Micron intends to double its High-Bandwidth Memory production capacity to meet surging AI accelerator requirements.

US memory chipmaker Micron plans to increase its monthly high-bandwidth memory (HBM) production to approximately 100,000 wafers by the end of 2026, up from roughly 40,000 to 50,000 wafers in 2025. According to reports dated September 4, this expansion comes as sustained demand for AI chips continues to constrain global HBM supply.

To support this ramp-up, Micron has placed additional orders with HBM equipment suppliers and is installing new machinery at its fabrication sites in Taiwan and Singapore. The company also intends to boost production of 12-layer HBM4 specifically for Nvidia’s Vera Rubin platform.

The 12-layer HBM4 architecture stacks a greater number of dynamic random-access memory (DRAM) dies within a single package, enhancing both density and bandwidth for AI accelerators. Nvidia describes Vera Rubin as its next data center AI platform.

Industry estimates place both Samsung Electronics and SK Hynix at a monthly output of 150,000 to 200,000 wafers. Meanwhile, Counterpoint Research projects that SK Hynix will command a 50% share of the global HBM market in the second quarter of 2026, followed by Samsung Electronics at 32% and Micron at 18%.

Source: Cailianshe