RamTrend

HBM · May 4, 2026

Samsung 36GB HBM3E 12H Targets Higher-Capacity AI Memory

Samsung’s 36GB HBM3E 12H raises both capacity and bandwidth for AI systems. The product signal is directly relevant to HBM competition, where stack height, thermal control and yield are becoming central constraints.

Price impact: 6Direction: upSource: Samsung Global Newsroom Semiconductors

Samsung announced development of a 36GB HBM3E 12H DRAM product with 12 stacked layers. The company states bandwidth of up to 1,280GB/s and capacity improvements compared with earlier 8-stack HBM3 products. For RamTrend readers, the main point is the push toward higher HBM stack density. AI training and inference systems require more memory close to accelerators, and higher-capacity HBM can improve system design flexibility when package space is limited. Samsung also highlights thermal compression non-conductive film and packaging improvements intended to manage higher stacks while maintaining package requirements. These manufacturing details matter because HBM supply is constrained not only by DRAM wafers but also by stacking, bonding, thermal and yield challenges. The price impact leans upward for premium memory. Higher-capacity HBM supports AI demand and supplier value mix, although successful mass production could gradually improve availability.

SamsungHBM3EHBM3HBMDRAMAI memory
Original sourceBack to news archive