SK Group's data centre arm eyes $350 billion AI infrastructure buildout

A massive infrastructure blueprint

SK Hyper, a new data centre unit of the South Korean conglomerate SK Group, is beginning a massive expansion of artificial intelligence infrastructure. The company is currently in talks with financial investors and major technology firms to fund a plan that aims to establish 15 gigawatts (GW) of data centre power in South Korea by 2035.

The scale of this project is significant, as the first phase involving 5 GW of capacity alone is expected to cost US$350 billion. This figure includes the cost of all necessary equipment required to power the facilities. To manage this enormous capital requirement, SK Hyper plans to share the overall costs with outside investors and the actual tenants who will use the data centres.

The company is not planning to commit to the entire buildout at once. Instead, SK Hyper intends to adjust the pace of its investment based on actual market demand. This strategy is designed to manage risk while ensuring that the infrastructure keeps up with the growing needs of the AI industry.

Securing capital and partnerships

SK Hyper is looking to establish a steady flow of funding to realize its ambitious goals. The firm intends to gather its initial portion of outside funding by the end of 2026 or the start of 2027. Following that initial milestone, the firm expects to conduct further fundraising over the next three to four years to sustain the expansion.

The company is currently engaging with a variety of strategic and financial partners to build this capital base. Chung Suk-geun, the chief executive officer of SK Hyper, stated that the company wants to be prepared for meeting the market demand by making data centres in Korea. These partnerships are essential for providing the liquidity needed for such a large-scale industrial project.

The group's internal resources also play a major role in the financial and operational viability of the project. SK Telecom, a major affiliate, has already committed to investing 750 billion won (US$567 million) into SK Hyper by 2030. This internal backing provides a foundation of stability as the company seeks more external investment from global technology players.

Leveraging group-wide expertise

One of the primary challenges for data centre operators is the high cost of electricity required to run massive computing clusters. SK Group plans to use its diverse portfolio of affiliates to offset these costs. By utilizing the expertise of its various companies in the fields of chips, energy, and telecommunications, the group aims to improve computing efficiency across its new facilities.

This integrated approach allows SK Hyper to offer a more competitive package than a standalone data centre provider might be able to offer. By optimizing how energy is used and how chips are managed, the group hopes to make the high-cost environment of AI training more sustainable for its clients. This synergy is a key part of the company's strategy to attract global technology firms to South Korea.

The company is also looking beyond the domestic market to find new revenue streams. SK Hyper intends to position South Korea as a base for AI training and operations to serve broader Asian markets, specifically mentioning Japan and Taiwan. Additionally, the company is exploring opportunities to offer integrated packages of computing capacity and AI models to governments in the Middle East and other parts of Asia.

Current projects and timeline

While the 2035 goal is long-term, SK Hyper is already moving forward with several immediate projects. The company is currently constructing a 900 megawatt (MW) site in Ulsan, a city in the southern part of the country. This is happening alongside a previously announced 100 MW project specifically for Amazon Web Services, one of the world's largest cloud providers.

To expand its footprint, SK Hyper is also in active discussions with the South Korean government to secure additional sites for development. These negotiations are part of the broader effort to establish the country as a hub for high-performance computing. The company is positioning itself to be a primary provider for both private enterprise and public sector needs.

The timeline for the new unit is aggressive, with the company aiming to begin generating revenue as early as 2027. This would occur ahead of the planned completion of the initial data centres in 2029. By starting revenue generation early, SK Hyper hopes to establish its market position and prove the viability of its infrastructure model to its growing list of partners.

Navigating the AI bubble fears

SK Hyper is launching this initiative during a period of significant market uncertainty regarding the future of artificial intelligence. Many investors have expressed concerns about an AI bubble, worrying that the trillions of dollars being spent on global computing buildouts may not be justified by future service demand. These fears have made it more difficult for infrastructure projects to secure easy funding.

The company's strategy of adjusting investment pace based on demand is a direct response to these market gyrations. By not committing to the entire plan upfront, SK Hyper seeks to protect itself from over-investing in capacity that might not be utilized. This cautious approach is intended to balance the need for rapid growth with the reality of a fluctuating economic environment.

Despite these concerns, the move highlights the belief that high-quality, efficient infrastructure will remain a necessity for the AI era. By focusing on regional demand and leveraging internal technical advantages, SK Hyper is attempting to create a stable model for data centre development. The success of this plan will depend on its ability to attract the necessary capital while maintaining a flexible approach to construction.