Analysis
Nvidia and SK Group expanded their strategic partnership in Seoul on July 25, a comprehensive tie-up both companies have described as worth more than $500 billion across AI factories and next-generation memory development, according to SK hynix's own newsroom and Nvidia's newsroom. The centerpiece is a long-term agreement for Nvidia and SK hynix to co-develop and optimize next-generation AI memory, including high-bandwidth memory (HBM), for workloads ranging from large-language-model training to agentic and physical AI.
Beyond memory, SK Telecom is set to build a 2-gigawatt Nvidia Vera Rubin DSX AI Factory to serve global compute demand -- a scale that puts a single Korean buildout in the same range as some of the largest US hyperscale campuses. The partnership builds on years of existing co-engineering between the two companies, which already supplies memory for some of the most advanced AI computing platforms running today.
The $500 billion figure spans a multi-year, multi-project partnership rather than a single committed check, and neither company has published a year-by-year breakdown of how it's allocated across memory R&D, factory construction and compute buildout. For a market already pricing AI infrastructure spending as a near-certainty, the practical significance is less the topline number and more the memory-supply commitment specifically -- HBM has been one of the tighter bottlenecks in the AI chip supply chain, and a locked-in, multi-year co-development deal with the world's largest memory maker is a real hedge against that constraint.