Why Malaysia became Southeast Asia’s data centre concentration
11 Jun 20265 min read
Summary
- NEXTDC opened KL1 in Kuala Lumpur on 14 May 2026, an A$1 billion (RM2.8 billion) build delivering 65 megawatts of IT capacity, designed for AI workloads with integrated rainwater harvesting alongside air cooling, per the NEXTDC press release.
- Malaysia approved 143 data centre projects between 2021 and June 2025 worth RM144.4 billion (about US$33 billion), according to the Ministry of Investment, Trade and Industry (MITI).
- The concentration is reshaping which Asia geographies host physical AI workloads and what that does to regional data sovereignty, logistics inferencing and the supply chain decisions that depend on real-time compute.
Australian colocation operator NEXTDC opened KL1 in the Klang Valley on 14 May 2026, its first international data centre and the most visible single move in the latest phase of Malaysia’s data centre build-out. The technical choices matter more than the headline. KL1 is air-cooled with integrated rainwater harvesting, designed for AI workloads at 65 megawatts of IT capacity, and it sits inside a national pipeline that the Ministry of Investment, Trade and Industry says has now approved 143 data centre projects worth RM144.4 billion (about US$33 billion) between 2021 and June 2025.
A data centre is a building filled with computer servers that does the actual processing work behind every cloud service, every video stream, every artificial intelligence model the world is rushing to build. Each server consumes electricity and generates heat. The traditional limit on where you could build a data centre was electricity supply. Cool a data centre with chilled water systems and you also need substantial water. The next generation of AI models, particularly the ones running on the latest Nvidia chips, generate more heat per server than older workloads and therefore consume more cooling water, often by a factor of three to five.
Malaysia’s data centre concentration is now the largest in Southeast Asia, and it has been built on cheap electricity, available land in the Iskandar and Klang Valley corridors, and proximity to Singapore’s connectivity infrastructure without Singapore’s land and water constraints. NEXTDC’s KL1 is positioned as the first Uptime Institute Tier IV-certified data centre in Peninsular Malaysia (the Uptime Institute is the global standards body for data centre design; Tier IV is its highest resilience grade, requiring fully redundant power, cooling and network paths). The Malaysia-based technology publication Digital News Asia confirms the design integrates air-cooled chillers and rainwater capture into the same building envelope. The design choice is the tell-tale sign. The operator engineered around water rather than around power, treating water as the binding constraint on the next phase of the build-out.
The pipeline that sits behind KL1 explains why. Malaysian business publication BusinessToday and the country’s national news agency Bernama report that the 143 approved projects are concentrated in Johor, Selangor and Negeri Sembilan, with hyperscale operators including Microsoft, ByteDance and Google securing capacity through colocation and build-to-suit arrangements. Major builds since 2024 include announcements from Bridge Data Centres (Singapore-headquartered, US-owned), AIMS (Malaysian incumbent), Yondr (US specialist), NTT Global Data Centres (Japanese telecoms group) and YTL Power International (Malaysian conglomerate operating the Kulai-Sedenak cluster). The Iskandar Puteri cluster, a federal-state special economic zone in Johor anchored on Singapore-adjacent land at the southern tip of Peninsular Malaysia, has drawn commitments measured in the billions of US dollars from the same hyperscaler customers that previously concentrated their Southeast Asia capacity in Singapore.
The supply chain consequence is the part that has been underreported. Where physical AI workloads run determines which countries can operate real-time AI applications without delay: a logistics dispatcher routing a fleet, a factory inspecting goods on a production line, a retailer pricing items dynamically. “Inferencing” is the industry term for running an AI model to produce a decision, and it has to happen geographically close to where the decision is needed. A factory in Penang that uses computer vision to inspect circuit boards depends on a model running close enough to the factory to keep round-trip latency under a few milliseconds. A logistics operator routing trucks across Peninsular Malaysia depends on the same locality. As the data centre concentration spreads across Singapore, the Klang Valley and Iskandar, the operational dependencies follow. The manufacturing belt and the data centre belt are now overlapping in ways they were not five years ago.
The water question is becoming the binding constraint that power was a decade ago. Malaysia’s National Water Services Commission tracks water demand growth across Selangor and Johor that already strains the existing reservoir and distribution network. KL1’s rainwater capture is one operator’s answer; other operators are looking at closed-loop water recycling, seawater cooling for coastal sites in Johor, and direct-to-chip liquid cooling that uses less water but more sophisticated hardware. The next phase of Malaysian data centre permitting, signalled by the announcement of a single licensing authority covering data centre approvals, will incorporate water sustainability criteria more explicitly than power consumption. Operators that did not engineer for water from the design stage will face a slower approval cycle than operators that did.
The geopolitical question follows. Where physical AI workloads land changes how regional data sovereignty is structured. Singapore has held the position of Southeast Asia’s primary inferencing and training hub since the late 2010s. The Klang Valley and Iskandar concentration is absorbing the growth tier of customers that no longer fit inside Singapore’s land and water budget. Singapore retains the high-end mandate; Malaysia takes the spillover. Malaysian Digital Authority licensing, Malaysian Communications and Multimedia Commission oversight, and Malaysian Personal Data Protection Act compliance now bind a meaningful share of Southeast Asia’s AI workload stack.
