The Great Split Reaches Central Banks

“The next Great Split will be defined not by who adopts AI first, but by who turns it into productive capacity.”

Central banks are entering the AI economy as both policymakers and technology adopters. Two consequential developments on October 28 reveal an important new dimension of the AI Supercycle.

The Bank for International Settlements has published a major international survey demonstrating that central banks are moving decisively from AI experimentation into operational deployment. Yet their dependence on external technology providers, rising implementation costs and uneven governance capabilities are creating new institutional vulnerabilities. Meanwhile, the Bank of Thailand has upgraded its economic forecasts because of AI-driven exports and investment, while warning that the benefits remain concentrated and have not translated into broad-based economic growth.

The emerging Great Split is therefore operating at two levels: between economies capturing AI investment without sufficiently distributing its productivity gains, and between central banks adopting intelligent systems faster than their governance capabilities are maturing. These are distinct but increasingly interconnected challenges.

The most consequential institutional development is the BIS Irving Fisher Committee's new report, AI Readiness in Central Banks, published October 28. Its findings establish a measurable transition.

More than two-thirds of reported central-bank AI projects are now in production, including smaller-scale deployments, compared with approximately half in 2024. Almost two-thirds of surveyed central banks now classify AI as a high strategic priority. Approximately 65% report having dedicated AI infrastructure. Yet 72% primarily use external or externally customised AI models, and almost 80% identify IT expenditure as a risk.

These findings are based on the committee's 2026 survey, rather than deployments that occurred specifically today.

The significance extends beyond operational efficiency. Central banks are increasingly using AI to support economic research, statistics, data analysis and supervisory activities. However, the survey establishes an important distinction: adoption is progressing faster in research and general-purpose functions than in core monetary-policy and prudential-supervision applications.

AI is improving central banks' analytical infrastructure. It has not yet demonstrated that it can independently improve monetary-policy decisions. That distinction deserves particular attention.

The BIS research identifies growing reliance on a relatively concentrated group of frontier AI providers. This creates a new category of institutional risk.

As central banks increasingly integrate external AI models into research, supervision and market intelligence, they could become exposed to common technological limitations. A model failure, cybersecurity incident, data-quality problem or shared analytical bias could affect multiple institutions. This is a potential vulnerability, not evidence that such failures have already produced coordinated policy mistakes.

The governance challenge is particularly revealing. Only 4% of surveyed central banks rate their capacity to monitor AI compliance as high. Most describe it as moderate, developing or limited. The report consequently emphasises stronger oversight mechanisms, particularly for autonomous AI agents, together with data interoperability and reduced dependence on individual providers.

For Europe, the implication is substantial. The ECB and national supervisors need AI capabilities capable of identifying financial transmission mechanisms without introducing excessive technological concentration into the institutions responsible for safeguarding financial stability.

The Bank of Thailand's October 28 monetary-policy decision provides an unusually clear demonstration of the Great Split. The Monetary Policy Committee unanimously maintained its benchmark interest rate at 1.00%, while upgrading its economic growth forecasts.

Expected GDP growth for 2026 increased from 2.3% to 2.4%. The 2027 projection increased from 1.8% to 2.1%. The central bank explicitly attributes part of this improvement to stronger exports and private investment associated with the technology and AI cycle.

But its accompanying assessment is equally important. The benefits remain concentrated in AI-related manufacturing and services, while overall economic growth remains below potential. Household purchasing power remains weak, employment conditions fragile and SME credit constrained. The central bank also projects headline inflation of 2.1% in 2026 and 1.3% in 2027, with energy prices and El Niño presenting temporary upside risks.

Thailand demonstrates that strong participation in the AI supply chain does not necessarily produce sufficient domestic demand to justify tighter monetary policy.

The transmission is: GLOBAL AI INVESTMENT → STRONGER TECHNOLOGY EXPORTS → HIGHER INDUSTRIAL ACTIVITY. But the subsequent diffusion remains incomplete: CONCENTRATED INVESTMENT → LIMITED PRODUCTIVITY SPILLOVERS → WEAK HOUSEHOLD DEMAND → CONTINUED MONETARY ACCOMMODATION.

This is not evidence that AI causes deflation. It demonstrates that an economy can experience AI-driven export expansion while remaining constrained by weak consumption, uneven credit access and insufficient productivity diffusion.

The Bank of Thailand's response is therefore significant: monetary policy remains accommodative despite stronger AI-related external demand. Its decision highlights why central banks cannot equate technology-sector strength with economy-wide economic strength.

