Samsung’s (OTCPK:SSNLF) fifth-generation, 12-layer HBM3E has passed Nvidia’s (NASDAQ:NVDA) qualification tests, meaning it can now become a critical supplier for Nvidia’s most powerful artificial intelligence accelerators, according to The Korea Economic Daily.
Samsung’s HBM3E had already passed qualification tests for Advanced Micro Devices (AMD) MI350, according to the report, which cited people familiar with the matter. It can now be used with Nvidia’s B300 Blackwell Ultra GPU. Nvidia had already approved 12-layer HBM3E chips from SK Hynix and Micron Technology (MU).
However, after achieving this milestone, Samsung must now perfect HBM4, which is slated to appear in Nvidia’s next generation of GPUs, the Vera Rubin. The Vera Rubin will succeed the Blackwell and is expected for release during the second half of 2026.
Samsung plans to send some HBM4 samples to Nvidia as soon as this month, the report said. The South Korean company is in talks to supply HBM4 to Broadcom (AVGO) and Google (GOOG)(GOOGL) as well. Samsung partnered with Taiwan Semiconductor Manufacturing (TSM) to develop its HMB4 offering.