SK, GS, and Naver Join Government-Led Plan to Invest Over 1,000 Trillion Won in Nationwide AI Data Centers
June 30, 2026
SK, GS, and Naver Join Government-Led Plan to Invest Over 1,000 Trillion Won in Nationwide AI Data Centers
South Korea is embarking on one of the world’s most ambitious artificial intelligence infrastructure projects, with the government and major private conglomerates pledging over 1,000 trillion Korean won (approximately $720 billion) to build massive AI data centers across the country by 2035. The initiative, announced at a national briefing session held at the presidential office Cheong Wa Dae, positions AI data centers as a cornerstone of the nation’s future economic strategy, alongside semiconductor manufacturing and digital talent development.
The project brings together SK Group, GS Group, and Naver, which plan to begin operating data centers with a combined capacity of 8.4 gigawatts (GW) starting in 2029. The initial phase will focus on three key regions: Ulsan, Donghae, and Sejong. SK will build a 1GW facility in Ulsan, GS a 2.4GW center in Donghae, and Naver a 1GW center in Sejong. SK also plans to review additional sites in regional clusters such as the central region, Daegu-Gyeongbuk, Honam, and Gangwon, with the goal of constructing an additional 1GW data center in each area. The government aims to accelerate site selection and permitting to begin construction in the first half of 2028, with phased operations commencing the following year. By 2030, South Korea’s total AI data center capacity is expected to reach 14.37GW, accounting for approximately 25% of the Asia-Pacific region’s projected 58GW capacity. SK Group Chairman Chey Tae-won confirmed the scale of the commitment, stating, "We plan to invest approximately 1,000 trillion Korean won in the AI data center project." SK separately aims to expand its own AI data center portfolio to 15GW by 2035.
Unlike conventional data centers used primarily for storage and cloud services, these facilities are designed for the intensive training and inference of large AI models. They integrate cutting-edge hardware such as GPUs, high-bandwidth memory (HBM), servers, and advanced power and cooling systems. The government views this as a strategic opportunity to export a bundled "AI data center solution" that combines domestically developed neural processing units (NPUs), servers, networking equipment, and power and cooling technologies. An industry source noted, "While South Korea struggles to catch up in AI model development, the government likely sees global competitiveness in the AI data center sector." The economic ripple effects are expected to be substantial; a single 1GW AI data center requires 60 to 70 trillion won in investment and is projected to create 5,000 daily jobs during construction, with 1,000 workers needed for ongoing operations.
However, the project faces significant challenges. AI data centers consume tens of times more electricity than traditional facilities, and operating the initial 8.4GW of capacity would require the output of six new nuclear reactors. Large-scale cooling water is also essential due to the intense heat generated by AI servers. Local opposition, rising electricity costs, and water shortages have already delayed similar projects in the United States. GPU supply remains another critical bottleneck, as surging demand for Nvidia’s processors has led to shortages and rising costs. Lee Byeong-hoon, a professor at Pohang University of Science and Technology, warned, "Stable AI semiconductor supply is critical. South Korea should develop domestic NPUs to reduce reliance on overseas suppliers." An AI infrastructure industry source added a note of caution about the project’s long-term viability: "A virtuous cycle where companies profit from AIDCs and reinvest is essential for the project to proceed as planned. Without a clear revenue model, companies might scale back or halt investments midway."
Source: chosun
