SK Hynix looks beyond HBM as inference workloads diversify

SK Hynix is expanding its memory strategy beyond the conventional GPU-HBM pairing as inference workloads split between latency-sensitive services and capacity-heavy, long-context applications.