The Data Center Pillar of South Korea's Megaproject
South Korea's Science Minister Bae Kyung-hoon announced on June 29, 2026 that the country will build AI data center capacity exceeding 18.4 gigawatts by 2035, backed by total investment of more than 1,000 trillion won ($648 billion), per Al Jazeera and Reuters reporting. The data center initiative is one of three concurrent "triple axis" announcements: Samsung Electronics and SK Hynix are committing 800 trillion won ($518B) for new chip fabs in the southwest; a further 81 trillion won targets a chip-packaging hub near Seoul; and the Science Ministry is leading the data center build. President Lee Jae Myung framed the combined initiative as a national "great leap" - "We must secure the core elements of AI faster than any other country."
Investment Structure and Timeline
The initial phase is concrete: SK Group, GS Group, and Naver have pledged 550 trillion won ($356B) to deliver 8.4 GW of AI data center capacity by 2029. The 2035 target adds a further 10 GW to reach 18.4 GW total, with cumulative investment crossing 1,000 trillion won. The capital is private-sector commitments aligned with government targets, not direct state expenditure - execution depends on power permitting, land availability, and sustained corporate investment through the decade.
Practitioner Signal: Scale and Vertical Integration
18.4 GW dedicated to AI workloads in a single country is at the outer edge of sovereign compute ambition globally. The strategic logic is vertical integration: Samsung and SK Hynix together supply roughly 80% of global high-bandwidth memory (HBM) production - the chip architecture that feeds AI accelerators at training and inference speed. Korean-hosted data centers drawing on domestic HBM supply chains creates a tightly coupled AI infrastructure stack. For practitioners tracking GPU and HBM supply chains, sustained Korean demand at this scale is a multi-year demand signal for AI accelerators and advanced memory.
Political Context and Execution Risk
The broader megaproject carries political controversy. The opposition has criticised President Lee's decision to locate chip fabs in Honam, the region where 85% of voters backed him in last year's election, accusing the government of industrial patronage over commercial logic. Lee rejected this on social media over the weekend. The data center component does not face the same geographic critique, but it shares the political environment of the wider announcement. As with any decade-long national infrastructure pledge, the gap between commitment and deployed gigawatts will depend on regulatory execution, power grid capacity, and whether private-sector partners sustain investment through market cycles.
Key Points
- 1What: South Korea's Science Ministry targets 1,000 trillion won ($648B) total in AI data centers by 2035, starting with 550T won from SK Group, GS Group, and Naver for 8.4 GW capacity by 2029.
- 2Why: President Lee frames the initiative as a national race - 'securing core elements of AI faster than any other country' - tied to vertically integrating Korean HBM chip strength with domestic compute.
- 3So what: A pledged 18.4 GW buildout reshapes AI accelerator and HBM demand signals, power market dynamics, and sovereign compute competition in Asia-Pacific through the decade.
Scoring Rationale
One of the largest sovereign AI data center pledges globally - 1,000 trillion won ($648B) targeting 18.4 GW by 2035 - with concrete initial commitments from SK Group, GS Group, and Naver. Multi-year HBM and compute demand signal for AI infrastructure practitioners. Score tempered by 2029-2035 horizon and the public-private pledge-vs-deployed-capital gap.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems