SK hynix announces multiyear deal with Nvidia for AI factories - The Korea Times

NVIDIA and SK Hynix have announced a multi-year technology partnership aimed at advancing next-generation memory technologies for artificial intelligence (AI) factories. This collaboration marks a sig... Read original →

SK hynix’s deep integration with NVIDIA elevates HBM4 from a memory component to a foundational element of AI compute architecture. Technically, this forces tighter co-optimization between sub-3nm logic nodes and EUV multi-patterning, compelling TSMC and Samsung to rebalance CoWoS packaging capacity. On compliance, if U.S. export controls extend to HBM4 parameters like stack count or TSV density, Korean firms face surging costs from onshore U.S. fabs and forced tech localization. Micron will likely accelerate joint validation with AMD and Azure, while Samsung may leverage its logic-memory vertical integration to undercut rivals with bundled AI factory solutions. Within 18 months, an HBM 'certification moat' will emerge—only vendors co-developed with top AI chipmakers will qualify for premium deployments, accelerating market consolidation.