AI Becomes a Central-bank Stress Scenario

“The next phase will be decided not by the scale of capital expenditure, but by the system's ability to sustain it until returns materialise.”

The Big Take — The AI Supercycle has entered a new phase of monetary and financial scrutiny. Two consequential central-bank developments, combined with today's financial-market repricing, point toward a fundamental shift.

On October 23, the Bank of Japan formally incorporated a potential collapse in AI investment expectations into its financial-system stress testing, examining how shocks could propagate through non-bank financial intermediaries into commercial banks. Meanwhile, the Reserve Bank of Australia has explicitly identified AI investment and technology-related price increases as contributors to inflation, strengthening its case for monetary restraint.

The Great Split is becoming a divergence between economies capable of absorbing AI's financial risks and those exposed to its inflationary costs without capturing the corresponding productivity gains. The central question is no longer whether AI will transform productivity. It is whether the financial system can withstand the transition from extraordinary capital expenditure to sustainable economic returns.

The most consequential institutional development comes from the Bank of Japan's October 23 Financial System Report. For the first time in this monitoring sequence, the BOJ has presented an explicit assessment combining deteriorating AI investment expectations, foreign non-bank financing exposures and international financial-market contagion within its macroprudential stress-testing framework.

The report identifies an increasingly important change in how the AI infrastructure expansion is financed. Major technology companies historically relied heavily on internally generated cash flows. They are now using more corporate borrowing, long-term data-centre leasing commitments and financing arrangements involving private funds and other non-bank intermediaries. The BOJ warns that a deterioration in expected returns could trigger portfolio adjustments among these intermediaries, amplifying market shocks and transmitting losses internationally.

Crucially, its stress tests indicate that Japan's banking system currently retains sufficient capital and funding resilience even under severe combined scenarios. The report does not establish that an AI-driven banking crisis is underway.

The hidden transmission mechanism: AI profitability disappointment → AI equity and credit repricing → non-bank portfolio adjustments → cross-border financial contagion → bank balance-sheet pressure → tighter credit conditions.

The important development is not simply recognition of these risks. A major central bank is explicitly incorporating AI-related profitability shocks and their transmission through non-bank finance into its assessment of banking-system resilience. That is a meaningful evolution in macroprudential analysis.

The second-order effect: The next financial correction could be amplified by the increasingly complex financing structures supporting AI infrastructure. Corporate borrowing, private credit, long-term leasing commitments and concentrated equity exposures create multiple channels through which declining investment expectations could affect financial institutions. For investors, the implication is to distinguish companies with self-financing capacity from businesses whose growth depends on continuing access to increasingly expensive external capital. For supervisors, the priority is to map exposures across banks, private funds, technology companies and infrastructure-financing vehicles.

The BOJ also identifies frontier-AI-enabled cybersecurity threats as a macroprudential concern, reinforcing the international shift toward treating operational resilience as a component of financial stability. Its response includes continued on-site examinations and off-site monitoring.

The second important development comes from Reserve Bank of Australia Deputy Governor Andrew Hauser's October 21 intervention, whose implications have become more consequential as energy prices and financial volatility rise.

Hauser identified three sources of inflationary pressure: prolonged geopolitical disruption, extraordinary AI investment and insufficient domestic productive capacity. He stated that technology-related price increases are beginning to feed through into consumer inflation. Australia is also experiencing a substantial increase in data-centre construction and associated investment demand. The problem is that the country's productivity performance has not yet demonstrated an equivalent acceleration — AI capital formation is increasing demand before industrial AI is expanding supply.

Hauser acknowledged that the inflation outlook had deteriorated materially relative to the RBA's August projections. He did not announce another policy-rate increase, but his remarks reinforce the possibility of further tightening.

The monetary consequence: Australia illustrates a central risk of the AI Supercycle. If investment in computing infrastructure, electricity and technology increases demand faster than productive capacity expands, inflation can remain elevated even when employment growth moderates. This produces an uncomfortable policy choice. Central banks may need to restrain aggregate demand despite the possibility that technological investment will eventually raise economic productivity. The monetary-policy cost of the transition is incurred before the productivity dividend is realised.

This asymmetry is increasingly relevant to the Federal Reserve, ECB and Bank of England. It also reinforces the distinction between a possible increase in the natural interest rate caused by sustained productive investment and a temporary increase in financing costs caused by inflation, risk premiums or capital scarcity. No formal AI-driven reassessment of the neutral rate accompanied these latest developments.

