Building Sustainable Data Centers for Malaysia’s AI Future

• EdgeConneX is deploying 500+MW of AI-ready capacity across Cyberjaya and Johor, with facilities designed to handle GPU workloads from 20kW to 600kW+ per rack.
• A closed-loop, advanced cooling system saving an estimated 16.6 million liters of water daily across both sites compared to traditional evaporative systems.

Enterprises and hyperscalers deploying GPU-intensive workloads need facilities that can handle 50-600 kilowatts per rack, sustained cooling for liquid-cooled systems, and the flexibility to scale as chip architectures evolve.

Most existing data centers in Southeast Asia weren’t built for that, and retrofitting legacy facilities doesn’t work at the densities AI requires. EdgeConneX is building two hyperscale campuses in Malaysia: Cyberjaya and Johor, designed specifically for those workloads. The approach reflects a broader brand philosophy: infrastructure should adapt to customer needs and local conditions.

The technical gap between traditional data centers and what AI demands is significant. Legacy facilities handle 10-15kW per rack. Modern AI training clusters need 50-100kW. Next-generation GPU architectures are pushing toward 200-600kW and beyond. The difference isn’t just about plugging in more power–AI workloads generate exponentially more heat, require different network topologies for moving massive datasets between nodes, and need infrastructure that can evolve as chip technologies change every 18-24 months.

What AI-ready Infrastructure Actually Means

As a NVIDIA DGX certified partner, EdgeConneX’s infrastructure meets NVIDIA’s reference standards for deploying advanced GPU clusters–power delivery, cooling compatibility, and network architecture. For customers running DGX systems, that reduces technical risk and deployment complexity.

It also signals EdgeConneX has been designing for AI workloads long enough to understand what customers actually need.

That programmability matters because AI workloads in Malaysia span both training–large GPU clusters running for extended periods–and inferencing, which requires lower latency and high throughput for real-time applications. A facility locked into one cooling approach can’t efficiently serve both. Customers need infrastructure that can handle a 30kW CPU rack in one aisle and a 200kW GPU cluster in the next, all within the same data hall.

EdgeConneX’s Ingenuity solution delivers that much needed flexibility. Racks can start at 20-200kW for near-term hybrid workloads and scale to 600kW+ as GPU deployments expand, without major retrofits. The electrical distribution, mechanical systems, and even building design–higher ceilings for liquid cooling infrastructure, reinforced floors for heavier equipment–are engineered to accommodate those changes over 20–30-year facility lifespans.

Flexibility as Infrastructure Strategy

The data-center-as-a-backplane concept addresses a specific customer problem: AI infrastructure changes fast, and facilities that lock customers into rigid configurations create obsolescence risk. For companies deploying AI workloads, change is a given as chip architectures will evolve, cooling approaches will become more efficient, and workloads will move between training and inference.

EdgeConneX’s architecture treats the building as a flexible platform. Power blocks are modular. Cooling configurations are programmable. Electrical distribution is designed to handle future density increases, not just today’s requirements. That means a customer can deploy CPU-based cloud workloads initially at 20-30kW per rack, add GPU clusters for AI training at 100-200kW as their models mature, then scale to 600kW+ as next-generation chips arrive: all without changing facilities.

The approach also future-proofs against technology lock-in. AI chip architectures are evolving rapidly. What works for today’s NVIDIA Rubin clusters may not work for whatever NVIDIA, AMD, or other manufacturers release next. By designing infrastructure that can adapt over multi-decade timelines, EdgeConneX reduces the risk that customers get stranded on obsolete platforms or forced into expensive mid-contract migrations.

EdgeConneX’s deployment experience–gigawatt-scale operations globally, with existing facilities across North America, Europe, and Asia-Pacific–means the company has run this playbook before. The Cyberjaya and Johor sites are implementations of a design philosophy the company has refined across dozens of markets, adapted for Malaysia’s specific technical requirements, resource landscape, and customer demand patterns.

Cyberjaya and Johor: Sustainably Built for Scale from Day One

EdgeConneX’s Cyberjaya and Johor campus’ are both anchored by a 275kV substation, the first of its kind in the greater Kuala Lumpur region. That dedicated electrical infrastructure allows the facility to operate independently from the national grid, critical for reliability at hyperscale and for avoiding the grid strain that’s becoming a constraint in other Southeast Asian markets.