The BIS and Thai developments converge on a fundamental economic distinction. AI adoption is advancing across public institutions and private businesses. AI investment is already influencing industrial production, international trade and financial conditions. Yet the relationship between adoption and aggregate productivity remains uncertain.

Three forces must be distinguished. Investment-driven inflation arises when AI capital expenditure increases demand for electricity, infrastructure, financing and specialised equipment faster than supply expands. Productivity-driven disinflation emerges when successful AI deployment sustainably reduces unit production costs and expands effective productive capacity. Demand-driven disinflation occurs when constrained credit, weaker household income or deteriorating investment expectations reduce economic activity.

Only the second mechanism represents the durable supply-side improvement underlying Industrial-AI deflation.

The distinction also matters for neutral interest rates. Sustained productivity growth and stronger investment demand could raise equilibrium real interest rates. Uneven income gains, weak investment diffusion and greater precautionary savings could offset that effect. Today's publications provide new evidence about adoption and monetary transmission, but neither establishes a formal revision to neutral-rate estimates.

The European implications are particularly consequential when today's findings are considered alongside the ECB's October 27 Bank Lending Survey. European banks are tightening corporate credit standards while anticipating greater financing requirements for AI-related digital infrastructure, particularly data centres.

Thailand demonstrates the risks of concentrated technological growth. The BIS demonstrates the risks of institutional dependence on a relatively narrow technological ecosystem. Europe faces both challenges.

Its industrial strategy must therefore accelerate AI adoption across manufacturing, services and SMEs rather than concentrating primarily on computing infrastructure. Its financial authorities must simultaneously strengthen supervisory capabilities, data interoperability and institutional resilience.

The objective is not merely to finance more AI investment. It is to increase the proportion of AI capital formation that becomes measurable, widely distributed productive capacity.

Five Takeaways for Investors and European Policymakers

What to watch

• The BIS establishes a new institutional benchmark. Central-bank AI readiness is becoming measurable through production deployment, governance capabilities, infrastructure and external dependencies. Investors should expect increasingly sophisticated supervisory scrutiny of financial institutions' AI operating models.
• AI adoption does not guarantee productivity diffusion. Thailand's upgraded forecasts demonstrate the benefits of global technology demand, but also the limitations of an investment cycle whose gains remain concentrated.
• External technology dependence becomes a financial-stability consideration. The BIS findings strengthen the case for examining common model providers, operational concentration and resilience across financial institutions.
• Monetary-policy divergence will increasingly reflect the distribution of AI gains. Economies experiencing strong industrial demand may require tighter policy, while those with concentrated export growth alongside weak domestic demand may maintain accommodation.
• Europe's priority must be productive capital allocation. Expanding computing infrastructure is necessary, but sustainable competitiveness requires financing Industrial-AI adoption, workforce adaptation, lower energy costs and broader participation in productivity gains.

The immediate monetary-policy calendar remains consequential. The Federal Reserve and Bank of Canada are scheduled to announce their decisions later on October 28. The ECB's decision follows on October 29. These meetings will provide further evidence of how policymakers balance persistent energy inflation, financing conditions and uneven economic growth.

As of this report, the two confirmed new developments are the BIS research publication and the Bank of Thailand's policy decision. Neither should be interpreted as announcing a new global monetary-policy framework.

The October 28 developments establish an important turning point. Central banks are no longer merely studying artificial intelligence as a potential source of future productivity. They are deploying it within their own institutions while confronting its increasingly visible effects on capital formation, trade and economic growth.

But the BIS reveals that technological adoption can advance faster than institutional governance. Thailand demonstrates that AI-related growth can advance faster than economy-wide productivity. Together, these findings expose the next fundamental challenge of the AI Supercycle. Technological acceleration is not sufficient. Economic and institutional capabilities must evolve at comparable speed.

For investors, the decisive distinction is between AI exposure and AI-generated economic returns. For central banks, it is between incorporating AI into existing analytical processes and developing the capacity to understand a structurally changing economy. For Europe, it is between financing technological infrastructure and building the productive capabilities required to sustain competitiveness.

Primary sources: BIS — AI Readiness in Central Banks, October 28, 2026 · Bank of Thailand — Monetary Policy Decision, October 28, 2026.

The next Great Split will not be defined by who adopts artificial intelligence first, but by who transforms intelligence into productive capacity, financial resilience and better economic decisions.