Financial markets are now testing the assumptions underlying the AI infrastructure expansion. Following Thursday's sharp technology selloff, the Nasdaq declined approximately 2.6%, while semiconductor shares fell almost 7%. Investors simultaneously moved toward US Treasuries, pushing benchmark yields below 5%.

By October 23, markets had also reduced expectations for additional monetary tightening. Reuters reported that the implied probability of a December Federal Reserve increase had fallen to approximately 85%, while the corresponding ECB probability had declined to 69%, from 90% earlier in the week. These are market-implied probabilities, not central-bank commitments. At the same time, Brent crude climbed toward $114, and European natural-gas prices had risen almost 30% during October.

The fundamental asymmetry: Two opposing forces are operating simultaneously. AI financing concerns → lower equity valuations → safe-haven demand → lower benchmark sovereign yields. But: energy price increases → higher production costs → inflation persistence → monetary restraint.

A third mechanism complicates the picture. Even as benchmark Treasury yields decline, corporate credit spreads can widen. Consequently, the risk-free cost of money can fall while the effective financing costs of highly leveraged AI infrastructure companies increase. Lower sovereign yields do not necessarily mean easier financial conditions for the AI economy. This distinction may determine the next market rotation.

For Europe, these developments introduce a particularly consequential challenge. The global AI Supercycle can raise capital demand, energy consumption and technology-related input prices without producing an equivalent increase in European productive investment.

Europe could therefore experience three simultaneous pressures. First, imported inflation from energy and technology-related demand. Second, volatility in international financing conditions as investors reassess AI profitability. Third, sovereign fragmentation that prevents lower benchmark yields from translating into cheaper capital throughout the euro area. France's renewed sovereign-market stress makes the third channel especially important.

The ECB must distinguish the disinflationary consequences of financial tightening from any genuine increase in productive capacity. A contraction in credit-financed AI investment is not Industrial-AI deflation. It may reduce inflationary demand while simultaneously damaging growth, earnings and future productive capacity.

The critical European policy response should therefore combine deeper capital markets, energy resilience, more diversified AI investment financing and stronger supervision of interconnected financial institutions. Europe must also avoid excessive dependence on a narrow group of external technology providers and financing intermediaries. The BOJ's analysis offers a useful supervisory reference point: stress testing should examine not only direct bank exposures but also the indirect channels linking AI-related assets, leveraged investors and global markets.

Five Strategic Takeaways

What to carry forward

• Financial stability — AI has become a formal macroprudential stress scenario. The BOJ's October 23 assessment establishes a concrete precedent for examining AI profitability shocks through international non-bank financial intermediation.
• Monetary policy — Investment inflation is arriving before Industrial-AI deflation. Australia provides direct evidence that AI-related demand is entering central-bank inflation assessments while productivity improvements remain insufficient to offset near-term pressures.
• Capital markets — The next rotation will be driven by financing resilience. Investors should differentiate between companies financing AI through operating cash flows and those dependent on continuous borrowing, refinancing and favourable capital-market valuations.
• Europe — Sovereign fragmentation can amplify technological divergence. The ECB must monitor how global financial tightening interacts with domestic credit spreads, sovereign risk and Europe's comparatively limited AI infrastructure investment.
• EU industrial policy — Productive capital formation becomes a financial-stability priority. A deeper Savings and Investments Union, competitive electricity prices, industrial AI adoption and stronger cross-border supervisory coordination should be treated as complementary elements of Europe's economic resilience.

The October 23 developments mark an important institutional threshold. The AI Supercycle is no longer being assessed exclusively as a prospective source of productivity growth or an extraordinary investment boom. Central banks are increasingly examining its capacity to generate inflationary demand, alter capital-market conditions and transmit financial instability.

The Bank of Japan is explicitly testing the consequences of deteriorating AI investment expectations. Australia is incorporating AI-related investment demand into its inflationary risk assessment. Global financial markets are beginning to differentiate more aggressively between the promise of AI productivity and the financing structures supporting that promise.

These developments point toward a common conclusion. The next phase of the AI Supercycle will be determined less by the scale of capital expenditure than by the financial system's ability to sustain that expenditure until productive returns materialise.

For investors, that makes capital quality and financial resilience increasingly decisive. For central banks, it makes the distinction between inflationary investment, productivity-led disinflation and credit-driven contraction essential. For Europe, it makes domestic capital formation, energy security and financial integration matters of strategic urgency.

The Great Split is entering its most consequential phase: the winners will not be those that invest the most in artificial intelligence, but those capable of transforming that investment into sustainable productivity without compromising monetary and financial stability.