[The Dawn of the Hyper-Scale AI Data Center Era] The World Is Now in a Data Center "War"

Stargate Project Emerges as a Landmark
Ultra-Large-Scale National Strategic Initiative in the US
Big Tech Giants Like AWS and Google Accelerate Their Moves
Japan and Malaysia Also Join the Race

In the state of Texas, USA, the construction of a large-scale data center has begun as part of the Stargate Project. Photo by Reuters Yonhap News

The conventional data centers that formed the backbone of the internet era were built with CPU-based general-purpose servers. However, in the AI era, data centers are centered around graphics processing units (GPUs), utilizing tens of thousands of GPUs. Nvidia's GB200 NVL72 (Blackwell) system consumes over 120 kW of electricity in just one rack. This figure is more than ten times higher than the previous 10 kW-class systems. Large-scale power consumption and heat generation are inherent to their operation. Without a stable power grid, transmission infrastructure, and advanced cooling technologies, it is impossible to run these centers.

To connect tens of thousands of GPUs as one, ultra-fast memory and networking are essential. High Bandwidth Memory (HBM), led by SK Hynix and Samsung, Nvidia’s InfiniBand, and next-generation networking technologies such as Co-Packaged Optics (CPO) serve as the arteries for data flow.

◇ Stargate Project Sparks National Data Center Competition = The most symbolic case in the race for ultra-large AI data centers is the “Stargate” project, announced in January this year. This project, involving OpenAI, Oracle, SoftBank, and the United Arab Emirates’ MGX, drew significant attention when it was unveiled directly by President Donald Trump at the White House. Led by Oracle, the first phase of data center construction has already begun near Abilene, Texas. Upon completion, the project aims to build more than 20 centers, operate over 2 million GPUs, and supply more than 5 GW of power.

Stargate has been elevated to a national strategic initiative. It sets out multiple objectives, including the creation of over 100,000 jobs in the United States, strengthening AI and cloud competitiveness, and expanding global large-scale data centers.

AWS and Google, both strengthening their cloud businesses, are making astronomical investments to secure market leadership. AWS announced investments of $20 billion in Pennsylvania and $10 billion in North Carolina this year alone. Industry experts predict that Amazon’s total capital expenditure in 2025 will reach $100 billion. AWS is also expanding its cloud data infrastructure in Australia, with a planned investment of 20 billion Australian dollars to expand data center infrastructure there by 2029. The Ulsan data center, a joint project between AWS and SK Group, is part of this trend.

Google is investing $7.5 billion in infrastructure this year and will build additional data centers worth $3 billion in Virginia and Indiana. In the Middle East, Saudi Arabia has begun construction of a large-scale data center for the AI company Humane, backed by the Public Investment Fund (PIF). The UAE is also expanding its AI infrastructure with a $1.5 billion project, a collaboration between OpenAI and local firm G42.

◇ Data Center Race Spreads to Japan and Southeast Asia = Japan is attracting global big tech companies with government subsidies and regulatory easing. Microsoft has announced a $2.9 billion investment to expand AI and cloud infrastructure in Japan. In Southeast Asia, Malaysia is aggressively positioning itself as a new AI hub by attracting large-scale data centers, particularly in the Johor region.

AWS, in partnership with SK Group, will build a 100 MW AI data center in the Ulsan Mipo National Industrial Complex. The total investment is about 7 trillion won, with AWS contributing $4 billion. Construction will begin in 2025, aiming to operate 41 MW in the first phase by 2027 and complete 103 MW by 2029. The center will be able to accommodate up to 60,000 GPUs, with the potential for future expansion up to 1 GW.