Micron has demonstrated a 512GB DDR5 module for future servers. The vertically stacked RDIMM reaches up to 9,200MT/s and could give a 24-slot dual-socket system as much as 12TB of directly attached DDR5 memory.
Infineon will sell its NOR Flash and F-RAM memory business. Winbond will pay $1.12 billion for the operation, which serves automotive, industrial, and infrastructure customers and is expected to transfer during the second half of 2027.
Astera Labs has expanded Leo controllers for memory-intensive AI workloads. New devices target accelerator-attached, expanded, pooled, and reusable server memory.
Chinese AI-chip suppliers are raising prices as HBM shortages intensify. Huawei, Cambricon, and smaller accelerator developers face higher memory procurement costs, restricted supply, and increasingly difficult allocation decisions.
Samsung plans High-NA EUV insertion into DRAM production by 2028. It is also joining development of a larger 12-inch photomask platform.
Everspin and Teledyne HiRel are expanding mission-critical MRAM availability together. Their partnership brings 256Mb PERSYST STT-MRAM into an aerospace and defence memory portfolio, with customer availability planned for the fourth quarter.
GlobalFoundries and RAAAM have taped out GCRAM test silicon successfully. Their FDX implementation targets 40% less memory area and power reductions reaching 60% against conventional SRAM.
CXMT has moved its LPDDR6 memory technology into volume production. The Chinese DRAM manufacturer specifies speeds reaching 12,800Mbps and capacities up to 16GB, with Xiaomi providing the first announced deployment.
SK hynix has broken ground on its Indiana HBM facility. The $4 billion project combines advanced packaging, testing, and an R&D testbed, with cleanroom opening targeted for 2028 and mass production in 2029.
Primemas has detailed rack-scale CXL memory for demanding PNNL workloads. The Abaco project combines high-density controller silicon, Micron DDR5, and memory pooling to exceed 100TB per rack.