JLL, DC Byte, Yardi Map APAC’s Hyperscale Data Centre Race to 2030

Major US cloud providers are planning Asia Pacific data centre capacity for 2030 and beyond, with operators facing record-low vacancies and pressure to deliver campuses exceeding 100 megawatts. JLL’s Asia Pacific head of data centre research Mikhail Jaura and DC Byte lead analyst Vivian Wong spoke d

Major US cloud providers are already planning Asia Pacific data centre capacity for 2030 and beyond after mapping out requirements through 2028 and 2029, as operators face record-low vacancies and pressure to deliver larger campuses faster, according to JLL. Discussions are increasingly focused on campuses of more than 100 megawatts, JLL’s Asia Pacific head of data centre research Mikhail Jaura said Tuesday during the Mingtiandi APAC Data Centre Forum on MTD TV, sponsored by Yardi.

“The conversations that we’re having right now is capacity planning for 2030 and onwards,” Jaura said. Operators are telling JLL that they lack enough inventory to supply the market, he added. Cloud remains the largest source of take-up in Asia Pacific excluding mainland China, accounting for about 40 per cent compared with roughly 22 per cent for AI, according to DC Byte lead analyst Vivian Wong, who also spoke during the session.

Mingtiandi’s special report, The Hyperscale Race in Asia Pacific, puts the region’s combined operational and planned IT capacity, ex mainland China, at 110 gigawatts in the second quarter of 2026. About two-thirds of that total was at an early stage, while roughly 11GW was operational. Wong said deals are exceeding 100MW, with customers seeking “very aggressive turnaround timelines”. Southeast Asian markets are generally delivering faster than mature locations, typically taking 18 to 24 months and sometimes less, depending on utility connections and contractors.

Gigawatt campuses typically begin with smaller installations and expand according to customer demand. “Realistically, the developer will also not be able to secure the resources for the entire gigawatt campus at once,” Wong said. Institutional investors remain reluctant to back speculative development, Jaura told MTD TV, with land, power and labour constraints testing ambitious proposals. Greater scrutiny of water and electricity allocations in Johor is also slowing the pipeline. “There’s very little sort of appetite for speculative build,” he said.

Platforms established between 2015 and 2020 are now stabilising and holding long leases, creating opportunities for core investors seeking predictable income. “Here in Asia, I would say we’re sort of at the start of the exit cycle,” Jaura said. Data centre asset trades across APAC ex mainland China totalled US$13.9 billion in 2025, down 37 per cent from US$22.2 billion in 2024, the report’s MSCI Real Assets figures show. Japan accounted for more than half of the US$3.7 billion traded in the first half of 2026.

National data privacy and financial regulations are another reason for providers to maintain multiple locations, according to Yardi’s Asia Pacific senior regional director Bernie Devine. Operating data centres across multiple zones helps meet compliance requirements for clients including banks and listed companies. Mingtiandi’s research tracked 25 new cloud availability zones in Malaysia, Thailand, Indonesia and the Philippines since 2022, compared with 18 across the rest of Asia Pacific. Malaysia added 10 and Thailand eight, while Singapore added none, though it remains Southeast Asia’s largest market by zone count.

Enterprise AI adoption, robotics and mobility applications could substantially increase computing requirements beyond today’s generic chatbots, Devine said. As companies gain confidence in tools that perform repetitive tasks, their deployment should expand. “The demand for compute will just go up enormously,” he told the forum. Wider enterprise AI adoption also depends on better data management. Moving information between systems creates integration costs and risks losing accuracy, while consistent, structured data gives AI the context needed to produce reliable results. Companies are finding that “a lot of the errors are coming from data problems”, Devine said.