Samsung Electronics and SK Hynix have been confirmed as the sole suppliers of HBM4 memory for Nvidia’s next-generation Vera Rubin AI accelerator, excluding U.S. rival Micron Technology from this flagship segment.
The selection represents a significant consolidation of the AI memory market, with the two Korean firms set to dominate the most lucrative tier of Nvidia’s upcoming hardware cycle. The supplier split is expected to favor SK Hynix with roughly 70 percent of Nvidia’s HBM4 allocation, while Samsung takes approximately 30 percent.
Micron, which currently supplies HBM3E for Nvidia’s existing platforms, does not appear on the Vera Rubin vendor list. “Micron isn’t even being discussed as a Vera Rubin HBM4 supplier,” an industry source told the Korea Economic Daily. Analysts attribute Micron’s exclusion to difficulties meeting Nvidia’s speed and thermal specifications, particularly around base die design.
Samsung passed Nvidia’s qualification tests for both 10 Gbps and 11 Gbps HBM4 variants and began limited shipments in February. SK Hynix is optimizing its 11 Gbps HBM4 product and expects to begin full-scale production this month. Nvidia pushed suppliers to exceed 10 Gbps operating speeds for Vera Rubin, well above the 8 Gbps JEDEC industry standard.
Vera Rubin will incorporate 8 HBM4 stacks per GPU totaling 288 GB of memory per GPU. The full Vera Rubin Superchip combines two GPUs for a total of 576 GB of memory. The hardware is slated for release in the second half of 2026 and is expected to be formally unveiled at Nvidia’s GTC developer conference on March 16.
Micron is expected to supply HBM4 for mid-tier Rubin accelerators, such as the Rubin CPX, rather than the top-tier Vera Rubin. Research firm TrendForce noted in February that Nvidia may eventually bring all three suppliers into its HBM4 ecosystem to ensure stable supply. The March 8 vendor list confirmation suggests Samsung and SK Hynix will dominate the initial ramp.
For Samsung, the selection follows struggles with HBM3E yields that left it largely shut out of earlier Nvidia supply chains. For SK Hynix, the allocation reinforces its position as Nvidia’s primary memory partner. The Korea Economic Daily published the original report on March 8.
