Southeast Asia Should Double Down on the AI Supply Chain

Exports are booming in Asia. Strong growth in AI-related exports — the physical products needed for the global data centre buildout — is driving broader growth in the region. This surging demand is already offsetting the negative impacts from chaotic US trade policy and the Iran war. In fact, the demand is pushing AI supply chains to their limits; production capacity needs to expand. This is an opportunity for Southeast Asia to double down on export-led growth.

While frontier AI capabilities are largely being developed in the US and China, much of the AI infrastructure is being built in, and then exported from Asia. Global data centre capacity — the physical backbone of AI — is set to increase by a compound annual growth rate of 24.6 per cent between 2025 and 2027. Estimates suggest it could potentially double by 2030, creating an additional USD1–2 trillion in IT equipment purchases. In response, orders have risen sharply across semiconductor and electronics supply chains in Asia. AI-related trade is expanding at a much faster rate than trade in other products.

Japan, South Korea and Taiwan produce the most valuable and complex inputs and products along the AI supply chain, such as the most advanced semiconductors. Current capacity is already hitting physical bottlenecks, pushing up prices as demand outpaces supply. Some types of these chips are now five times more expensive, while memory component costs are up 300 per cent this year. The broader suite of less complex but still essential AI supply chain products such as central processing units are also facing production capacity constraints.

Southeast Asia has benefited from AI-related demand by providing less advanced inputs and participating in the assembly, testing and packaging (ATP) stage of the semiconductor supply chain. Some economies — Malaysia, the Philippines, Singapore, Thailand and Vietnam — already enjoy a comparative advantage in chip exports. New Lowy Institute research indicates global semiconductor supply chains are a promising growth opportunity. Southeast Asia is well positioned to continue moving into higher value-added activities.

As AI infrastructure demand outpaces the supply Asia can provide, Southeast Asia will be key for developing new capacity, especially for less advanced electronics manufacturing and semiconductor ATP. An Asian Development Bank report suggests exports of AI-related goods are the main channel developing countries in the Asia-Pacific can benefit from the AI boom. This stems from a boost to industrial output and employment.

As AI infrastructure demand outpaces the supply Asia can provide, Southeast Asia will be key for developing new capacity, especially for less advanced electronics manufacturing and semiconductor assembly, testing and packaging.

However, there are some risks to consider.

The Iran war has revealed that Southeast Asia’s industrial production networks are more exposed to an oil shock than those in China, which is the region’s key competitor in comparable AI-related exports. Accelerating the clean energy transition would help to reduce the region’s dependence on imported fossil fuels, building the economic resilience needed to remain competitive and shield its manufacturing sector from external shocks more effectively.

Semiconductor exports from the region are also facing headwinds. China continues to reduce imports as it builds a fully indigenous semiconductor ecosystem. Imports declined by over 20 per cent between 2021 and 2024. Deepening regional integration across the semiconductor supply chain would help offset declining Chinese demand by increasing economies of scale, thereby reducing unit costs and allowing greater specialisation. Deeper integration could also increase Southeast Asia’s consumption of its own exports.

Much has been written about whether AI investments are a bubble waiting to bring the global economy crashing down. There are good reasons to be sceptical of this view. Experts tend to agree that AI will be economically transformative, albeit with mixed views on the overall impacts on inequality and labour markets. Therefore, it is hard to imagine a future where large increases in production capacity across AI-related products are not needed, even with cyclical dynamics accounted for. Data centres require products across the semiconductor supply chain and other electronic inputs.This demand is unlikely to weaken in the long term as AI diffuses through the global economy. As such, Southeast Asia would need to increase its production capacity.

To better shield the broader economy from any downside risks, the private sector should be the primary driver of supply chain expansions. Governments have an essential role. They can crowd in multinational firms through favourable business environments, support skills development, provide common-use and specialised public infrastructure, and, where necessary, provide some policy support. However, reports that semiconductor firms engage in “incentive shopping”, extracting the most favourable tax and subsidy terms when establishing new projects is concerning. Governments would be remiss to commit substantial public financing for such projects considering current market dynamics. Unmet demand is likely sufficient to drive production expansions without major public financing.

Greater regional coordination, such as adopting an ASEAN ‘best practice’ investment facilitation mechanism, could also rationalise efforts to expand a regional AI-supply chain while further reducing the fiscal burden of attracting foreign investment. Improving regulatory conditions and transparency would help build long-term investor confidence and simultaneously uplift capabilities in lower-income economies. That would allow a more efficient allocation of foreign capital by encouraging lower value-add activities to migrate towards lower-cost economies like Cambodia and Vietnam. That would reduce the subsidy required to attract foreign investment in these economies while simultaneously freeing up input, construction, and fiscal capacity in more advanced economies like Malaysia and Singapore to support only higher-value activities.

Southeast Asian economies are already notable AI supply chain participants. The global AI boom is a strong signal that countries already reaping the benefits, such as Malaysia and Vietnam, should double down. Those less exposed to this boom like Indonesia and Cambodia should get on board quickly.

2026/291