23 OCT 2026 — Central banks should keep policy tight while the AI boom is still mostly a spending story, and loosen only once the productivity payoff is visible in the data, the Monetary Authority of Singapore's (MAS) chief economist said on 20 October. Edward S. Robinson, who is also MAS's deputy managing director for economic policy, made the argument in remarks at the 2026 Beijing Financial Street Forum, which MAS has published in full.
The speech set out a general sequence for policymakers. It did not announce a decision, and it did not say when, or whether, MAS itself would ease.
What Robinson said about policy
His prescription came in two steps. "Monetary policy should be appropriately restrictive in the initial stages of the boom, while demand effects dominate," Robinson said. Only later, he said, might policy "turn more accommodative as productivity gains demonstrably take hold". Because the risk of getting the calibration wrong is high during that transition, he argued for moving gradually and letting the data lead.
The word doing the work is "demonstrably". Robinson warned against assuming that AI has already moved the benchmarks central banks steer by: potential output, the natural rate of unemployment, the neutral interest rate and, for Singapore, the neutral path of the exchange rate. Strong activity today, he said, may still be cyclical demand rather than a lasting rise in what the economy can produce.
Investment first, productivity later
Robinson framed the boom as a race between demand now and supply later. Spending on data centres and energy networks is surging, and higher asset prices are lifting demand through wealth effects. The productivity dividend is still unclear.
In Asia, he said, the industries that make AI hardware are already adding to growth. AI-driven electronics account for more than 70% of the region's export growth this year, a figure Chia Der Jiun, MAS's managing director, gave at the annual report media conference on 28 July, when he put the 2024 share at 46%. That growth, Robinson said, mostly reflects more capital and machines run harder, not broad efficiency gains. Workers and equipment are "working harder", he said, which is not yet the same as "working smarter".
The longer-run case for AI
Across the wider economy, Robinson said, AI is spreading faster than the personal computer or the internet did, citing research on generative AI uptake. Early evidence shows efficiency gains in particular applications, but he judged it too soon to see an economy-wide lift. He pointed to electrification, whose productivity benefits became large only after factories were rebuilt around it.
The upside he cited is sizeable. With broad enough adoption and the right investment, OECD estimates suggest AI could add up to 0.6 percentage points a year to total factor productivity growth, which he said could offset some of the drag from ageing populations in Asia. He also raised the prospect of AI making healthcare, education and legal advice cheaper to provide. "Perhaps the seemingly intractable Baumol cost disease could finally find a cure in AI," he said, referring to the tendency of labour-heavy services to grow steadily dearer.
Financial stability comes first
Robinson tied the case for caution to markets as well as prices. Hopes of large future productivity gains may already be built into asset valuations, he said, and over-optimistic investors could borrow too heavily. A later reassessment could send asset prices and investment into reverse and tighten credit.
His answer was to use macroprudential tools alongside interest rates, to guard against borrowing secured on revenues that may never arrive. He also called for budget, regulatory and structural measures to spread AI use through digital infrastructure and skills training. "We cannot assume that this time is different," he said near the close.
Chia had made a related point in July, noting that companies tied to AI made up about 40% of the S&P 500 by market value, and over 30% of the MSCI EM Asia Index.
Where Singapore's own policy stands
The remarks followed MAS's October policy statement by six days. On 14 October MAS kept the rate of appreciation of its exchange-rate policy band unchanged, with no change to its width or centre, after tightening in April and July. The statement gave advance estimates of 5.4% year-on-year GDP growth in the third quarter, led by technology-related sectors.
MAS forecast core inflation to average 1.5% to 2.0% this year, and both core and headline inflation to average 1.5% to 2.5% in 2027. It named the AI boom among the upside risks, saying domestic inflation could stay high if its spillovers broaden and unit labour costs climb faster.
Read together, the two documents describe a central bank that sees the boom lifting growth and prices now and wants hard evidence before treating it as a supply-side gift. Robinson also flagged a complication: if AI's gains are spread unevenly, job losses and income worries could raise precautionary saving and pull the neutral rate down instead of up. Policymakers, he said, should be ready for either result.
View author profile → · Editorial policy