For a regional infrastructure planner, the question is what comes after the current pipeline. If 143 projects worth RM144.4 billion (about US$33 billion) have already been approved, and if the next phase of permitting weighs water sustainability heavily, the operators that built first will hold the operational advantage longer than the operators that arrive in the next 24 months. Singapore’s spillover to Johor is already visible in cross-border power and water arrangements. Two adjacent geographies are positioning to compete for the next wave of regional data centre capacity. Vietnam’s central provinces (Da Nang and Quang Nam in particular) offer cheap hydropower and proximity to a growing domestic AI market. Indonesia’s Riau Islands, immediately south of Singapore, offer geographic adjacency to Singapore’s cable landing infrastructure and lower land costs than Johor. Whether either jurisdiction can match Malaysia’s permitting speed and Tier IV-grade engineering depth will define the second half of this build-out cycle.
The headline number in Malaysia’s data centre boom is the megawatts. The number that will decide the next phase is the cubic metres.
A data centre is a building filled with computer servers that does the actual processing work behind every cloud service, every video stream, every artificial intelligence model the world is rushing to build. Each server consumes electricity and generates heat. The traditional limit on where you could build a data centre was electricity supply. Cool a data centre with chilled water systems and you also need substantial water. The next generation of AI models, particularly the ones running on the latest Nvidia chips, generate more heat per server than older workloads and therefore consume more cooling water, often by a factor of three to five.
Malaysia’s data centre concentration is now the largest in Southeast Asia, and it has been built on cheap electricity, available land in the Iskandar and Klang Valley corridors, and proximity to Singapore’s connectivity infrastructure without Singapore’s land and water constraints. NEXTDC’s KL1 is positioned as the first Uptime Institute Tier IV-certified data centre in Peninsular Malaysia (the Uptime Institute is the global standards body for data centre design; Tier IV is its highest resilience grade, requiring fully redundant power, cooling and network paths). The Malaysia-based technology publication Digital News Asia confirms the design integrates air-cooled chillers and rainwater capture into the same building envelope. The design choice is the tell-tale sign. The operator engineered around water rather than around power, treating water as the binding constraint on the next phase of the build-out.
The pipeline that sits behind KL1 explains why. Malaysian business publication BusinessToday and the country’s national news agency Bernama report that the 143 approved projects are concentrated in Johor, Selangor and Negeri Sembilan, with hyperscale operators including Microsoft, ByteDance and Google securing capacity through colocation and build-to-suit arrangements. Major builds since 2024 include announcements from Bridge Data Centres (Singapore-headquartered, US-owned), AIMS (Malaysian incumbent), Yondr (US specialist), NTT Global Data Centres (Japanese telecoms group) and YTL Power International (Malaysian conglomerate operating the Kulai-Sedenak cluster). The Iskandar Puteri cluster, a federal-state special economic zone in Johor anchored on Singapore-adjacent land at the southern tip of Peninsular Malaysia, has drawn commitments measured in the billions of US dollars from the same hyperscaler customers that previously concentrated their Southeast Asia capacity in Singapore.
The supply chain consequence is the part that has been underreported. Where physical AI workloads run determines which countries can operate real-time AI applications without delay: a logistics dispatcher routing a fleet, a factory inspecting goods on a production line, a retailer pricing items dynamically. “Inferencing” is the industry term for running an AI model to produce a decision, and it has to happen geographically close to where the decision is needed. A factory in Penang that uses computer vision to inspect circuit boards depends on a model running close enough to the factory to keep round-trip latency under a few milliseconds. A logistics operator routing trucks across Peninsular Malaysia depends on the same locality. As the data centre concentration spreads across Singapore, the Klang Valley and Iskandar, the operational dependencies follow. The manufacturing belt and the data centre belt are now overlapping in ways they were not five years ago.
The water question is becoming the binding constraint that power was a decade ago. Malaysia’s National Water Services Commission tracks water demand growth across Selangor and Johor that already strains the existing reservoir and distribution network. KL1’s rainwater capture is one operator’s answer; other operators are looking at closed-loop water recycling, seawater cooling for coastal sites in Johor, and direct-to-chip liquid cooling that uses less water but more sophisticated hardware. The next phase of Malaysian data centre permitting, signalled by the announcement of a single licensing authority covering data centre approvals, will incorporate water sustainability criteria more explicitly than power consumption. Operators that did not engineer for water from the design stage will face a slower approval cycle than operators that did.
The geopolitical question follows. Where physical AI workloads land changes how regional data sovereignty is structured. Singapore has held the position of Southeast Asia’s primary inferencing and training hub since the late 2010s. The Klang Valley and Iskandar concentration is absorbing the growth tier of customers that no longer fit inside Singapore’s land and water budget. Singapore retains the high-end mandate; Malaysia takes the spillover. Malaysian Digital Authority licensing, Malaysian Communications and Multimedia Commission oversight, and Malaysian Personal Data Protection Act compliance now bind a meaningful share of Southeast Asia’s AI workload stack.
For a regional infrastructure planner, the question is what comes after the current pipeline. If 143 projects worth RM144.4 billion (about US$33 billion) have already been approved, and if the next phase of permitting weighs water sustainability heavily, the operators that built first will hold the operational advantage longer than the operators that arrive in the next 24 months. Singapore’s spillover to Johor is already visible in cross-border power and water arrangements. Two adjacent geographies are positioning to compete for the next wave of regional data centre capacity. Vietnam’s central provinces (Da Nang and Quang Nam in particular) offer cheap hydropower and proximity to a growing domestic AI market. Indonesia’s Riau Islands, immediately south of Singapore, offer geographic adjacency to Singapore’s cable landing infrastructure and lower land costs than Johor. Whether either jurisdiction can match Malaysia’s permitting speed and Tier IV-grade engineering depth will define the second half of this build-out cycle.
The headline number in Malaysia’s data centre boom is the megawatts. The number that will decide the next phase is the cubic metres.