Southeast Asia’s data centre boom is built on borrowed time

Every hyperscale data centre requires three physical resources: power, water and land. It also requires one facilitating condition – the support of a government willing to regulate and accommodate demand for those coveted resources. Southeast Asia is witnessing an AI infrastructure boom, taxing its capacity to deliver all four.

While each ASEAN country takes a different path to AI mastery, they all risk arriving at the same destination: the creation of a voracious industry that rewards economic elites but brings few direct benefits to affected local communities and loses its social license to operate.

Hyperscalers are hunting for strategic ecosystems, moving beyond just cheap resources. Amazon, Google, Microsoft and Oracle have committed more than $US 160 billion ($A 231 billion) to Southeast Asia, competing directly with Chinese firms such as Alibaba and ByteDance. Regional governments are using AI infrastructure to further their own agendas too.

Take three examples of different pathways to managing investment. Under Singapore’s National AI Strategy 2.0, the city state has cast AI buildout as a national capability, rather than a service leased to foreign tech. Thailand’s strategy is to bet on building AI applications for its healthcare, tourism and agriculture. And the Philippines unveiled a $34.4 billion roadmap in September 2026 to scale its data centre capacity thirtyfold by 2033.

When both the stakeholders – that of hyperscalers and the government – are racing toward the same prize, neither has obvious reasons to pause and weigh the resource cost first. And here is where the governance gap gets created.

Singapore's route is the most sophisticated. Between 2019 and 2022, the city-state imposed a moratorium on new domestic data centres, citing energy and land constraints on an island with no domestic power generation slack to spare. The moratorium has since been lifted with tighter efficiency conditions attached. But Singapore’s underlying limits have not disappeared.

What has changed is where the growth now takes place. In April 2026, Indonesian state utility PLN Batam signed a deal with DayOne Data Centres for roughly 450 megawatts of capacity on Batam island, an hour by ferry from Singapore. Two months later, Nvidia, Firmus Technologies and DayOne announced a 360-megawatt ‘AI factory’ on the same island, designed to house up to 170,000 Nvidia chips from 2027. Batam is positioned inside a free-trade zone with Singapore’s financial and logistics infrastructure a short hop away.

But this simply displaces the intractable resource burdens. Batam’s water and grid capacity are already under severe strain. Indonesia gains investment and jobs hosting the hardware, but Batam absorbs the environmental costs of an industry whose profits, tax base and reputational credit remain firmly registered in Singapore. This arrangement allows Singapore to host the region’s AI economy on paper while its physical footprint lands elsewhere. This is unfortunately not a loophole. It is the model working as intended.

Thailand’s route is different. It is racing to catch up with the demand for resources instead of exporting it. Google has committed roughly $1 billion to data centre and cloud infrastructure, split between Bangkok and Chonburi, while Amazon has pledged $5 billion. These commitments are a part of a larger wave. Of the THB 1.87 trillion ($A 79.5 billion) in approved investments by the Board of Investment in 2025, 40% was for data centres – the largest commitment of any single industry.

And approximately 70% of approved data centre projects are clustering in the Eastern Economic Corridor, which is a manufacturing zone under water stress from existing industrial demand. Unlike Singapore, Thailand has no willing neighbour to which it can export the consequences. Its only option is to regulate the infrastructure that it has begun to build.

The Board of Investment reworked its incentive criteria  for data centre     s last year to favour those of ‘high quality’ – the most energy and water efficient and those that “demonstrate tangible benefits for Thailand” in training, research development, and supporting domestic businesses. These high-quality centres receive an eight-year tax holiday; others receive five years. 

In August, the energy minister announced a major liberalisation of the electricity market to allow any company to purchase power directly from renewable energy suppliers rather than solely through the state utility. This privilege had previously been extended to data centres alone under the 2,000 megawatt cap pilot programme, which was scrapped. The minister also created a new tariff category aimed at ensuring data centres paid the true cost of electricity and avoided additional demand forcing up prices for other consumers. 

The government was visibly rewriting its rules in real time. Evidently, the investment arrived faster than any framework built to govern it.

The Philippines is where this expansion exacts its highest human toll. It has formally joined Pax Silica, a United States-led framework for securing AI and semiconductor supply chains, as its latest Southeast Asian partner. This membership determines whose supply chains Washington trusts, acting as a distinction with consequences for trade, investment and technology access well beyond one project.

Manila has proposed a 1618-     hectare data centre and semiconductor hub at New Clark City, which is pitched as the country’s entry point. But the site's own developer, the Bases Conversion and Development Authority, says the hub will not host AI data centre     s. Citing insufficient electricity supply, it pledged to manufacture the components that feed AI infrastructure built elsewhere.  

The environmental group Kalikasan estimates the project would displace roughly 20,000 people – Aeta indigenous communities and farmers combined. This is no longer a case of a hypothetical cost weighing against a hypothetical benefit. It is a concrete example of the potential human cost attached to a project whose economic returns are still in the speculative stage, and whose supply chain rationale is as much geopolitical as it is commercial.

The master deal between Manila and Washington, still being negotiated, will decide whether the communities facing displacement will get a fair price before the construction begins.  It is positioned to be one of the clearest early tests of whether the region’s AI boom can be governed on terms other than its own momentum.

Despite these risks, the investment is still indispensable. Singapore, Thailand and the Philippines – and their fellow ASEANs – want the jobs, tax revenue, geopolitical positioning and, most importantly, technological modernisation that come with data centres and the AI tools they enable. Even when the benefits might be hard to quantify, there is a fear of being left behind.

But the constraints are also becoming more widely understood and debated in the region.

Singapore has a genuine physical limitation: an island with little room left to expand. The Philippines has heard the displacement estimates of Kalikasan and the denials of Manila’s own development authority that data centres are coming to New Clark City. Independent reporting has documented water stress in Thailand’s Eastern Economic Corridor. Clearly awareness is not the missing piece. It is sequencing. Regulatory frameworks, water allocation plans and community consultation processes are being built retroactively, which results in a loss of the host government’s negotiating power. 

The costs don't just disappear because they are deferred. They resurface later – in Batam’s water table, in the Eastern Economic Corridor’s next dry season, and in whatever New Clark City becomes once construction starts. Each government has arrived at a different version of the same arrangement: host the AI boom, as long as its worst costs land somewhere else in time and space. Good governance dictates that accountability arrives before the money does.

Debomita Dasgupta is policy researcher based in Melbourne, working on climate policies and sustainability governance in the Indo-Pacific. Her writings have appeared in Lowy Institute's The Interpreter, Australian Outlook and the DevPolicy Blog.