NVIDIA and SK Group commit more than $500 billion to Korean AI factories

NVIDIA and SK Group unveiled an expanded strategic partnership valued at more than $500 billion, spanning AI factories and next-generation memory, at South Korea’s AI Summit on July 24.

The partnership ranks among the largest single commitments yet to national AI infrastructure, and it ties NVIDIA’s accelerator roadmap to Korea’s dominance in high-bandwidth memory at a moment when that memory is the industry’s tightest bottleneck.

At the center is a 2-gigawatt AI factory that SK Telecom will build on NVIDIA’s DSX platform and Vera Rubin accelerated computing, powered by SK hynix HBM4 memory, with the first facility due online in 2027. SK hynix separately entered a long-term AI memory supply and co-development agreement with NVIDIA.

NVIDIA will also invest $1 billion in Naver, which plans to more than triple its GPU data-center capacity from 55 megawatts to 200 megawatts. Alongside Brookfield, Naver and NVIDIA will expand Korea’s national AI factory to roughly 200 megawatts, or about 100,000 Vera Rubin GPUs.

SK Group Chairman Chey Tae-won said SK would collaborate with NVIDIA on “a world-class AI factory.”

The headline figure is the partners’ own and stretches across multiple projects and years rather than a single upfront outlay; the first SK Telecom facility is not scheduled to run until 2027, and much of the memory arrangement is framed as long-term co-development rather than committed volume.