Artificial intelligence, explained.
In Malaysia, data centres need RM4.6 million of investment for every job created (Amar Syazwan Rosman/Unsplash)
Exports of AI hardware are lifting regional growth, but the rules that decide who profits are being set without it.
Subscribe to The Informer for monthly expert analysis, and to Events for advance notice of visiting world leaders and distinguished guests.
You may unsubscribe from Lowy Institute newsletters at any time. For information on our privacy practices and how to unsubscribe, see our Privacy Policy.
East Asia’s manufacturing escalator has slowed – AI alone will not restart it East Asia’s manufacturing escalator has slowed – AI alone will not restart it
At the UN General Assembly last month, artificial intelligence (AI) was centre stage (Opens in new window). Brazil and France worried about power concentrated in a handful of firms and states. Developing countries asked for access to computing power, know-how and infrastructure, and for a real voice in setting standards to shape AI to fit their developmental needs. The United States treated AI dominance as a strategic and economic imperative.
Less than a week later, the World Bank’s October East Asia and Pacific Economic Update, “Riding the AI Wave” (Opens in new window), looked to identify benefits. While much of the world debates who will shape AI, East Asia is busy riding its wave. Putting the story together, it shows what AI export-oriented growth can and cannot do for emerging economies.
According to the World Bank, China, Indonesia, Malaysia, the Philippines, Thailand and Vietnam exported US$1.4 trillion of AI-related goods in the 12 months to April. In early 2026, US imports of AI-enabling goods rose almost 60% while its other imports fell by nearly a quarter. AI-related exports from Vietnam and Malaysia have at least doubled from their 2020–22 average. The World Bank’s upgrade to the region’s 2026 growth forecast rests largely on these chips, servers and routers driven by a wave in AI investment.
A closer look at what is being exported is warranted. Assembled computers, servers and routers make up 69% of Thailand’s AI-related exports, and 60% of Vietnam’s. China’s AI-related exports of US$785 billion are mostly intermediates, much of it shipped to ASEAN countries for assembly. The design, software, intellectual property and data that make these goods valuable are largely created and captured elsewhere.
This is the familiar lesson of global value chains: what matters is which tasks a country performs, not which products it ships overseas.
When developing countries ask the UN for a voice in setting AI standards, they are making a trade policy demand too.
Manufacturing was not only good for exports, it used to be the escalator (Opens in new window) that moved workers off farms and into better-paid, more productive work. The World Bank is candid that this escalator has slowed.
Countries that reached their peak industrial employment share in 2015 and 2019 did so at incomes of roughly US$900 to US$5,900 per head, far short of high-income status. AI hardware does not change this. In Malaysia (Opens in new window), data centres need RM4.6 million of investment for every job created, against less than RM1 million in electronics and automotive jobs. Tellingly, when the World Bank looks for jobs at scale, it turns to tourism and agribusiness, not AI.
This is where services come in. Services (Opens in new window) already account for roughly 26–33% of the value of global manufactured exports. The report confirmed that services provide large support to the region’s growth, but it also conceded that its AI trade numbers miss AI-related services altogether: software development, business process outsourcing and data work.
This is precisely where the Philippines and Vietnam are building new capabilities. The report also cites Vietnam’s (Opens in new window)experience around its accession to the World Trade Organisation of opening telecommunications, finance and other services, which raised productivity not only in those sectors but in the manufacturing sectors that rely on them. Notably, while Vietnam still lags in services openness (Opens in new window) compared to other East Asian countries, it sees the export-led growth path through services capability, not around it.
Anthropic CEO Dario Amodei gives an on-screen presentation to the UN Security Council meeting on Artificial Intelligence and International Security, 23 September 2026 (Manuel Elías/UN Photo)
Which brings us back to New York. The UNGA debate focused on safety, ethics and access, all real concerns. But the economic terms of the AI economy are being set somewhere else, and mostly outside multilateral forums.
The region’s AI export boom rests on unilateral decisions of importing markets. Most AI-related goods remain exempt from US tariffs. What is granted unilaterally can be withdrawn unilaterally. The boom also rests on an investment cycle, with the largest US hyperscalers spending more than US$1 trillion over 2025 and 2026, which the World Bank warns could reverse if it proves to have run ahead of demand.
Meanwhile, the rules that will decide who captures value from AI – on cross-border data flows, digitally delivered services, standards and export controls – are being written unilaterally or at best in bilateral and mini plurilateral deals and national security reviews. Most developing countries have little say in any of them.
When developing countries ask the UN for a voice in setting AI standards, they are making a trade policy demand too, whether or not they frame it that way. Standards, services commitments and data rules are a large part of trade governance. Treating AI governance and trade governance as separate conversations suits those who already dominate both.
So what still should work for emerging economies as their growth drivers? Openness still works, but openness to services, technology and ideas now matters more than cheap labour and economies of scale. Countries need to compete on tasks rather than sectors, invest in physical and digital connectivity and skills, and reform services regulation so that their manufacturers and digital exporters can draw on efficiently provided inputs. And they need to treat the multilateral rulebook as something worth fighting for: not because it is perfect, but because the alternative is a world in which the terms are set by the few with the most compute.
East Asia has shown, once again, that it can build what the world wants to buy. The harder task is making sure it also has a hand in writing the rules under which it is sold. And, just as crucially, it must put more of what it builds to work at home.
About the author
Mia Mikic
Mia Mikic is a trade economist with an interest in inclusive development and a proven track record and experience in academia and international civil service. She is a member of the Friends of Multilateralism Group (Opens in new window), the Board of Advisers for the Trade Policy Research Forum (Opens in new window) and the Advisory Board of the Centre on Inclusive Trade and Development (Opens in new window), Georgetown Law, United States.
The most-pressing world events explained by Lowy Institute experts and global contributors, in your inbox, every Thursday.
You may unsubscribe from The Interpreter at any time. For information on our privacy practices and how to unsubscribe, see our Privacy Policy.
