As extra firms in Asia Pacific undertake synthetic intelligence to spice up their operations, the stress on knowledge centres is rising quick. Conventional services, constructed for earlier generations of computing, are struggling to maintain up with the heavy power use and cooling calls for of contemporary AI methods. By 2030, GPU-driven workloads might push rack energy densities towards 1 MW, making incremental upgrades not sufficient. As an alternative, operators are actually turning towards purpose-built “AI manufacturing unit” knowledge centres which can be designed from the bottom up.
AI Information spoke with Paul Churchill, Vice President of Vertiv Asia, to higher perceive how the area is getting ready for this shift and what sorts of infrastructure adjustments lie forward.
Explosive market development is setting the tempo
The AI data-centre market is projected to surge from $236 billion in 2025 to almost $934 billion by 2030. This development is pushed by speedy adoption of AI in industries like finance, healthcare, and manufacturing. These sectors depend on high-performance computing environments powered by dense GPU clusters, which require way more power and cooling capability than conventional servers.
In Asia Pacific, this demand is amplified by authorities investments in digitalisation, the enlargement of 5G, and the rollout of cloud-native and generative AI functions. All of that is pushing compute wants increased at a tempo the area has by no means seen earlier than.
Churchill defined that assembly this demand requires extra than simply bigger services. It requires smarter infrastructure methods which can be scalable and sustainable. “Infrastructure leaders should transfer past piecemeal upgrades. A future-ready technique includes adopting AI-optimised infrastructure that mixes high-capacity energy methods, superior thermal administration, and built-in, scalable designs,” he stated.
Cooling and energy challenges are rising
As rack densities enhance from 40 kW to 130 kW, and doubtlessly as much as 250 kW by 2030, cooling and energy supply have gotten vital points. Conventional air cooling strategies are not sufficient for these situations.
To handle this, Vertiv is growing hybrid cooling methods that blend direct-to-chip liquid cooling with air-based options. Programs can alter to altering workloads, cut back power use, and keep reliability. “Our coolant distribution models allow direct-to-chip liquid cooling whereas making certain reliability and serviceability in high-density environments,” Churchill stated.
Energy supply can also be turning into extra complicated. AI workloads fluctuate quickly, so infrastructure must react in actual time. Vertiv is evolving its rack energy distribution models and busway methods to deal with increased voltages and enhance load balancing. Clever monitoring helps operators handle hundreds extra effectively, cut back wasted capability, and prolong uptime – a key consideration in components of Southeast Asia the place energy grids are much less secure.
Information centres are being redesigned for AI
The rise of liquid-cooled GPU pods and 1 MW racks, like these deliberate by AMD and hyperscalers equivalent to Microsoft, Google, and Meta, indicators a deeper architectural shift. As an alternative of retrofitting older services, new knowledge centres are being designed particularly to assist AI.
“The way forward for data-centre structure is hybrid, and these infrastructures require services to be constructed round liquid move,” Churchill stated. This consists of new ground layouts, superior coolant distribution, and extra subtle energy methods.
The following-generation services will combine cooling, energy, and monitoring from the chip stage to the grid. For Asia Pacific, the place hyperscale campuses are increasing quickly, this sort of built-in design is important to maintain up with efficiency expectations and sustainability targets.
From incremental upgrades to AI manufacturing unit knowledge centres
By 2030, Asia Pacific is anticipated to overhaul the US in knowledge centre capability, reaching virtually 24 GW of commissioned energy. To deal with this development, enterprises are transferring away from advert hoc upgrades towards full-stack AI manufacturing unit knowledge centres.
Churchill stated this transition ought to occur in phases. Step one is built-in planning, bringing collectively energy, cooling, and IT administration somewhat than treating them as separate methods. The method simplifies deployment and supplies a powerful base for scaling.
The second step is to undertake modular and prefabricated methods. These enable firms so as to add capability in phases with out main disruptions. “Corporations can deploy factory-tested modules alongside present infrastructure, steadily migrating workloads to AI-ready capability with out disruptive overhauls,” he stated.
Lastly, sustainability have to be constructed into each stage. This consists of utilizing lithium-ion power storage, grid-interactive UPS methods, and higher-voltage distribution to enhance effectivity and resilience.
DC energy positive aspects new relevance for AI knowledge centres
Vertiv not too long ago launched PowerDirect Rack, a DC energy shelf designed for AI and high-performance computing. Switching to DC energy can reduce power losses by lowering the variety of conversion steps between the grid and the server. It additionally aligns with renewable power and battery storage methods, which have gotten extra widespread in Asia Pacific.
That is particularly helpful in energy-constrained markets like Vietnam and the Philippines. In these areas, versatile energy options are important to maintain services operating easily. As Churchill famous, DC energy is “not simply an effectivity play – it’s a technique for enabling sustainable scalability.”
Sustainability is turning into a central precedence
With AI driving up power use, data-centre operators are dealing with stricter rules and rising grid constraints. That is notably true in Southeast Asia, the place energy reliability and tariffs range broadly.
Vertiv is working with operators to combine different power sources like lithium-ion batteries, hybrid energy methods, and microgrids. These can cut back dependence on the grid and enhance resilience. There may be additionally rising curiosity in solar-backed UPS methods and superior power storage applied sciences, which assist steadiness hundreds and handle prices.
Cooling effectivity is one other main focus. Hybrid liquid cooling methods can cut back each power and water use in comparison with older strategies. “Our focus is on delivering infrastructure that meets efficiency calls for whereas aligning with ESG targets,” Churchill stated. “We’re collaborating with our companions to make sure that AI-driven development within the area stays accountable, sustainable, and aligned with long-term digital and environmental targets.”
Modular options assist speedy enlargement
Many rising economies in Asia Pacific face challenges like restricted land, unstable energy provide, and shortages of expert labour. In these settings, modular and prefabricated data-centre methods provide a sensible resolution.
Prefabricated modules can reduce deployment instances by as much as 50%, whereas bettering power effectivity and scalability. They permit operators to increase steadily, including capability as wanted with out heavy upfront funding. The flexibleness is very invaluable for AI workloads, which might develop shortly and unpredictably.
By combining compact design with energy-efficient operation, modular methods give operators a option to construct AI-ready capability quicker and with much less threat – an important benefit because the area’s digital economies develop.
Making ready for a demanding future
The AI surge is reshaping how knowledge centres are constructed and operated in Asia Pacific. As workloads intensify and sustainability pressures mount, firms can not depend on outdated infrastructure. The transfer towards AI manufacturing unit knowledge centres, powered by superior cooling, DC energy, and modular methods, displays a shift in how the area is getting ready for the subsequent period of computing.
(Photograph by İsmail Enes Ayhan)
Need to study extra about AI and massive knowledge from business leaders? Try AI & Big Data Expo happening in Amsterdam, California, and London. The great occasion is a part of TechEx and co-located with different main expertise occasions. Click on here for extra data.
AI Information is powered by TechForge Media. Discover different upcoming enterprise expertise occasions and webinars here.