The Cyberjaya substation enables Phase one to come online in Q1 2027 with capacity to scale significantly in the future. The modular power plant design means data halls can be served by one or multiple 12MW blocks, allowing customers to start with lower-density deployments and add capacity as their AI workloads grow. That scalability matters when customers are deploying infrastructure for workloads that may not fully materialize for 12-18 months but need the facility ready when they do.

Johor adds another 180MW, with Phase one slated for 2027. The site is positioned to serve customers who need proximity to Singapore but want Malaysia’s cost structure and regulatory environment. Both facilities adopt a closed-loop, non-evaporative cooling system from the ground up, a design choice that eliminates the water consumption typical of evaporative cooling systems while maintaining competitive power usage effectiveness.

Asia Pacific: Building for the Next Wave of AI Infrastructure

Beyond technical specifications, EdgeConneX operates with predominantly local teams across its Asia-Pacific facilities, supported by regional and global expertise where needed. Community engagement is embedded in the company’s operations model. Initiatives launching in Malaysia in the second half of 2026 will focus on education and local workforce development, building on programs EdgeConneX runs in markets like Dublin, where the company sponsors youth football teams; Atlanta, where it partners with local schools on STEM education; and Silicon Valley, where it supports agricultural restoration projects along the San Joaquin River.

The goal is for data centers to function as community infrastructure, not just commercial real estate. That means hiring locally, engaging with regional suppliers, and ensuring the facilities contribute to the areas where they operate beyond just tax revenue and power consumption.

16.6 Million Liters: Why Cooling Approach Matters

Some traditional data center cooling relies on evaporative systems—large-scale air conditioning towers that use water to dissipate heat. In Malaysia’s hot, humid climate, that creates both a resource constraint and an operational risk as scarcity pressures mount. EdgeConneX’s advanced cooling system enabled by non-evaporative, closed-loop heat rejection technology eliminates that exposure entirely. Cyberjaya saves an estimated 9 million liters of water per day compared to an equivalent water-cooled facility.

Johor saves 7.6 million liters daily. Combined, that’s 16.6 million liters per day not drawn from local water supplies—a meaningful difference in regions where infrastructure development competes with residential and agricultural demand for the same resources. The technical execution uses non-evaporative heat rejection technology and highly efficient chillers capable of handling ultra-dense, liquid-cooled racks, which is the modern gold standard for high-performance computing.

The system rejects heat through air instead of water, whether the data hall is being cooled by air or liquid at the rack level. That allows EdgeConneX to maintain competitive PUE metrics without the water consumption typical of evaporative systems. The approach aligns with broader sustainability commitments. EdgeConneX targets water neutrality across its portfolio by 2030 and has achieved UL 2799 zero-waste-to-landfill certification at facilities including its Jakarta site, which hit a 64% landfill diversion rate.

Malaysia’s Cyberjaya and Johor facilities are slated for UL 2799 certification within 18-24 months of operations startup, following the same operational playbook EdgeConneX uses across its 90+ global data center portfolio. From a competitive standpoint, its cooling philosophy is a differentiator. Most data center operators in Southeast Asia still rely on evaporative systems because they’re proven at scale and well-understood by operations teams.

EdgeConneX understands that customers–particularly hyperscalers facing their own sustainability mandates–will prioritize water efficiency as procurement criteria tighten around environmental impact. The math is straightforward: 16.6 million liters per day is roughly 6 billion liters annually across both sites, a resource footprint that shows up in corporate sustainability reporting and regulatory compliance filings.

What Comes Next

EdgeConneX’s Cyberjaya and Johor facilities are designed to handle the infrastructure demands AI workloads create–ultra-high density, advanced cooling, modular scalability–while operating as community partners in the regions where they’re built. The company’s gigawatt-scale deployment experience and local-first operational model position it to serve customers who need both technical performance and long-term operational reliability.

The first facilities are set to come online in Q1 2027. What customers ultimately deploy within them—training clusters, inference workloads, or hybrid AI-cloud environments—will depend on the pace of their own AI ambitions. EdgeConneX’s role is to stay ahead of that demand: delivering infrastructure that can evolve with changing requirements and operating it in a way that creates lasting value for the communities around each site.

Malaysia’s data center market is entering a period of sustained growth, opening the door for operators to apply their expertise and innovate around the region’s evolving AI infrastructure needs. As demand becomes more specialized, the greatest opportunities will be for providers that can deliver adaptable, high-performance environments designed for the realities of next-generation workloads.