NVIDIA and SK Telecom to build gigawatt-scale Korean AI Cloud on DSX platform

NVIDIA and SK Telecom unveiled plans for a gigawatt-scale AI Cloud in Korea built on the NVIDIA DSX platform, with the first 'AI factory' slated to come online in 2027. The deal is part of NVIDIA's broader sovereign-AI infrastructure strategy, in which it partners with national telcos and governments to stand up domestic compute capacity rather than leaving regions dependent on US hyperscalers.

The announcement accompanied a wave of GTC Taipei 2026 reveals from Jensen Huang, including the Vera Rubin computing platform and Vera CPU, a Microsoft PC collaboration, the Alpha Mini 2 open model for self-driving cars, and Blackwell RTX GPUs with 6,144 CUDA cores and a claimed petaflop of AI performance. Korea is becoming a strategic battleground: Anthropic just opened a Seoul office and Samsung is committing to AI-driven factories, making the country a focal point for the week's infrastructure narrative.

The Korean cloud fits a pattern of nation-scale compute deals NVIDIA has been signing globally, complementing European efforts like the Bull/Foxconn Vera Rubin NVL72 buildout. For SK Telecom, it's a bet that domestic enterprises and government will pay a premium for in-country, sovereign AI capacity amid rising geopolitical sensitivity around where models run and data lives — the same 'sovereign AI' theme Cohere is exploiting around the Anthropic suspension.

The caveat is timeline and power: gigawatt-scale facilities are enormous energy and capital commitments, and 2027 is far enough out that competitive hardware (and Amazon's custom-silicon ambitions) could reshape the economics before the first factory ships. Watch for confirmed capacity figures, anchor customers, and whether the buildout actually lands on schedule.