The Productivity Dividend Meets the Cost of Capital

THE BIG TAKE — The most consequential development is not another interest-rate decision. It is new evidence that artificial intelligence is beginning to reshape monetary transmission itself.

On October 30, ECB Executive Board member Philip Lane presented an extensive assessment of AI adoption, financing and productivity at the Federal Reserve Bank of San Francisco. His findings reveal that AI-related activity is now contributing materially to European credit growth, but AI-intensive companies are also more sensitive to monetary tightening than comparable businesses.

Meanwhile, new Bank of England evidence shows AI is already altering recruitment, professional-services pricing and corporate cost structures. At the Federal Reserve, Dallas Fed President Lorie Logan has explicitly called for at least two additional interest-rate increases, despite uncertainty about AI investment and its eventual returns.

Together these developments expose a central contradiction. The economy is becoming more productive in selected activities, but the capital required to distribute those gains remains expensive, unevenly allocated and increasingly sensitive to monetary policy. This is a material new development in the Great Split thesis.

Philip Lane's October 30 speech, AI Adoption in Europe, provides the strongest new empirical evidence. Euro-area digital investment has increased by 73% in real terms since 2014, rising from approximately 2.3% to an estimated 3.3% of GDP. Businesses in the ECB's Corporate Telephone Survey report that AI has increased their capital-expenditure plans by approximately 10% this year, and AI-related borrowing accounted for approximately one-quarter of annual aggregate corporate credit growth in Q1 2026.

AI is no longer merely a prospective productivity shock. It is already influencing investment demand, borrowing requirements and the transmission of monetary conditions to the real economy.

The hidden pattern: tightening can penalise the companies building future productivity. AI investment often requires substantial upfront expenditure before returns materialise. If borrowing costs stay elevated, businesses may postpone investments that could eventually reduce unit costs. The mechanism is: higher interest rates → higher AI financing costs → lower innovation investment → delayed productivity gains → weaker future disinflationary capacity.

This is a potential second-order consequence of monetary restraint, not a reason to abandon inflation targets. But it changes how policymakers should assess the longer-term costs of tightening: restrictive policy can reduce inflationary demand today while delaying the productive capacity required to contain inflation tomorrow. The new ECB evidence makes that trade-off more measurable.

Lane confirms that European investors are significant purchasers of American technology equities and euro-denominated bonds issued by US companies. Purchases of these corporate bonds have increased alongside declining European purchases of US Treasury securities. This does not prove investors are directly selling Treasuries to buy hyperscaler bonds, but it demonstrates a meaningful change in the composition of European international investment flows.

The ECB also finds financing remains particularly difficult for intangible investments — training and organisational capabilities that cannot readily serve as loan collateral. European banks are better equipped to finance tangible infrastructure than the less collateral-intensive activities necessary for widespread AI adoption.

The second-order effect: capital formation without domestic productivity formation. European savings help finance American technological infrastructure; American providers expand their platforms; European businesses purchase access. Yet smaller European companies may lack funding for the organisational transformation required to use those capabilities productively.

The Bank of England's October 30 Agents' Summary offers a counterpoint. Businesses increasingly report measurable productivity improvements from AI, particularly in professional services — but transmission to prices is not straightforward. Some firms are changing their pricing models rather than automatically cutting prices; value-based pricing is becoming more common, while customers increasingly demand a share of savings. Labour-cost savings are being partially offset by higher spending on cloud computing, software, cybersecurity, data infrastructure and governance. Employment effects are emerging predominantly through reduced recruitment and trainee intakes rather than widespread workforce reductions.

The critical distinction: lower unit costs do not automatically produce lower prices. Businesses may pass productivity gains to customers, retain them through higher margins, or see them absorbed by additional technology spending and restructuring. AI-driven improvements in individual processes do not necessarily produce immediate economy-wide disinflation.

