South Korea is making a massive bet on AI infrastructure by partnering with NVIDIA to build gigawatt-scale data centers that will serve enterprise, industrial, and government clients worldwide. Naver, South Korea's largest internet company, is leading the effort alongside memory chip manufacturer SK Hynix and construction firm Doosan Group. The first facility launches in early 2027, with plans to expand from 55 megawatts to 1,000 megawatts of computing capacity.

Why Is South Korea Building These AI Factories Now?

Jensen Huang, NVIDIA's CEO, visited South Korea in June 2026 and locked in three major partnerships simultaneously. The timing reflects what NVIDIA calls "surging global demand" for AI services, physical AI systems, robots, and self-driving cars. Naver already had momentum entering these negotiations; in January 2026, the company completed South Korea's largest AI computing cluster, packing 4,000 NVIDIA B200 Blackwell GPUs into a single system designed for large-scale AI workloads.

South Korea's geographic advantage is significant. SK Hynix and Samsung together manufacture two of the world's three largest memory chip supplies, and memory chips are essential for AI computers to function. Having these manufacturers within the country gives South Korea a major competitive edge that most other nations lack.

What Are the Three Companies Contributing to This Project?

• Naver: Building and operating the Gak Sejong data center, starting with a 55-megawatt expansion and scaling to gigawatt capacity using NVIDIA's DSX platform, targeting enterprise and government clients
• SK Hynix: Developing next-generation memory chips purpose-built for AI workloads and signing multi-year contracts; will also use NVIDIA's CUDA-X libraries and PhysicsNeMo framework to accelerate chip simulation before physical production
• Doosan Group: Handling physical infrastructure including data center construction, supplying components for NVIDIA's Blackwell chips, and providing energy solutions for the platforms

This division of labor reflects the scale of the undertaking. Naver focuses on software and services, SK Hynix on memory innovation, and Doosan on construction and energy infrastructure.

What Are the Major Challenges Facing This Project?

The biggest obstacle is power supply. Building AI computers at gigawatt scale requires enormous amounts of electricity, and South Korea's power grid is already under strain. The country has delayed major factory projects in the past due to power constraints. Data centers in other Asian countries like Singapore and Japan have hit the same wall, running out of power capacity entirely.

GPU supply presents a second challenge. Even with NVIDIA directly involved in the partnership, global GPU availability remains tight. Scaling from 55 megawatts to gigawatt capacity takes years, and any component delay cascades through the entire timeline. The early 2027 launch date for the first facility appears achievable, but hitting larger capacity goals afterward carries significant execution risk.

How to Track Progress on South Korea's AI Infrastructure Buildout

• Q3-Q4 2026: Watch for announcements confirming power supply and capacity agreements for the Gak Sejong expansion, which will signal whether the project has secured necessary electricity infrastructure
• Early 2027: SK Telecom's first data center is scheduled to go live, marking the earliest verifiable checkpoint for whether the partnership is meeting its timeline commitments
• 2027-2028: Monitor Naver's progress scaling beyond 55 megawatts, which will indicate whether the gigawatt ambitions are on track or facing delays

Why Does This Matter Beyond South Korea?

This announcement signals a fundamental shift in how countries view AI computing capacity. The United States, China, Japan, Taiwan, and Singapore are all racing to lock in next-generation infrastructure, treating AI computing power the same way they once treated oil reserves: as a strategic asset.

South Korea's move with NVIDIA positions it as a serious node in the global AI network, not merely a chip supplier. As AI workloads grow exponentially, the countries and companies that control computing capacity will have significant leverage over where AI products get built, deployed, and priced. Control of gigawatt-scale infrastructure translates directly into geopolitical and economic influence in the AI era.