Samsung and TSMC Join Forces for Nvidia’s Custom HBM
TREE NEWS reports: Samsung Electronics will partner with TSMC to produce the base die for Nvidia’s next-generation custom high-bandwidth memory (HBM), known as NVHBM, using TSMC’s 3nm process. This marks the first time Samsung has collaborated with TSMC from the HBM manufacturing stage. Nvidia has also requested that Samsung, SK Hynix, and Micron uniformly adopt TSMC’s 3nm base die and standardize the core die on the sixth-generation DRAM 1c process to enhance consistency across suppliers.
Industry Implications: A Shift in Supply Chain Dynamics
The move represents a significant shift in the semiconductor supply chain. Traditionally, Samsung has produced both the core die and base die for HBM in-house. By adopting TSMC’s 3nm process for the base die, Samsung is aligning with Nvidia’s requirements for higher performance and consistency. This collaboration could pave the way for more standardized HBM solutions, benefiting the broader AI and high-performance computing markets.
However, Samsung will maintain its own foundry strategy for standard HBM and custom HBM for other clients like AMD. This dual approach allows Samsung to stay competitive while catering to Nvidia’s specific needs. The partnership also highlights the growing importance of advanced packaging and process technology in the AI chip race.
Forward-Looking Perspective
As AI models grow in complexity, the demand for high-performance memory will only increase. This collaboration between Samsung and TSMC, driven by Nvidia’s specifications, could set a new standard for HBM production. It also underscores the interconnectedness of the global semiconductor industry, where even competitors must collaborate to meet the demands of leading AI chip designers.
Looking ahead, we can expect further innovations in HBM technology, with a focus on higher bandwidth, lower power consumption, and greater integration. The move towards 3nm base dies and 1c DRAM core dies is a step in that direction, and it will be interesting to see how other players in the ecosystem respond.