There is also a labour-market consequence. If businesses reduce entry-level recruitment while retaining experienced employees, aggregate unemployment may initially remain stable even as opportunities for younger workers deteriorate. The first significant employment effect of AI may be fewer jobs created rather than existing jobs destroyed — a measurement problem, since traditional indicators may underestimate the structural adjustment beneath stable headline statistics.

Lorie Logan's October 30 intervention argues that US monetary policy remains insufficiently restrictive and that at least two additional quarter-point increases are required under her current outlook. She observes that longer-term Treasury yields have risen substantially during 2026, reflecting, in her assessment, expectations of stronger nominal growth and higher neutral interest rates. She also acknowledges the investment outlook could strengthen or weaken as businesses learn more about AI's commercial applications.

This creates a distinction between the Fed and ECB. America's stronger investment cycle may support expectations of higher equilibrium rates; Europe's weaker capital-market integration and more uneven diffusion could produce a different outcome. However, neither central bank has established a definitive AI-specific revision to its neutral-rate estimate. Higher market yields should not automatically be read as proof that the equilibrium real interest rate has permanently increased.

The energy constraint. A separate October 30 Dallas Fed report documents competition between data-centre developers and energy producers for labour, equipment, infrastructure, water and other resources in the Permian Basin — concrete regional evidence of inflationary demand pressures from AI infrastructure construction. Building the AI economy can raise resource costs before operating it reduces production costs. That timing difference is becoming central to the next phase of monetary-policy analysis.

The Bank of England and Financial Conduct Authority's AI survey, published October 26, provides context that should not be overlooked. Among responding financial institutions, 86% report using AI, up from 75% in 2024. Approximately 80% report having deployed AI agents, though only 2% report agents with broad autonomy. Third-party implementations now account for 53% of reported high-materiality AI systems, and the three largest cloud and computing providers account for 79% of reported provider relationships. These are survey findings, not evidence of a current systemic incident.

The supervisory implication is significant. Institutions may become more operationally efficient while increasing their dependence on concentrated external technology infrastructure — creating channels for correlated operational failures, cyber incidents and liquidity disruption. The ECB's October 30 speech similarly documents rising cybersecurity and IT incidents within banking operational-risk events. The next financial-stability challenge is not simply the amount of credit financing AI, but the interaction between that credit, shared technology dependencies and the speed of automated financial decisions.

The Great Split — Five Strategic Takeaways

Who should do what

• For monetary policymakers: Distinguish AI-driven investment demand from realised productivity growth. The two can have opposite inflationary implications, particularly when their timing differs.
• For European investors: Evaluate AI exposure through sustainable returns on invested capital, cash-flow generation, financing sensitivity and whether productivity gains are retained in margins or passed to customers.
• For EU policymakers: Address the financing gap for intangible investment. Training, organisational transformation and proprietary applications are essential to productivity diffusion, but conventional collateral-based lending is poorly suited to many of these activities.
• For banking supervisors: Intensify monitoring of AI-related credit exposures, common cloud dependencies, third-party concentration and operational resilience.
• For European governments: Treat competitive electricity, deeper capital markets and widespread Industrial-AI adoption as complementary components of competitiveness rather than separate initiatives.

The October 30 developments materially strengthen the Great Split thesis. The ECB shows AI-related investment is influencing credit growth and monetary transmission. The Bank of England confirms AI is changing employment, operating costs and pricing behaviour before the aggregate consequences become fully visible. The Federal Reserve is confronting the possibility that stronger investment and productivity may alter the equilibrium interest-rate environment.

The next major divergence will be determined by three factors: the ability to finance technological adoption; the ability to convert adoption into sustained productivity; and the ability to maintain monetary and financial stability throughout that transition. Europe's challenge is particularly consequential because its financial system is still better equipped to finance tangible assets than the intangible capabilities on which much of the AI productivity dividend depends. The objective should not be to shield technological investment from monetary discipline — it should be to build capital markets capable of financing productive transformation across the interest-rate cycle.

The defining competitive divide of the AI Supercycle will separate economies in which capital finances productivity from those in which the cost and allocation of capital prevent productivity from emerging.