South Korea is not waiting for overseas hyperscalers to define what its AI infrastructure looks like. SK Telecom and NVIDIA announced plans to build gigawatt-scale AI cloud capacity inside the country, with the first facility scheduled to come online in 2027. For a nation whose industrial identity runs through memory chips, robotics, and advanced manufacturing, the timing feels deliberate.
NVIDIA’s DSX full-stack platform powers the infrastructure, operating differently from a conventional cloud data center. Rather than handling mixed enterprise workloads, an AI factory optimizes for one thing specifically: generating the computational output that AI training, inference, and agentic systems need to run continuously at scale. The distinction matters because it shapes hardware configuration, cooling requirements, and energy planning from the ground up, not as an afterthought.
SK Telecom is directing the capacity toward three areas where South Korea already has serious footing. Sovereign AI development, industrial robotics, and semiconductor manufacturing are not abstract targets here. SK hynix, part of the same SK Group family, has been running digital twin projects for its semiconductor fabs using NVIDIA Omniverse to sharpen manufacturing operations. That kind of deployment is exactly what the new AI factory infrastructure is designed to support at greater scale.
Jensen Huang, NVIDIA’s founder and CEO, put the telecom angle plainly. Connectivity networks, he argued, now have the potential to become the backbone of national AI clouds rather than simply moving data between endpoints. For SK Telecom, that reframing shifts its identity from a service carrier to a piece of critical national infrastructure, a different kind of business entirely.
SK Group Chairman Chey Tae-won noted that the partnership secures full-stack AI capability spanning chips all the way through to data center operations. Beyond the 2027 build, both companies plan joint research into next-generation AI factory architecture, focusing on silicon efficiency, memory optimization, and how data centers need to evolve to handle future AI workload demands.
The sovereignty thread running through the announcement is hard to miss. South Korea joins a growing list of countries actively building domestic AI infrastructure rather than routing critical workloads through foreign-controlled platforms. Having a national telecom operator anchor that effort gives it more credibility than a government policy document alone ever could.
