Generative AI expected to change format and location of future data center deployments
Generative AI has attracted a lot of hype recently due to its capacity to create realistic and creative content. According to an EY-Parthenon analysis, the growth of AI is expected to cause a surge in demand for computing resources and drive demand for data centers.
Today, generative AI engines need hundreds of megawatts to be trained. Yet, many data centers are not well equipped to handle such higher rack densities and the heat generated by AI workloads. AI workloads leverage graphic processing units instead of standard central processing units, which will drastically accelerate the pace of rack densification.
While racks in Southeast Asia today typically only utilize a smaller energy supply — between three kilovolt-amperes (kVAs) and 10kVAs — tomorrow’s AI-enabled racks may need to support rack densities as high as 30kVAs to 50kVAs. This increase will present downstream implications and requirements for stable and secure power and more advanced cooling systems.
Specifically, this will impact the site selection process; instead of central locations, data center site consideration criteria will favor locations with access to stable and secure supplies of electricity and water. This may likely drive future AI-focused data centers’ locations to move from traditional tier-one cities to tier-two cities and city outskirts.
Greater “localization” of data centers expected
Data center hubs first concentrated in Asia-Pacific cities that have major local economies and a high density of submarine cables. As the business landscape in Southeast Asia matures, demand has started to extend beyond the traditional hub markets and new data centers have started to be more “localized”, i.e., sited nearer to their users. While the demand for data center capacity is still largely driven by tier-one or capital cities today, there is growing demand for data center capacity in edge locations in tier-two or tier-three cities that are closer to end users.
Across Southeast Asia, there are three main segments of data center users that drive data center utilization, namely cloud operators, content and gaming companies, and other enterprises. These three different customer segments drive localization at different rates.
Content and gaming companies play a large role in driving localization, with an increasing presence near their end consumers in tier-two and tier-three cities.
Cloud operators are starting to deploy cloud production regions in emerging markets. However, beyond the capital cities of these emerging markets, cloud operator demand for data center capacity is still lagging.
Demand for data center capacity from other enterprises continues to grow with public cloud penetration driving their digitization. Currently, small- and medium-sized enterprises and large enterprises (including finance-related firms) remain largely focused on increasing their data center usage within capital cities. While there are immediate roadblocks to localization, it is inevitable given emerging trends, such as Internet of Things and autonomous vehicles.