TradingKey - Against the backdrop of intensifying competition in AI computing power, collaboration across the upstream and downstream segments of the chip industry supply chain is becoming increasingly close. Samsung Electronics and Advanced Micro Devices ( AMD) recently signed a memorandum of understanding (MoU), once again drawing market attention back to key sectors such as high-bandwidth memory, advanced manufacturing processes, and AI infrastructure.
The collaboration between Samsung Electronics and AMD this time goes deeper than it appears on the surface. The two companies signed an MoU on Wednesday with a clear direction: to advance memory, computing, and potential manufacturing segments together around next-generation AI infrastructure. Key highlights mentioned in the official statement include HBM4 high-bandwidth memory, DDR5 solutions for next-generation EPYC processors, and a point of particular market interest: the possibility of further cooperation at the foundry level in the future. The agreement was signed at Samsung's advanced manufacturing base in Pyeongtaek, with AMD's Chair in attendance, underscoring the importance both parties place on this partnership.
Looking at the details, HBM4 remains the core focus. Samsung is set to become a major memory supplier for AMD's next-generation Instinct MI455X accelerators, which is perhaps the most significant aspect of this collaboration. Simultaneously, both parties are advancing memory compatibility and optimization for the next-generation EPYC platform (including Helios). The underlying logic is straightforward: the bottleneck for AI computing power has long since moved beyond the chips themselves; bandwidth and data scheduling efficiency have become increasingly critical. Memory is no longer a "follower" but a key variable directly impacting overall system performance.
Looking further, the foundry aspect is also noteworthy. While it is currently only in the "exploratory stage," once implemented, it would mean that Samsung Electronics is not just providing memory to AMD, but could also further participate in the chip manufacturing process. Samsung has previously supplied HBM3E for AMD products such as the MI350X and MI355X; the upgrade to HBM4 essentially extends the collaboration from current products to the next-generation platform cycle. For Samsung, this is not just about securing orders, but also a significant move to gain leverage in competing with SK Hynix for high-end HBM market share.
From another perspective, this type of cooperation illustrates a trend: the AI supply chain is being "repackaged." In the past, focus was primarily on GPUs or CPUs themselves, but now, segments like memory, packaging, foundry services, and system architecture are becoming increasingly difficult to separate. AMD requires stable and leading HBM resources to support its product iteration cycle, while Samsung needs such clients to strengthen its position in high-end storage and advanced manufacturing. Both parties get what they need, but more importantly, once such a tie-up is formed, it often spans multiple product cycles rather than being a simple one-off supply agreement.