Samsung SDS plans to build three hyperscale artificial intelligence data centers across South Korea, committing roughly 3.7 trillion won ($2.8 billion) to expand national AI computing capacity and strengthen domestic cloud infrastructure.
The investment reflects a broader strategic shift as the company positions itself as a sovereign AI infrastructure provider, ensuring critical computing resources remain locally owned and operated. Demand for large-scale GPU infrastructure has surged as enterprises accelerate adoption of generative AI and high-performance computing workloads.
According to industry sources and government agencies on March 11, Samsung SDS is developing AI-focused data centers in Haenam in South Jeolla Province, Gumi in North Gyeongsang Province, and Dongtan in Gyeonggi Province, with each facility expected to support several tens of megawatts of computing capacity.
Together, the projects could exceed 140 megawatts of AI infrastructure, making the initiative one of the most significant private investments in South Korea’s emerging AI computing ecosystem.
National AI Computing Center Anchors the Strategy
First, a consortium led by Samsung SDS recently secured preferred bidder status for the government’s 2.9 trillion-won National AI Computing Center project spearheaded by the Ministry of Science and ICT.
The consortium includes Naver Cloud, Clush, KT Corporation, Kakao, Samsung Electronics and Samsung C&T.
Samsung SDS selected the Solaseado Data Center Park in Haenam as the location for the national facility. The center is expected to deploy roughly 15,000 advanced graphics processing units by 2028, providing computing capacity for domestic AI research and enterprise workloads.
The project’s estimated budget has already risen from more than 2 trillion won to roughly 2.9 trillion won, reflecting the rapidly increasing costs associated with advanced AI infrastructure. Funding will combine public-private investment and policy finance loans.
Industry observers expect the facility to require at least 40 megawatts of power capacity, given that the Solaseado industrial complex was designed to host 25 data center buildings with 40-MW capacity each. Local authorities are also accelerating plans for a 154-kilovolt substation and transmission network, bringing the timeline forward from 2030 to 2028 to support the facility.
“Since the consortium has been selected as the preferred bidder for the National AI Computing Center project by the Ministry of Science and ICT, negotiations will take place over the next two weeks,” an industry official said. “After that, the total project cost and power capacity are expected to be finalized.”
Expansion Beyond Seoul Area With Dongtan and Gumi AI Hubs
Meanwhile, Samsung SDS is expanding its proprietary AI data center network alongside the national project. The company’s 2025 annual report, released March 10, revealed plans to invest about 800 billion won in two additional AI facilities located in Dongtan and Gumi.
The Dongtan high-performance computing data center recently completed construction of a west building expansion. The new facility will begin operations in the first quarter of this year, complementing the east building completed three years earlier. Each structure supports 20 megawatts of power capacity, bringing the site’s total capacity to 40 megawatts.
Samsung SDS had already invested 300.8 billion won in the Dongtan center by the end of last year, leaving approximately 4.9 billion won in remaining capital expenditure, suggesting the project is largely complete. The company is also constructing a second HPC infrastructure hub in Gumi, designed for 60 megawatts of computing capacity. The site will act as a foundation for Samsung SDS’s GPU-as-a-service offering, which forms a core pillar of its cloud portfolio.
A regulatory filing disclosed in January showed the company had initially allocated 427.3 billion won for the project’s construction and equipment. However, the latest annual report revises the total investment to 463.9 billion won through 2029, signaling expanded infrastructure scope.
Enterprise AI Services Drive Infrastructure Spending
Analysts view the data center expansion as a calculated move to strengthen Samsung SDS’s enterprise AI ecosystem.
Revenue from the company’s cloud service provider segment, which includes AI-related offerings, reached 271.3 billion won in the fourth quarter of last year, representing 17% year-over-year growth. Two platforms anchor this strategy: FabriX, Samsung SDS’s enterprise generative AI system, and Brity Copilot, an AI-powered productivity agent designed for enterprise workflows.
The company has also expanded its enterprise distribution strategy through partnerships with global AI providers. Recently, Samsung SDS became an official reseller of OpenAI’s ChatGPT Enterprise, offering the service bundled with the FabriX platform to corporate customers.
Consequently, the company expects AI workloads running on its infrastructure to scale quickly as enterprise adoption accelerates.
Infrastructure Strategy Aligns With Long-Term AI Growth
Analysts believe the strategic value of Samsung SDS’s data center network will grow as Brity Copilot evolves into a broader personal-agent platform. The system already integrates five agent-driven features, including an interpreting agent capable of real-time foreign-language translation.
At the same time, financial analysts anticipate improving capital efficiency despite the scale of the investment. According to projections from Samsung Securities, Samsung SDS’s return on invested capital could rise from 15.0% in 2025 to 17.8% in 2026 and 22.4% in 2027.
Therefore, the company’s expanding network of AI data centers may become a cornerstone of South Korea’s emerging sovereign computing infrastructure as demand for domestic AI capacity accelerates.
