Nvidia to Source 70% HBM4 from SK Hynix and 30% from Samsung
Nvidia will source 100% of HBM4 memory for its next-gen Vera Rubin AI chips from SK Hynix (≈70%) and Samsung (≈30%), completely cutting Micron due to inability to meet >11 Gbps data-transfer specs. Samsung plans mass production and initial shipments by late February.
1. Nvidia's HBM4 Supplier Allocation
Nvidia has allocated roughly 70% of its next-generation HBM4 memory orders to SK Hynix and the remaining 30% to Samsung, ensuring full coverage for its upcoming Vera Rubin AI accelerators.
2. Micron Falls Out on Stricter Specs
Raising HBM4 data-transfer speed targets above 11 Gbps disqualified Micron from meeting Nvidia’s technical requirements, resulting in Micron’s projected share being cut to zero.
3. Samsung Prepares Production and Shipments
Samsung is ramping up mass production of HBM4 chips and expects to begin shipping to Nvidia by the third week of February, aligning memory deliveries with the Vera Rubin launch timeline.