The AI Productivity Gap

“The decisive variable is no longer the velocity of AI investment, but the velocity of its returns.”

The most consequential monetary development is a new empirical distinction: AI is already contributing meaningfully to economic growth, but its measurable productivity dividend remains considerably smaller than businesses anticipate.

The Bank of England's November 5 Monetary Policy Report provides unusually important evidence. Its new framework suggests emerging AI-related activity has accounted for just under one-fifth of British GDP growth since the start of 2026 — yet business surveys attribute only 0.1–0.2 percentage points of annual productivity growth to AI since early 2024. Capital is being committed on the expectation of future productivity, while interest rates are being set by inflation that is already materialising.

The Report introduces two consequential analyses. First, an AI-activity proxy — spanning programming, information services and software — estimates new AI-related activity at just under 20% of GDP growth since the start of 2026. It is a proxy, not a direct measure of value added, but it shows AI is now macroeconomically significant.

Second, on productivity, the Bank finds much of Britain's recent improvement is cyclical — slower hiring and the reversal of labour hoarding. AI-driven gains exist but remain modest: businesses estimate AI has raised productivity by roughly 0.5% cumulatively over three years, while anticipating far faster gains ahead.

At an ECB conference in Frankfurt, New York Fed President John Williams acknowledged that heavy AI investment is lifting costs for certain technology products and services. He emphasised that inflation expectations remain anchored and wage growth broadly consistent with the Fed's objective, and endorsed no specific next move.

The significance is analytical: the Fed increasingly confronts a technology cycle that can raise prices before generating the productive capacity to reduce them. Markets reported alongside his remarks placed the probability of a December increase above 50% — market pricing, not a Fed commitment. Stronger growth does not necessarily mean lower rates; the direction of the neutral rate remains uncertain.

The Bank raised Bank Rate 25 basis points to 4.00%, voting 8–1, with Catherine Mann favouring a 50bp move. It now expects CPI to approach 5% in Q2 2027, principally on sustained energy pressures. Britain is therefore tightening during an investment expansion that could eventually improve its productive potential — the central paradox of the AI Supercycle. Central banks cannot disregard current inflation because future productivity is expected; investors cannot assume technology will automatically deliver lower rates.

Businesses report little aggregate displacement from AI so far, but increasingly say adoption is reducing recruitment and the replacement of departing staff. Expected productivity gains partly reflect anticipated slower hiring rather than faster output alone. Productivity-led growth supports higher real incomes; employment-led adjustment can lift measured productivity while weakening household income. The Split may appear first in recruitment and income distribution rather than aggregate unemployment.

A November 3 BIS speech by Alexandre Tombini warns that excessive AI investment, complex financing and circular investment relationships can obscure concentrations of risk, alongside operational vulnerabilities in increasingly autonomous systems. Investors should distinguish projects backed by established operating cash flows from those dependent on optimistic utilisation, refinancing or interconnected arrangements.

For Europe, the United States remains the centre of frontier capital formation; Britain is beginning to measure AI's contribution. If global demand lifts real rates, European firms could face costlier capital before achieving equivalent productivity gains. The response is not preferential accommodation but stronger productive-capital formation — integrated markets, competitive energy, and financing for adoption.

Five Key Takeaways

For investors and European policymakers

• The productivity gap is now an investable variable — distinguish AI capex from measurable gains in output, margins and returns.
• Inflationary investment demand can precede Industrial-AI deflation.
• Employment divergence — slower recruitment with rising productivity — could become a leading indicator.
• Financial concentration deserves closer scrutiny: credit structures, refinancing dependence and interconnected exposures.
• Europe must accelerate the conversion of savings into productive investment.

The next Great Split will be determined by the distance between the capital invested in intelligence and the productivity it ultimately delivers — and the decisive variable is no longer the velocity of AI investment, but the velocity of its financial returns.