Samsung, the world's biggest memory chip maker, has unveiled a new memory chip concept called zHBM for future AI servers. It promises extremely high data-transfer speeds and memory density compared to even HBM5 chips, which have not yet launched.
At the Future of Memory and Storage (FMS) 2026 expo in Santa Clara, California, Samsung Electronics unveiled zHBM, short for Z-axis High Bandwidth Memory. The concept is designed for advanced AI computing devices, including AI servers and data centers. Samsung claims that zHBM can offer approximately eight times (8x) higher speeds and more than 10 times (10x) the memory density of HBM5 chips.
These improvements are made possible by stacking HBM chips directly on top of an AI accelerator rather than placing them horizontally beside it on an interposer. This arrangement shortens the distance between the processor and memory, which could improve data-transfer speeds.
Samsung says zHBM chips could offer 10 times more memory density than HBM5 chips through next-generation wafer-bonding technology. They could also provide three times better energy efficiency and 50% lower thermal resistance than conventional memory configurations.
zHBM chips could also support customer-specific designs, allowing customized components to be integrated into the layer between the AI accelerator and the memory chip. This could enable even higher memory capacity and improve accelerator performance.
Samsung has not announced when it plans to begin producing zHBM chips, but they could be introduced after HBM5 chips become available to customers sometime in 2028. So, we are looking beyong 2028 to zHBM chips being ready for an actual commercial launch.
Post a Comment