Nvidia and SK hynix ink multi-year memory co-development and supply agreement

NVIDIA and SK Hynix have entered a multi-year collaboration agreement to co-develop next-generation memory technologies and secure supply for NVIDIA's upcoming platforms. The deal addresses the growin... Read original →

NVIDIA and SK hynix’s alliance marks a strategic pivot toward memory-compute co-design in the AI arms race. Technically, joint HBM4 and LPDDR5X development forces upgrades across EDA, TSV packaging, and CoWoS capacity—accelerated by CuLitho and PhysicsNeMo, cutting memory design cycles by over 30%. On compliance, the partnership sidesteps some U.S.-South Korea export controls, yet SK’s Wuxi fab remains vulnerable if Washington restricts HBM exports to China. Competitively, Samsung and Micron will rush HBM4 samples and push DDR5+CXL alternatives, while TSMC may deepen Micron integration to counter SK’s exclusivity. Within 18 months, memory will consume >45% of AI server BOM costs, shifting architecture paradigms from ‘chip-first’ to ‘memory-defined,’ spurring vertical software stacks like CUDA-X for memory.