Nvidia Audits Samsung's HBM4 Packaging to Power Rubin GPUs and Diversify Supply Chain

Nvidia has launched intensive audits of Samsung Electronics' HBM4 packaging production lines, focusing on the high-bandwidth memory critical for its upcoming Rubin architecture GPUs, including the Vera Rubin model slated for release in the second half of 2026. Samsung, a key player in the memory market, is in the final adjustment stages for HBM4 supply, with mass production targeted to begin as early as February 2026 following Nvidia's certification. This validation process enters a critical phase as Nvidia seeks to secure reliable sources for its AI accelerators.

The move matters significantly for Nvidia's supply chain strategy, as it aims to lessen dependence on TSMC for advanced packaging while navigating surging demand for AI hardware. HBM4 represents a leap in memory tech, integrating logic dies to act as co-processors that handle data processing more efficiently, enabling higher speeds above 11 Gbps per pin after Nvidia tightened specs in late 2025. Samsung's turnkey approach—using its own 4-nm logic dies, hybrid bonding, and 1c DRAM for better efficiency—gives it an edge, potentially capturing around 30% of Nvidia's HBM4 needs alongside SK Hynix's dominant 70% share, sidelining Micron.

Looking ahead, successful audits could accelerate Rubin GPU mass production, strengthening Nvidia's lead in AI computing with Rubin promising massive performance gains over predecessors. This intensifies competition among memory giants like SK Hynix, Samsung, and Micron, all racing to scale HBM4 capacity—such as Micron's push for 15,000 wafers by year-end—ensuring a robust supply for the next wave of AI infrastructure.