Memory and Foundry Talks
AMD Chief Executive Lisa Su is visiting South Korea to secure additional supplies of high-bandwidth memory for AI accelerators. Meanwhile, Samsung is seeking a broader partnership to turn its memory supply relationship into foundry chip production orders.
When asked about discussions with Samsung executives, Lisa Su stated that the companies plan to talk about their broader partnership, noting that AMD maintains many collaborative ventures with Samsung.
Prior to visiting corporate headquarters, executive leadership attended an event where they highlighted local memory makers as important strategic partners. This reflects the high significance of data center-grade DDR5 and HBM memory for the company's AI and data center businesses.
Long-Term Supply Strategy
Describing Samsung Electronics and SK hynix as key supply-chain partners, Su emphasized that extreme AI demand requires long-term planning three to five years ahead with industry partners.
The non-binding memorandum of understanding signed earlier covers four major collaborative areas over a three-year period:
• Supply of next-generation memory products, including HBM4 and advanced DDR5;
• Co-development and optimization of memory solutions for AI and high-performance computing systems;
• Exploration of foundry manufacturing opportunities for future AMD platforms;
• Co-optimization of advanced packaging and related semiconductor technologies.
Manufacturing Risks and Leverage
While Samsung has become a principal supplier of HBM4 for advanced accelerators, expanding collaboration beyond memory supply is complex. It involves considerable redesign efforts for Samsung Foundry's process technologies alongside ramp-up risks at a new contract chipmaker. Although AMD already utilizes Samsung for certain programmable logic devices, moving advanced production carries inherent risks, though Samsung's position as an HBM supplier provides commercial leverage.
AMD has previously noted in financial reports that transitioning product manufacturing to new foundries carries risks of delayed introductions, lower yields, or reduced performance.
Current Accelerator Shipments
Meanwhile, shipments of next-generation accelerators equipped with massive HBM4 capacity alongside rack-scale Helios AI systems are already underway to interested customers.
Leadership confirmed that these newly shipping systems rely entirely on HBM4 architecture.
As negotiations continue for expanded memory allocations, industry observers are watching to see if AMD will adopt Samsung Foundry for advanced processors in exchange for securing high-bandwidth memory capacity.
