Deputy Governor Lu Lei, distinguished officials, speakers and guests, good morning.

Introduction

1. The AI boom is playing out as a race to balance present demand against future supply effects. On the demand side, upbeat expectations have prompted an unprecedented run of AI investment and expansion. On the supply side, however, verified gains in productivity from the new technology have been more limited, at least so far. As difficult as it is to measure present commitments against as-yet-unrealised future benefits, this calculation is made even more challenging by today's global environment.

An Age of Perma- and Poly-shocks

2. We entered 2026 still adjusting to trade measures introduced over the preceding year. The conflict in the Middle East, which began in end-February, rapidly developed into a global energy shock.

3. This is not a passing phase. We are now in the era of “perma-” and “poly-shocks”, where disturbances arrive more frequently, last longer, and overlap to a greater extent, testing the absorptive capacities of economies and financial systemsBrown, G., El-Erian, M., & Spence, M., with Lidow, R. (2023). Permacrisis: A plan to fix a fractured world. Simon & Schuster..

4. Yet the global economy has remained remarkably resilient. Supply chains are often characterised as propagators of shocks, but they can also act to mitigate shocks when firms use them to substitute inputs and reconfigure production. Recent research finds that diversified supplier networks materially dampen the impact of supply disruptionsAhn, J., & Tan, B. J. (2025). Supply chain diversification and resilience (IMF Working Paper No. 25/102). International Monetary Fund.. So it proved in 2025: firms rerouted and diversified sourcing, and world merchandise trade grew by an estimated 2.4%, confounding earlier forecasts of a contractionWorld Trade Organization. (2025). Global trade outlook and statistics: Update.. However, this resilience is not inexhaustible, as buffers can be depleted, and repeated adaptation is costly.

AI and Productivity

5. A third shock of a very different character to these trade and energy perturbations, is the emergence of AI at a seemingly fortuitous moment. The AI investment and production boom has become a major force sustaining global demand. For several Asian economies, it is a powerful counterweight to the drag from trade tensions and higher energy costs.

6. During the information-technology revolution of the 1990s, a distinction was made between industries that produce the new technologies and those that use themJorgenson, D. W., & Stiroh, K. J. (2000). Raising the speed limit: U.S. economic growth in the information age. Brookings Papers on Economic Activity, 2000(1), 125–236.. In the case of AI, the former is already boosting growth in Asia. AI-driven electronics account for over 70% of Asia’s export growth this yearMonetary Authority of Singapore. (2026). Remarks by Mr Chia Der Jiun, Managing Director, MAS, at the MAS Annual Report 2025/2026 Media Conference on 28 July 2026. . Beyond manufacturing, Asia, including China, is also making important progress in the underlying AI technologies, from advanced chips to large language models.

7. For now, the strength in the AI-producing sectors appears to mainly reflect capital accumulation and higher capital utilisation, rather than a broad-based increase in total factor productivity. New data centres are being built, advanced equipment is being used more intensively, and workers and machines are, in effect, “working harder”. This raises output but is not yet the same as “working smarter” through sustained efficiency gainsKlein, M. C. (2026). AI, productivity, and rates: Part I. The Overshoot..

8. On the AI-using side, AI is diffusing across the wider economy faster than the personal computer and the internet didBick, A., Blandin, A., & Deming, D. J. (2026). The rapid adoption of generative AI. Management Science.. While early evidence points to efficiency gains in specific applications, it remains too soon to identify a decisive economy-wide step-up in productivityBrynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889–942..

9. Economic history counsels patience. A century ago, the productivity benefits of electrification became substantial only after factories and workflows were redesigned around the new technologyDavid, P. A. (1990). The dynamo and the computer: An historical perspective on the modern productivity paradox. The American Economic Review, 80(2), 355–361.. AI may require a similar process, with economy-wide productivity gains depending on the breadth of adoption, complementary investments in skills and infrastructure, and organisational redesignOrganisation for Economic Co-operation and Development. (2026). OECD compendium of productivity indicators 2026..

10. With sufficiently broad adoption and appropriate investment, AI could raise total factor productivity growth by up to 0.6 percentage points per annum, according to OECD estimatesFilippucci, F., Gal, P., & Schief, M. (2024). Miracle or myth? Assessing the macroeconomic productivity gains from artificial intelligence (OECD Artificial Intelligence Papers No. 29). OECD Publishing.. This could meaningfully offset the negative effects from demographic ageing and other structural headwinds in Asia.

