Micron Aims to Double HBM Wafer Capacity to 100K Monthly by Late 2026
Micron is reportedly planning to double its monthly HBM wafer production capacity from 40,000–50,000 wafers in late 2025 to 100,000 by the end of 2026. This aggressive capacity expansion is primarily focused on 12-high (12Hi) HBM4 memory targeted for next-generation AI hardware, such as NVIDIA's upcoming Rubin architecture. High Bandwidth Memory (HBM) is a crucial component for AI accelerators, and memory capacity limits directly bottleneck modern server performance. By expanding production, Micron aims to narrow the market share gap with top competitors SK Hynix and Samsung while securing a key role in supply chains for future AI systems. Micron plans for 12Hi HBM4 to account for 50% of its total HBM output by late 2026, up from an estimated 20–30% earlier that year. To achieve this, the company is significantly expanding equipment purchases for its manufacturing facilities in Taiwan and Singapore.
## BACKGROUND
High Bandwidth Memory (HBM) stacks multiple DRAM chips vertically to deliver significantly higher bandwidth and power efficiency than standard memory modules. 12Hi HBM4 represents the fourth-generation HBM design using 12 stacked layers, engineered specifically to support massive parallel workloads required by advanced AI reasoning architectures like NVIDIA Rubin.