Asia’s AI boom has a timing problem
Investors and policymakers should take three mismatches seriously
ASIA’S artificial-intelligence boom is generating economic gains today on the promise of much larger gains tomorrow.
Surging demand for semiconductors, advanced electronics and digital infrastructure is supporting exports and investment across Asean+3 (China, Japan, South Korea and Asean), helping economies withstand an uncertain global environment.
But the region’s growing exposure to the AI investment cycle raises a question that has received less attention: what happens when investment, infrastructure, productivity and financial markets move at different speeds?
The risk is not necessarily that the AI boom will end. It is that expectations may move faster than the economies supporting them.
The scale of the opportunity is substantial. According to the Asean+3 Macroeconomic Research Office (Amro), the region accounted for almost half of global exports of AI-enabling goods, such as chipmaking inputs, memory and other core computing products, in 2025.
In the first half of 2026, these products contributed roughly two-thirds of the increase in Asean+3 merchandise exports.
This demand has helped sustain regional growth despite higher energy costs and persistent trade uncertainty. Amro projects Asean+3 growth of 4.1 per cent in both 2026 and 2027.
Yet the impressive headline numbers conceal three timing mismatches that policymakers and investors should take seriously.
Revolution vs expectations
The first is between technological investment and the power infrastructure needed to support it. AI capabilities and investment plans can expand rapidly. Electricity systems cannot. New generation capacity and transmission networks require years of planning, regulatory approvals, financing and construction.
This creates a two-sided risk. If power infrastructure fails to keep pace, data centres and other AI-related investments may face higher costs or operational constraints.
If infrastructure is built around overly optimistic projections, economies could be left with expensive underutilised assets.
In South-east Asia, AI-related demand adds to electricity needs already rising with manufacturing and urbanisation. In Malaysia, applications for data-centre power supply had exceeded 11 gigawatts by December 2024, equivalent to about 40 per cent of Peninsular Malaysia’s generating capacity.
The second mismatch is between the immediate economic stimulus from AI investment and the time required to realise its productivity benefits.
Manufacturing advanced chips, building data centres, and installing computing equipment generate orders and revenues immediately. Yet the productivity gains promised by AI depend on changes in business processes, workforce skills, organisational structures, and the diffusion of technology across firms.
These adjustments are neither automatic nor instantaneous.
Amro’s analysis of the International Monetary Fund’s AI Preparedness Index puts Singapore, Japan, Korea and Hong Kong above 70 on a 0-100 scale, within the same group as the US.
By contrast, six of the region’s 14 member economies – including the Philippines and Vietnam – score below 50. Economies with lower AI readiness may take longer to translate AI investment into broad-based productivity gains.
A country can therefore experience a powerful AI-driven investment boom long before the technology produces a comparable increase in economy-wide productivity.
That gap does not imply that the investment is misguided. But it creates uncertainty about the timing and distribution of future returns.
This brings us to the third mismatch: financial markets move faster than either infrastructure or productivity.
Equity valuations and financing decisions incorporate expectations about future earnings. If those expectations are revised, financial conditions can change abruptly, even while the underlying technological transformation remains intact.
Recent volatility in global AI and semiconductor stocks illustrates how rapidly this repricing can occur.
Effects ripple
For Asean+3, the implications extend beyond the fortunes of individual technology companies.
The region’s position in global production networks means that a change in AI investment plans could simultaneously affect semiconductor exports, equipment orders, corporate earnings, equity prices, and cross-border capital flows.
These effects could reinforce one another. Weaker export demand would affect corporate revenues, while falling valuations or tighter financing conditions could discourage investment.
Economies heavily exposed to the same global technology cycle could experience these adjustments at roughly the same time.
Amro’s scenario analysis illustrates the region’s exposure to a slowdown in technology investment. If US technology investment were to return to its more moderate 2024 growth pace, rather than maintain its recent momentum, Asean+3 growth in 2027 could be 1.5 percentage points below the baseline.
This is a scenario, not a forecast. Nor does it imply that AI-related investment is unsustainable.
Indeed, AI could deliver substantial and lasting productivity gains. The danger lies in assuming that the path from technological innovation to economic returns will be smooth, or that financial markets will always adjust gradually.
Priorities for policymaking
A technological revolution can be real even when some of the expectations surrounding it prove premature.
This distinction should guide policy.
First, governments need to ensure that essential infrastructure expands in line with realistic demand projections. Public investment and regulatory frameworks should facilitate reliable electricity supply without insulating private investors from commercial risks.
Second, policymakers should focus on converting AI investment into broader productivity gains. This means strengthening skills, encouraging technology adoption among smaller firms, and creating conditions for innovation to spread beyond the leading technology exporters.
Third, financial supervisors should look beyond direct exposures to technology companies. Vulnerabilities may emerge through suppliers, infrastructure financing, equity markets, and non-bank financial intermediaries, such as insurers and private credit funds.
Surveillance should account for the possibility that a common shock could affect several channels simultaneously.
These priorities require neither excessive intervention nor complacency. They call for careful sequencing, credible policies and the flexibility to adjust as technologies and market conditions evolve.
Regional cooperation can help identify shared vulnerabilities and facilitate the exchange of information. But each economy must also assess its own exposure and preserve the policy space needed to respond to unexpected developments.
Asia has good reason to welcome the opportunities created by AI. Its manufacturing capabilities, technological expertise, and production networks place it in a strong position to benefit from the transformation.
But sustaining that advantage requires more than attracting investment or expanding computing capacity.
The ultimate test of Asia’s AI boom will not be how fast investment grows, but how much lasting economic value it creates – and how well the region manages the risks along the way.
The writer is director and CEO of Amro.
AI tools were used in translating the writer’s draft from Japanese to English before submission. The analysis and views expressed are the writer’s own.
This essay is part of New Global Order, which explores how the changing world landscape is reshaping business, politics and beyond.
TRENDING NOW
South-east Asia has lost Nvidia’s best salesman
Demand, tech capabilities position Apac and AI as Skyscanner’s next growth frontiers: CEO Bryan Batista
Could Jensen Huang work for Elon Musk? It would be ‘a titanic ego death match’, says biographer
Jensen Huang is gone. Nvidia’s controls still need explaining