11. For Asia, AI diffusion into non-tradable sectors could prove especially important. According to the Balassa-Samuelson theorem, faster productivity growth in tradables raises economy-wide wages and the price of non-tradable servicesBalassa, B. (1964). The purchasing-power parity doctrine: A reappraisal. Journal of Political Economy, 72(6), 584–596; Samuelson, P. A. (1964). Theoretical notes on trade problems. The Review of Economics and Statistics, 46(2), 145–154.. AI could modify this pattern by lifting productivity directly in domestic services, narrowing sectoral gaps and spreading welfare gains more broadly. By extending the reach of expertise, AI may enable a wider range of workers to perform higher-value tasks previously reserved for specialised professionals. This could lower the cost of services such as healthcare, education and legal advice, while allowing the benefits of productivity growth to reach a larger share of societyAutor, D. (2024). Applying AI to rebuild middle class jobs (Working Paper No. 32140). National Bureau of Economic Research.. Perhaps the seemingly intractable Baumol cost disease could finally find a cure in AI.

Implications for Monetary Policy

12. For central banks, the core challenge is to weigh the emergence of AI as an important favourable long-run supply shock, against the current macroeconomic impact of higher demand. Investment in data centres and energy networks is surging, while higher asset valuations are raising demand through wealth effects, even as the productivity dividend remains unclear.

13. Financial stability considerations reinforce the need for caution. Expectations of large future AI-driven productivity boosts may become embedded in asset valuationsLenzu, S. (2026). Artificial intelligence and monetary policy: A framework and perspective on cyclical transmission, structural transition, and financial stability (Staff Report No. 1192). Federal Reserve Bank of New York.. If investors become overly optimistic, leverage could rise while valuations overshoot. A subsequent reassessment could then provoke reversals in asset prices, credit conditions and investment rates.

14. Monetary policy calibration must remain evidence-based and state-contingent. It might be premature to assume that AI has already shifted the “stars” that guide monetary policy – including potential output (y*), the natural unemployment rate (u*), the neutral interest rate (r*), and, for Singapore, the neutral exchange rate path (e*). Strong activity may still reflect cyclical demand rather than a lasting increase in productive capacityIn Singapore, the balance of risks shifted towards inflation this year, as AI-related activity supported growth and the positive output gap was projected to widen. The Monetary Authority of Singapore accordingly tightened monetary policy in calibrated steps in April and July..

15. Eventually, the policy “stars” may themselves move. Sustained productivity growth could raise potential output and the natural interest rate through higher expected returns on investment. But if the gains from AI are unevenly distributed, the resulting labour displacement and income uncertainty could raise precautionary saving, exerting downward pressure on the natural interest rate. The net effect remains uncertain, and policymakers should be prepared for either outcome.

16. Joseph Stiglitz recently observed that the optimistic scenario rests on AI achieving “sufficient success” to be creatively disruptiveStiglitz, J. (2026, September 12). The AI economy. Financial Times.. Yet, technological success alone will not suffice, given the disruption that AI could bring. Accordingly, managing the AI transition will require a multi-layered policy response.

17. Monetary policy should be appropriately restrictive in the initial stages of the boom, while demand effects dominate. It may then turn more accommodative as productivity gains demonstrably take hold, and the “stars” shift. The risks of policy miscalibration during this transition would suggest a gradual and data-dependent approach.

18. At the same time, macroprudential policy should be used to guard against leverage built on over-anticipation of inherently uncertain future revenues. Budget and regulatory measures, coupled with structural policies, should promote AI diffusion, through digital infrastructure, skills training and labour-augmenting applications, and anticipate the risks. Finally, the proper regulatory guardrails must be set in place to preserve the trust in this new technology.

Conclusion

19. The optimism over AI is not entirely misplaced. In the Solow tradition, technological progress is ultimately what permits economies to produce more without proportionately more labour and capital. The greater promise of AI is to broaden the frontier of human creativity, by enabling more people to innovate, create, and exercise judgment in their daily work. In Edmund Phelps' terminology, the goal is not just higher productivity, but economic dynamism and mass flourishingPhelps, E. (2013). Mass flourishing: How grassroots innovation created jobs, challenge, and change. Princeton University Press..

20. Yet history also records repeated episodes in which expectations ran ahead of productivity and valuations ran ahead of fundamentals. We cannot assume that this time is different. The challenge for policymakers is to smooth out the transition while the demand and supply sides of the AI economy converge. Sound macro-policy cannot eliminate uncertainty but can promote price and financial stability as the necessary conditions for the long-run promise of AI.

21. Thank you.

***