[Tae-soon Lee (이태순), head of Vertiv Korea] As AI technology rapidly spreads across daily life and business, the need for AI data centres that are energy-efficient and environmentally sustainable is growing. AI data centres are now becoming essential infrastructure for digital transformation across industries, beyond simple IT infrastructure.

According to IDC, the Asia-Pacific (excluding Japan) data centre market is expected to post 22 percent average annual growth from 2024 to 2029. With demand for AI, cloud and digital services surging, hyperscalers and cloud providers are expanding data centres at an unprecedented pace.

As large-scale data centre construction and expansion projects follow one after another across India, Malaysia, Japan and Southeast Asia, this growth is accelerating further. Data centre operators are reshaping regional digital infrastructure by building high-density, scalable, AI-ready data centres to meet next-generation computing demand.

If companies can accurately grasp global AI data centre market trends, they will be able to pursue effective AI strategies based on them. Vertiv, which provides leading power and cooling solutions for AI data centres, has been sharing on-the-ground insights through its Vertiv Frontiers report. In its latest report, it introduced five trends that can gauge the future of AI data centres.

First, power infrastructure is becoming more advanced. Most data centres currently use a mixed AC and DC power distribution approach, but they are hitting limits as AI workloads increase and power density rises sharply. Unlike AC, which goes through multiple conversions and leads to energy loss and lower efficiency, high-voltage DC can reduce current, shrink conductor size and minimise conversion stages. As rack density rises further, the adoption of high-voltage DC is expected to expand. This shift is expected to accelerate as more on-site power generation facilities and microgrids are built.

Second, AI environments are becoming more distributed. While existing AI data centres have supported large language models and focused mainly on providing general-purpose AI services, they are expected to provide a range of services based on inference and tailored to company-specific requirements. As a result, in heavily regulated industries such as finance, defence and healthcare, the likelihood of choosing on-premises or hybrid AI environments is increasing due to issues such as data sovereignty, security and latency. As distributed AI environments expand, the importance of flexible, scalable high-density power and liquid cooling solutions is also increasing.

Third, energy strategies, including on-site power generation, are accelerating. AI data centres prefer the power grid to secure stable electricity, but they are increasing short-term on-site generation due to practical constraints. As power supply fails to keep up with surging demand, companies are seeking long-term energy self-sufficiency strategies centred on AI data centres. To address power availability problems, investment continues in on-site generation technologies such as natural gas turbines, DC microgrids and nuclear power (SMR). 'Bring Your Own Power' is expected to take hold as a major strategy going forward.

Fourth, the use of digital twins is spreading. As complexity increases with the adoption of highly integrated AI workloads and high-performance GPUs, the need to build and operate data centres quickly is growing. In this situation, using AI-based digital twin technology makes it possible to design and simulate data centres and IT infrastructure in a virtual environment and to build them as an integrated deployment through a modular prefabricated approach. That can significantly shorten the time required to implement AI services. It is expected to play a key role in improving ROI and efficiency as AI data centres scale to gigawatt-class capacity.

Fifth, adaptive, high-reliability liquid cooling technology is also evolving. As heat generation rises with the spread of AI workloads, liquid cooling is becoming an essential element for more data centres. Liquid cooling technology efficiently cools AI data centres and enables AI services, and AI technology is also being used to further advance the efficiency and stability of liquid cooling systems. Through AI-based prediction and control technology, potential failures can be detected in advance and fluids and components optimised. This is expected to further strengthen the stability and uptime of hardware and data workloads.

As AI technology rapidly evolves and expands, the AI data centre market is also changing quickly. These changes are being intensified by factors including accelerating ultra-high density, competitive gigawatt-class expansion, the perception of the data centre as a single computing unit and silicon diversification. At a time when information on AI technology and markets is pouring out like a flood, accurately understanding trends in the data centre market and establishing and executing AI strategies suited to a company’s goals and circumstances will be one key to securing clear competitiveness in the AI era.