Nvidia, SK Hynix Sign Multi-Year Deal for Next-gen AI Memory - The Information

NVIDIA and SK Hynix have signed a multi-year agreement to jointly develop next-generation AI memory chips, aimed at enhancing data throughput and reducing latency in AI computing tasks. As AI models g... Read original →

NVIDIA’s strategic lock-in with SK Hynix signals a decisive shift from generic AI architectures toward co-designed memory-compute systems. Technically, 3nm EUV will enable HBM4+ bandwidth beyond 2TB/s, pressuring TSMC to scale CoWoS capacity and forcing EDA vendors to accelerate heterogeneous integration simulation tools. On compliance, tightening U.S.-South Korea export controls on advanced lithography gear demand redundant supply chains—especially given ASML delivery bottlenecks. Competitively, Samsung may retaliate with aggressive HBM pricing, while AMD could fast-track its Micron collaboration. Within 18 months, this alliance will crystallize AI memory as a standalone high-barrier segment, sidelining players lacking foundry-designer vertical integration.