Samsung is lining up HBM4 shipments for Nvidia’s Vera Rubin AI lineup as soon as June, a handy turnaround for a business that was wobbling.
HBM4 is being sold as a “revolutionary” step for memory makers, and Samsung now looks to have “turned the tables” on rivals after earlier struggles to win customers.
Korean media claims Samsung’s HBM4 is first in line for Rubin, with supply expected as soon as next month.
A key brag is pin speed, with Samsung’s HBM4 rated at 11 Gbps+, well above the JEDEC baseline and apparently a requirement Nvidia pushed for.
With agentic AI pitched as the next big thing, Rubin gets a chunky memory spec bump, driven by Samsung’s faster interface and wider pipes.
Samsung’s HBM4 also uses a 4nm logic base die from its own foundry, giving it more control over delivery than SK hynix and Micron leaning on TSMC.
Nvidia has reportedly moved Vera Rubin into “full production”, and suppliers need to keep pace or get dropped.
The report says customer shipments tied to Vera Rubin start in August, and Rubin chips will be shown in full at GTC 2026, with Samsung’s HBM4 getting its moment.
