the natural rate. Its central conclusion: high public debt does not necessarily eliminate central banks' ability to control inflation, but the stock of debt can raise the equilibrium interest rate consistent with price stability. Tightening can also become less effective when higher interest payments expand household income. This is research, not a change in the Governing Council's reaction function.

The interaction with the AI Supercycle is consequential. AI capital formation may raise equilibrium real rates through stronger investment demand; elevated public debt may add further upward pressure; energy shocks can keep inflation elevated. Yet Europe may fail to capture the corresponding productivity expansion if financing conditions constrain investment — producing higher equilibrium financing costs without equivalent productivity growth. The research does not establish such a divergence, but strengthens the case for monitoring debt, productive investment and monetary transmission together.

The next test arrives on November 5. The Bank of England is widely expected to raise its benchmark rate from 3.75% to 4.00%. Reuters reports September CPI reached 3.5%, exceeding the Bank's earlier 3.0% projection; markets assign ~92% probability to a quarter-point increase. Goldman Sachs has revised its forecast to include a further 25bp increase in February 2027, though the central bank has made no such commitment.

The Bank confronts the same asymmetry emerging elsewhere: energy inflation raises production costs now, while AI adoption offers only uncertain future productivity gains. Monetary restraint may become necessary before technological disinflation can materialise — the inflationary costs of building an AI-intensive economy can arrive before its productivity dividend. The challenge is to prevent temporary energy shocks from becoming persistent inflation without needlessly constraining investment in future productive capacity.

The HKMA has launched Project SAGE — Securing Agentic Finance with Governance and Enablement — to support exploration of autonomous financial applications alongside appropriate governance standards. It has established a cross-sector expert group and published a discussion paper on supervisory considerations for banks deploying AI agents, raising questions of agent accountability, registries of trusted agents and a possible Know Your Agent framework. These remain subjects for consultation, not binding requirements.

Traditional supervision focuses on institutions making decisions; agentic AI introduces systems that execute financial tasks with varying autonomy. The critical distinction is between automating information processing and delegating the execution of decisions — the latter requires stronger controls over authorisation, intervention, accountability and system integrity. Supervision is beginning to move from governing institutions that use AI toward governing the autonomous systems through which they act. For European banks, the HKMA offers an important reference point.

Five Takeaways for Investors and Policymakers

the next financial rotation

• AI valuations — Monitor transmission from equity markets into credit, assessing valuations alongside refinancing exposure, private-credit concentration and institutional liquidity.
• European capital — Integration is now a productivity imperative; completing the Savings and Investments Union is a prerequisite for sustaining technological transformation.
• Monetary policy — Distinguish higher neutral rates (debt, investment demand, productivity) from higher risk premiums (sovereign stress, liquidity shortages).
• Agentic finance — Governance becomes a competitive advantage; banks will be judged on accountability, resilience, controls and common-dependency management.
• Industrial AI — Capital formation must survive the inflation cycle; Europe needs financing structures that sustain adoption despite energy volatility and restrictive policy.

November 2 shows that the economic value of artificial intelligence will depend not only on technological capability, but on the institutions through which capital is allocated, risks distributed and investment sustained. Europe cannot afford to finance the world's technological transformation while allowing its own productive investment to become the first casualty of every economic shock.