SGC Energy bets on on-site power to build South Korea's first AI data center of its kind

Interview with Hwang Se-hun, senior managing director of SGC Energy's data center development task force

Company plans to break ground on a 60 MW facility in Gunsan in the fourth quarter

Aims to leverage existing energy expertise to expand into AI infrastructure

On-site power supply and precast construction method seen as key competitive advantages

Executive says success in Gunsan could open the door to overseas expansion

The Herald Business takes a closer look at the companies shaping South Korea's industrial landscape — tracking splits, mergers, rebrands, new ventures and businesses being wound down — through on-the-ground reporting and data analysis.

SGC Songam Building, headquarters of SGC Energy. [SGC Energy]

"In an AI data center, the most expensive power is not power with a high unit price — it is power you cannot secure when you need it. The capabilities in power generation and plant operations that SGC Group has built over the years turn out to be the most critical competitive advantage in the AI data center business."

황세훈 SGC에너지 데이터센터 개발 TFT 상무

The global AI boom is rapidly reshaping the data center market. Where data centers were once the exclusive domain of IT companies, the ability to manage the enormous power consumption and heat generated by high-performance GPU servers has emerged as the decisive factor in whether a project succeeds or fails.

Against that backdrop, SGC Energy, a comprehensive eco-friendly energy company, has moved to build a large-scale AI data center in Gunsan, North Jeolla Province. At its annual general meeting in March, the company announced plans to restructure its business around AI data centers as a new growth engine to strengthen its long-term competitiveness. On Friday, The Herald Business sat down with Senior Managing Director Hwang Se-hun of SGC Energy's data center development task force at the company's headquarters in Yeomgok-dong, Seocho-gu, Seoul, to hear about the company's innovative AI data center model and its vision for the future.

Hwang Se-hun, senior managing director of SGC Energy's data center development task force. [SGC Energy]

SGC Energy pursues 300 MW AI data center cluster in Gunsan, banking on power plant synergies

SGC Energy plans to launch the Gunsan AI data center project at 60 MW and expand it in stages to a maximum of 300 MW. The company will build the data center in a modular format on a site it has already secured. It aims to complete the building permit and construction notification for the 60 MW facility by the end of the third quarter, with full groundbreaking scheduled for the fourth quarter.

On Tuesday, SGC AI Infra — SGC Energy's dedicated AI data center subsidiary — signed an investment agreement with North Jeolla Province, Gunsan city and the Korea Industrial Complex Corp. at the North Jeolla Provincial Government building in Jeonju. Korea Investment & Securities also attended the signing ceremony as a financial investor and lead arranger.

SGC Energy's move into AI data centers is not a diversification away from its core business, Hwang said. He described it as "a strategic choice to extend our existing energy competitiveness into the infrastructure business of the AI era," adding that the company aims to secure a new growth engine while maintaining the stability of its power generation operations.

Overview of SGC Energy's AI data center project in Gunsan, North Jeolla Province. [SGC Energy]

In the era of cloud computing and AI, the volume of data and power that must be processed has grown exponentially, making it impossible for AI data centers to operate without the support of infrastructure companies capable of delivering stable power and thermal management. "SGC Energy has the capability to build and operate power plants in the hundreds of megawatts range, and the plant construction expertise of our subsidiary SGC E&C adds further synergy," Hwang said.

He also credited the strong commitment of SGC Energy Chief Executive Lee Woo-sung as the driving force behind the project's rapid progress. In an energy industry that tends toward conservatism, the company's culture of continuous change — including the adoption of wood pellets as a renewable energy source and co-firing power generation — has fueled the push into new businesses and allowed the company to accelerate its contribution to the national AI transformation agenda, he said.

South Korea's first on-site power supply for a data center — and a precast construction breakthrough

The biggest advantages of SGC Energy's AI data center lie in what the company calls innovations in power and "air" — industry shorthand for construction time. Transmission line constraints and power shortages have effectively halted new data center construction in the Greater Seoul area. SGC Energy is pursuing what it says is South Korea's first on-site approach, directly connecting a power plant to a data center on a large plot within the Gunsan industrial complex.

The most important benefit of the on-site model is not simply lower power costs but the certainty of securing large-scale power reliably. Because any delay or interruption in power supply is itself a major business risk for an AI data center, having generation infrastructure on-site — and being able to draw on both self-generated power and the Korea Electric Power Corp. grid — is a structural advantage, Hwang said.

"If we supply electricity directly through a dedicated power plant rather than relying solely on the KEPCO grid, we can deliver high-quality power stably without voltage fluctuations caused by external factors," he said. "There is also a price advantage." The approach mirrors what global technology giants such as Microsoft and Oracle have already adopted, and SGC Energy aims to be the first to implement it in South Korea.

The company is also innovating in construction methods to keep pace with the rapidly shifting AI chip technology cycle. Data centers in the Greater Seoul area have traditionally been built as high-rise structures because of expensive land costs, requiring three to five years to complete. "SGC Energy is applying a single-story precast — modular — construction method," Hwang said. "Modules are manufactured to standardized specifications and then mounted on-site."

This approach will dramatically reduce project financing risks such as load-bearing issues and construction delays. It also cuts the construction period to around 15 months, and the company is pushing ahead with a target of commissioning the facility in the third quarter of next year.

A rendering of SGC Energy's AI data center in Gunsan, North Jeolla Province. [SGC Energy]

Gunsan positioned as a hub for power, networks and talent

SGC Energy chose Gunsan for its AI data center because the city offers a combination of power availability, land, network connectivity and accessibility. "Because the data center will be located within Gunsan's second national industrial complex, large plots are easy to use and a sufficient buffer from residential areas can be maintained, which relatively lowers the risk of community opposition during development," Hwang said. "It also aligns with the current government's policy of fostering semiconductors and AI in the Jeolla region."

He added that a submarine international fiber-optic cable landing station is being developed near the Gunsan site, with a target of entering global service in 2029. "As international network accessibility improves, we expect it to be a positive factor not only for AI data centers but also in attracting demand for cloud data centers and content delivery networks," he said. Improved rail connectivity between the Greater Seoul area and Gunsan is also anticipated, which is expected to help address one of the persistent challenges for regional data centers: securing skilled operations staff.

100% pre-leased and eyeing a 300 MW mega-cluster — then the world

SGC Energy is currently in talks with KT Corp's cloud unit on lease rates and detailed contract terms, on the basis of 100 percent pre-leasing under a long-term master lease structure. "I believe we were able to secure a tenant in advance because of the site's ability to guarantee large-scale power, the shortened construction timeline and ease of expansion enabled by modular design, and our years of experience operating power generation facilities," Hwang said. "The most important thing is not to build the data center first and then look for customers, but to execute the project with demand already secured," he added.

Hwang closed the interview by outlining his ambitions beyond the domestic market. With power shortages driven in part by AI data center demand affecting countries around the world, he said collaboration between companies could open a clear path to overseas expansion. "Based on the experience we gain from success in Gunsan, I believe we can take this business model overseas in partnership with IT companies and others," he said.

keg@heraldcorp.com