Southeast Asia's AI Boom: Limited Benefits from Data Center Construction Frenzy?

Read The Diplomat, Know The Asia-Pacific Southeast Asia might end up hosting substantial computing capacity without developing a correspondingly deep AI economy. Headlines have for years been touting the emergence of Asia as a global powerhouse in data center development. The nature of headlines is that all too often they prove to be horrendously wrong in the long or even the short run – but when it comes to data centers in Southeast Asia, they are repeatedly getting validated. The Asia-Pacific’s data center development pipeline reached a record 26.5 GW in the first half of 2026. Southeast Asia accounts for roughly half of the capacity under construction: Malaysia leads with 1,039 MW, followed by Thailand with 859 MW. All this investment is producing a division of labor among national markets: Singapore concentrates on high-value, power-constrained activity while Malaysia, Thailand, Indonesia and Vietnam are absorbing different forms of large scale growth. The question is not whether Southeast Asia will add more data centers. It undoubtedly will. The more pressing question is whether the infrastructure being built across separate national markets could become the foundation of a regional AI economy. Numbers of course, matter still. They always will. But not all numbers tell the exact same story. Under construction figures, for example, do not describe current operational capacity. At about 1.46 GW, Singapore retains the largest operational base in Southeast Asia, but its new capacity remains tightly controlled. Limits on power, land, and sustainability all but ensure that the country is unlikely to compete with its neighbors simply by adding volume. Instead, these constraints strongly favor higher value uses including premium workloads and research compute while reinforcing Singapore’s status as the region’s high trust, highly connected high tech hub, supported by its strengths in talent development, governance, research, and connectivity. It’s of little surprise, then, that Singapore’s second Data Center Call for Applications and plans for a low-carbon park on Jurong Island both point toward increasingly selective, high specification projects rather than an expansion bonanza. On August 21, the government provisionally awarded 200 MW to four data center operators out of more than 20 applications. This latest and all too predictable display of Singapore’s constraints is bound to gradually push more investment across the Causeway. Indeed, on the other side of the Johor Strait, Malaysia has emerged as the region’s leading scale market. More available land and comparatively lower costs combined with supportive industrial policy are attracting facilities designed for hyperscale operations and AI training workloads. Thailand, by contrast, had been expanding from a smaller base, with investment approvals and partnerships with domestic power producers supporting rapid growth particularly around Bangkok. Indonesia and Vietnam illustrate another approach in which domestic market size and questions of digital sovereignty matter alongside foreign investment. Indonesia is pairing the region’s largest domestic market of about 288 million with large foreign AI projects while still requiring government and other designated strategic data to stay onshore. CoreWeave’s planned 360 MW deployment in Greater Jakarta illustrates how the country is moving from a primarily domestic-demand story toward a major destination for foreign AI infrastructure investment. Meanwhile Vietnam, home to 102 million people, has sought to attract investment through regulatory liberalization, including allowing wholly foreign owned data centers. Capital, as it tends to, is moving across national markets, potentially enabling workloads to be distributed according to different national advantages, and constraints in one country increasingly affect investment decisions in its neighbors. As a consequence, instead of viewing them as isolated national opportunities, investors increasingly view ASEAN markets as complementary locations within a broader infrastructure strategy. On August 1, SoftBank’s Singapore subsidiary signed a memorandum of understanding with SC ZEUS, a Singapore-based data center developer, and Robust HPC, a Malaysian high-performance computing specialist, to develop AI data-center capacity focused initially on Malaysia, Thailand, Indonesia, and Vietnam. However logical it might be, this cross-border approach also exposes a problem. Instead of removing physical constraints, moving investment across borders redistributes them. Both Malaysia and Thailand are now facing limits in grid connections, transmission capacity, and the speed at which new power can be delivered. In early September Thailand asked developers to halt work on about 49 projects under construction and froze approvals for about 117 until it finalizes national rules on power, water, siting and local benefit. In Johor, the Malaysian state bordering Singapore, water supply has also been under pressure. Neither Bangkok or Kuala Lumpur are in a unique place. The United States, which accounts for approximately 45 percent of global data center capacity, is struggling to connect new load to the grid as demand continues to rise. Although energy generation potential is crucial for data center developers, it is far from being the only factor: whether firm power can reach the right location on the required timetable is equally, if not more, important. The ASEAN Power Grid seeks to make cross-border electricity trade more practical. An early example of what such a broader regional system could look like is offered by the Laos-Thailand-Malaysia-Singapore Power Integration Project, Rest assured, greater regional power connectivity will not eliminate local shortages or transmission bottlenecks, but it could give governments and developers a larger geography over which electricity supply and demand can be balanced. For an industry where expansion is increasingly constrained by access to reliable power, this matters almost as much as the movement of data. As data centers spring up across the region, a thornier question is also taking shape: what do countries actually capture from all this infrastructure? Data centers bring construction. They bring engineering jobs, energy investment, and tax revenue. They also bring ceremonial ribbons to cut at opening ceremonies. But much of the higher-value activity associated with AI lies elsewhere: software, model development, research, specialized services and the use of AI across domestic industries. This is similar to a country hosting large electronics assembly operations without capturing the design, chip-making, or software layers that creates most value. Southeast Asia might end up hosting substantial computing capacity without developing a correspondingly deep AI economy. Hence the terms attached to investment matter a lot. Governments can use access to land, power, and incentives to s