The Great Split Is Now in the Data

A consequential shift has emerged in the central-bank response to the AI Supercycle. The Federal Reserve is documenting an economy in which AI infrastructure investment is displacing other productive activity while AI adoption is beginning to restrain hiring. Simultaneously, the Financial Stability Board has established its final international framework for responsible AI adoption, and the Bank of England is confronting the implications for financial resilience and the neutral interest rate.

The Great Split is no longer exclusively a divergence between technological leaders and laggards. It is becoming a measurable divergence in capital allocation, employment, monetary transmission and financial-system resilience.

Four developments establish the change.

The Federal Reserve's October 14 Beige Book provides important evidence of a development we have been tracking: AI's inflationary and potentially disinflationary forces are beginning to coexist.

The distinction is visible in the labour market.

Five Federal Reserve districts reported that some firms had reduced staffing through attrition, postponed recruitment or anticipated fewer new jobs after improving productivity through AI.

These are qualitative reports from business contacts, not an economy-wide causal estimate. Nevertheless, they indicate that AI-related labour substitution is appearing in the Federal Reserve's assessment of economic conditions.

At the same time, the Kansas City Federal Reserve reported that data-centre construction was crowding out other production and intensifying competition for skilled workers.

Manufacturers were increasing working hours, introducing overtime and extending shifts because available labour was insufficient.

The Cleveland district identified data centres, power generation and defence spending as important sources of manufacturing demand.

The contrast is consequential.

AI infrastructure investment → stronger demand for capital and specialised labour, while AI adoption → higher productivity in selected tasks → weaker demand for some occupations.

Both developments can occur simultaneously.

This produces a more complex monetary environment than the conventional distinction between inflationary growth and disinflationary productivity.

The AI investment economy can experience rising input costs and labour shortages while AI-adopting businesses restrain recruitment.

The second-order consequence. The Great Split may therefore widen even within the same national economy. Capital-intensive activities absorb resources and generate demand. Businesses deploying AI selectively reduce their labour requirements. Households whose earnings fail to keep pace with inflation face weaker purchasing power.

The Fed's latest evidence shows how these different forces can coexist within an otherwise modest expansion. A single measure of employment or wage growth may increasingly conceal divergent economic realities.

This matters particularly after the October 14 inflation release. US headline CPI reached 3.7% in September, while core inflation increased to 2.6%.

Gasoline accounted for more than half the monthly headline increase, demonstrating that the latest inflation acceleration cannot be attributed principally to AI. Nevertheless, the coexistence of energy inflation, infrastructure demand and emerging labour substitution complicates the Federal Reserve's assessment of underlying inflation persistence.

The monetary-policy implication is clear: industrial-AI deflation cannot be inferred from weaker hiring alone. Sustainable disinflation requires demonstrable productivity growth that expands productive capacity.

Two October 15 interventions from the Bank of England introduce complementary dimensions.

Chief Economist Huw Pill argued that higher bond yields across advanced economies may partly reflect an increase in the real neutral interest rate.

He also cautioned investors against interpreting a prospective interest-rate increase as the beginning of another prolonged tightening cycle.

This is important because Pill distinguished the current inflationary environment from the dynamics that produced successive rate increases during 2022–23.

His remarks do not establish that AI has caused an increase in the neutral rate. They do, however, reinforce the importance of separating monetary-policy expectations from changes in the underlying equilibrium cost of capital.

For the AI Supercycle, the distinction is fundamental. If extraordinary investment demand raises equilibrium real rates, central banks may need to maintain higher nominal rates even as some productivity-related disinflation begins to emerge.

Lower inflation does not automatically imply a return to the exceptionally low real interest rates of the previous economic regime.

Meanwhile, Deputy Governor Sarah Breeden introduced a more operational financial-stability requirement. Speaking in Bangkok, Breeden warned that AI-enabled cyberattacks had become sufficiently sophisticated that banks could no longer rely exclusively on preventing intrusions.

Financial institutions must also be capable of restoring essential operations after their primary systems have been compromised. She specifically highlighted independent recovery capabilities, including so-called bare-metal recovery.

The implication extends beyond cybersecurity. As banks deploy autonomous systems across payments, credit, liquidity management and market operations, their dependence on shared technological infrastructure grows.

A serious disruption could affect institutions that are otherwise financially sound.

The emerging supervisory principle is that financial resilience must be measured not only by capital and liquidity buffers, but also by the capacity to restore critical services after technological failure.

The most consequential institutional development today comes from the Financial Stability Board.

On October 15, the FSB formally published its final report on responsible AI adoption, establishing twelve sound practices addressing organisation-wide governance and risks throughout the AI lifecycle.

The practices cover board accountability, risk assessment, explainability, model performance, human oversight, cybersecurity and third-party dependencies.

