Specifications
Capacity
48GB (16-High) / 36GB (12-High)
Pin Speed
10 GT/s (16-High); 11.7 Gbps (12-High)
I/O Interface
2,048-bit
Bandwidth
>2 TB/s per stack
Packaging
MR-MUF (mass reflow molded underfill)
Status
Mass production Q3 2026
Target
NVIDIA Vera Rubin platform
Overview
SK hynix is the primary HBM4 supplier for NVIDIA's Vera Rubin AI accelerator platform, holding an estimated 60–70% of initial HBM4 volume. Its HBM4 portfolio spans a 12-High 36GB stack demonstrated at 11.7 Gbps (the industry's fastest at the time) and a 16-High 48GB configuration that pushes TSV stacking to 16 dies.
HBM4 doubles the interface width to 2,048 bits with 32 independent channels, delivering more than 2 TB/s per stack. QS Compute supplies SK hynix HBM4 stacks for next-generation AI accelerator, GPU board, and advanced packaging (MR-MUF) programs.
Request a Quote — SK hynix HBM4 16-High 48GB
QS Compute — Global B2B AI computing hardware supply, 15-day sample lead time. Price available on quote; stock available to order.
Get Your Quote →