SK Hynix Pushes Full-Stack AI Memory With Nvidia CUDA-X and HBF Roadmap
SK Hynix is extending its HBM leadership into NAND flash with a full-stack AI memory push, standardizing next-generation HBF through work with SanDisk and an OCP joint effort. The HBF spec targets up to 512GB capacity and 3TB/s bandwidth, aiming for first AI inference samples next year and faster time-to-market using Nvidia CUDA-X and PhysicsNeMo from verification through fab operations.