They are nonbinding international guidance, not newly enforceable regulations. Nevertheless, the framework creates a common reference point for supervisors and financial institutions across jurisdictions.

The final version also strengthens expectations surrounding agentic AI. For higher-risk applications, the FSB calls for more intensive human oversight and clear triggers requiring human approval before autonomous systems execute certain actions.

This represents a meaningful refinement of the supervisory approach. The regulatory question is moving from whether financial institutions can use AI responsibly to whether they can retain effective control over autonomous AI execution.

The financial consequences. The implications are potentially significant.

First, financial institutions capable of implementing reliable controls may gain a competitive advantage as AI adoption accelerates. Second, technology providers offering demonstrable auditability, explainability and operational resilience could become increasingly valuable to regulated financial institutions.

Third, compliance and implementation costs may reinforce concentration among larger institutions unless proportional supervisory approaches preserve opportunities for smaller competitors.

Fourth, common dependence on a limited number of models, cloud providers and technology platforms could remain a systemic vulnerability even where individual institutions meet supervisory expectations.

The deeper issue is therefore one of system-wide interdependence. The FSB framework can improve governance within institutions. Supervisors must also understand how shared technological dependencies could transmit disturbances across institutions.

The IMF's October 15 Asia-Pacific outlook provides the international counterpart to the Federal Reserve's domestic evidence.

The IMF projects regional growth of 4.8% in 2026, moderating to 4.4% in 2027. The upward revision for 2026 reflects, in part, the exceptional strength of AI-related investment and technology exports.

But the gains are concentrated. South Korea and Taiwan have benefited substantially from semiconductor demand and their positions in global AI supply chains.

Other economies face higher energy and financing costs without capturing comparable increases in technology investment.

This is the international manifestation of the Great Split. Economies positioned inside the AI capital-formation cycle experience stronger investment, exports and income generation. Economies outside that cycle remain exposed to its energy requirements, capital-market consequences and competitive pressures.

The IMF also warns that economies benefiting most directly from AI demand may be disproportionately vulnerable to a correction in technology investment or valuations.

The implication for central banks is that the same global shock can generate different monetary-policy requirements. Some economies may need to contain excessive demand. Others may need to protect growth while managing imported inflation and capital outflows.

There is no universal AI monetary-policy reaction function.

For Europe, today's developments reinforce an increasingly urgent conclusion. The European policy challenge is not simply to accelerate AI adoption. It is to ensure that adoption translates into higher domestic productivity, stronger capital formation and sustainable economic returns.

The Federal Reserve's evidence illustrates why the distinction matters. AI can displace labour before its aggregate productivity benefits are fully visible. AI infrastructure can crowd out other investment before its economic returns justify the capital employed. Global financing requirements can increase Europe's cost of capital without generating equivalent European productive assets.

Europe must therefore connect industrial AI, capital-markets integration, energy infrastructure and workforce adaptation within a common investment strategy.

The ECB can manage monetary conditions. But monetary policy cannot substitute for the creation of productive capital.

Five Takeaways for Investors and EU Policymakers

where capital and supervision meet

• Monetary policy: The Federal Reserve's Beige Book offers new evidence that AI-related labour substitution and inflationary infrastructure demand are developing simultaneously. Central banks must distinguish productivity-led disinflation from demand weakness.
• Neutral rates: Huw Pill's remarks reinforce the possibility of a higher equilibrium real interest rate. Investors should not assume that moderating inflation will automatically restore the financing conditions of the previous decade.
• Capital allocation: The evidence of data-centre investment crowding out other production makes capital productivity increasingly important. The critical question is whether additional AI infrastructure generates returns sufficient to compensate for its opportunity cost.
• Financial supervision: The FSB's final framework and the Bank of England's recovery warnings elevate autonomous-system controls, technological dependencies and operational recovery capabilities into strategic board-level priorities.
• European competitiveness: The Great Split can be narrowed only if Europe converts savings, energy capacity, industrial expertise and technological adoption into productive investment. Supervisory resilience and capital formation must advance together.

The latest evidence marks a consequential evolution.

The AI Supercycle is beginning to alter the allocation of labour and capital before its long-term contribution to aggregate productivity has been fully established.

At the same time, central banks and international supervisors are adapting to a financial system in which autonomous technology could accelerate decisions, reshape risk transmission and increase operational interdependence.

The emerging policy challenge is therefore more demanding than controlling inflation during an investment boom. It is to preserve monetary and financial stability while the economy reallocates capital, labour and productive capacity on an unprecedented scale.

For investors, the decisive variable is increasingly the return on capital after accounting for financing costs, operational risk and the displacement of existing productive activity.

For Europe, the decisive variable is the speed at which AI adoption becomes measurable domestic productivity.

The next economic regime will be determined by which economies convert AI investment into productive capacity faster than the combined costs of capital, energy, labour displacement and systemic risk accumulate.