Nvidia’s Vera Rubin platform consumes 12-high HBM4 stacks, and Micron is reorganizing its memory fabs around that demand. Monthly high-bandwidth memory output is on track to reach about 100,000 wafers by the end of 2026, according to ETNews, which cited industry sources.
That rate would be roughly double the 40,000-50,000 wafers a month the company’s HBM lines managed last year. Micron has told investors that its HBM4 ramp is running about twice as fast as the HBM3E ramp that preceded it, and the taller stacks are expected to dominate its output mix as volumes grow.
The company remains the only US HBM producer, a status that brings CHIPS Act support and customers who want supply beyond Korea. Even so, Samsung and SK hynix keep a wide lead; both began HBM4 volume earlier and still run far larger HBM programs.
Nvidia has been widening its supplier list for future accelerators, giving Micron a door into designs it previously missed. Meeting the 2026 target would put the company within reach of a market that two Korean suppliers have long dominated.
