Across the Asia Pacific, AI workloads are pushing rack densities, which averaged 8 to 40 kilowatts (kW), are now climbing to 130 to 600 kW, with projections of 1 MW per rack around 2028 to 2029. The old model of homogeneous, globally replicated facilities can no longer keep up with these extreme demands.
Growth figures tell a clear story. The APAC region added nearly 2,300 MW to its data centre development pipeline in the first half of 2025. Operational capacity stands at approximately 12.7 gigawatts, with 3.2 gigawatts under construction and an additional 13.3 gigawatts in planning stages, as per the Cushman & Wakefield report.
Yet this growth is deeply uneven. Operators face distinct regulatory frameworks, power constraints, and environmental conditions from one market to the next.
Johnson Controls’ vice president and general manager of data centre solutions, Austin Domenici, told iTnews Asia that operators need to design for regional realities.
Whether it is power availability in Jakarta, water scarcity in Mumbai, or land constraints in Tokyo, the ability to localise infrastructure strategies while maintaining global standards is a competitive advantage, said Domenici.
Regional realities demand localised design
Asia Pacific is not one market but a network of very different ones. What unites them is the acceleration of AI and cloud adoption; what separates them is the operational context in which those workloads must run.
Singapore represents the most policy-driven approach.
According to Domenici, the government’s Green Data Centre Roadmap, SS 715:2025 energy efficiency standards, and the Tropical Data Centre Standard are driving adoption of waterless cooling, higher chilled water setpoints, and energy-efficient IT equipment.
Singapore will be adding at least 300 megawatts of new data centre capacity in the coming years, with another 200 megawatts reserved for operators adopting green energy solutions, said Domenici.
This cannot be merely considered as a regulation, but a deliberate national strategy to keep data growth aligned with sustainability goals.
India is scaling rapidly, with growth driven by hyperscalers and corporates. Bangalore and Chennai are emerging as key hubs.
Power availability and water scarcity remain key constraints, pushing operators to experiment with modular designs and localised cooling systems that can perform efficiently in high-temperature environments.
Japan is balancing high demand with land constraints and seismic resilience, supported by continued foreign capital investment, he added.
Continued foreign investment shows that international players are willing to adapt to Japan’s design norms in exchange for access to its stable, high-value market.
Each of these markets is solving the same underlying problem of managing higher-density computing, but through different architectural and policy frameworks.
Operators that can tailor infrastructure to local conditions while meeting global performance and ESG benchmarks will lead the next phase of regional expansion.
A new infrastructure model is required to cope with higher-density workloads
The rapid rise of AI workloads is not just increasing demand but is compressing timelines and testing the limits of traditional infrastructure models.
In APAC, where operational capacity stands at approximately 12.7 gigawatts and an additional 13.3 gigawatts is in planning, the pressure to deliver performance at scale is intensifying.
Domenici said, “The most significant challenges emerge not from individual system limitations, but from the lack of coordination between them.”
In high-density environments, cooling, power, and controls must operate as a unified ecosystem. When they do not, even small inefficiencies can escalate into systemic vulnerabilities.
What we are seeing is a shift away from siloed systems toward integrated, lifecycle-based approaches. Operators are investing in platforms that unify cooling, controls, and energy management to enable real-time visibility and continuous optimisation.
- Austin Domenici, Vice President and General Manager, Data Centre Solutions, Johnson Controls
This is not just about meeting ESG targets, but it is about building infrastructure that can adapt to unpredictable workloads, reduce non-IT energy consumption, and support long-term sustainability goals without compromising uptime, he added.
Across Southeast Asia, the urgency is clear. Cooling already accounts for up to 40 percent of total energy use in many facilities.
By 2027, analysts expect that half of IT buyers in the Asia Pacific will work only with vendors that meet ESG and responsible AI procurement criteria.
The most effective operators are adopting integrated lifecycle management.
Instead of reactive maintenance, they rely on continuous diagnostics, predictive analytics, and automation to manage workloads dynamically, said Domenici.
He added that this approach allows for responding to demand spikes, maintaining uptime, and managing energy use without compromising operational continuity.
Talking about Johnson Controls, Domenici said the company’s approach integrates modular thermal management, building controls, fire protection, and facility security and service into a unified lifecycle strategy.
He added that the research and development work at the Johnson Controls Advanced Development Engineering Centre (JADEC) helps operators shift from reactive upgrades to proactive infrastructure planning.
These include cooling systems with lower connected load ratings and full lifecycle support models that reduce deployment complexity and improve time to service.
Meanwhile, with innovative solutions like Coolant Distribution Unit (CDU), non-IT energy consumption can be reduced by nearly 50 percent even in the Asia Pacific data centre hubs with the warmest climates, said Domenici.
Toward scalable resilience and regional sustainability
Looking ahead, the question is how to keep scaling without rebuilding entire facilities, and the answer lies in modularity and efficiency.
Architectural flexibility allows operators to add or reconfigure capacity within live environments. Modular cooling units, prefabricated power modules, and distributed control systems enable meeting AI-driven loads without long construction delays.
The sustainability stakes are equally significant.
The International Energy Agency projects that by 2030, AI-optimised data centres could consume more electricity than Japan does today.
Water use is another constraint; in some Asian cities, it is already among the most limited resources.
According to Domenici, Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE) ratings are a primary consideration.
Efficient data centres can achieve PUE ratings between 1.2 and 1.4, compared to the global average of 1.58. For water, consumption can be as low as 0.2 litres per kilowatt hour of IT power, compared to a global average of 1.8 litres, said Domenici.
These benchmarks show that improvements are possible, especially in regions where power and water are constrained, he added.
Citing an example, Domenici mentioned Johnson Controls’ YORK YVAM system, which leverages magnetic-bearing technology and delivers effective cooling with 40 percent less energy and zero water consumption.
As Southeast Asia becomes a focal point for AI infrastructure, with the Singapore–Johor–Batam corridor already a major hub and Malaysia, Thailand, and Vietnam attracting significant investment, the ability to deploy repeatable, resilient, and regionally adapted systems will be critical.
However, Domenici mentioned resilience alone is no longer enough.
He added that the next phase of growth must consider how infrastructure contributes to the communities it operates in.
That includes minimising disruption, designing for low water and energy impact, and exploring opportunities for waste heat reuse and district energy integration.
The future of infrastructure will be defined not just by how it performs under pressure, but by how it supports broader sustainability and liveability goals, Domenici said.
