AMD is reportedly considering a significant shift in its manufacturing strategy, engaging in discussions with Samsung to produce some of its future processors. This potential partnership, slated to begin around 2028, comes as AMD seeks to diversify its supply chain and alleviate capacity pressures from its long-standing partner, TSMC. While TSMC has been instrumental in manufacturing AMD's current generation of Ryzen desktop processors and other products, increasing demand and constrained wafer supply, particularly for cutting-edge nodes, may be prompting this strategic exploration.
Currently, AMD utilizes a variety of TSMC's advanced process nodes for its diverse product portfolio. For instance, the Ryzen 7 9800X3D incorporates chiplets fabricated on both TSMC's N4P and N6 nodes, demonstrating a flexible approach to manufacturing. Similarly, its Ryzen AI 300 and 400-series laptop APUs, along with the latest Radeon GPUs, also leverage TSMC's N4P technology, and certain Zen 5c chips are even produced using the N3 process. Looking ahead, AMD has confirmed its use of TSMC's N2 process node for its next generation of Epyc server-grade processors, signaling continued reliance on its primary foundry for high-performance computing.
The potential collaboration with Samsung could address several strategic objectives for AMD. One key factor is the anticipated scarcity of TSMC's most advanced N2 node, which will likely be prioritized for high-demand products, including AMD's own Epyc server processors. By engaging with Samsung, AMD could secure additional manufacturing capacity for other product lines, such as consumer-grade CPUs, to avoid potential supply bottlenecks. This dual-foundry approach could offer greater flexibility and resilience in managing production volumes for different market segments.
The types of processors that Samsung might manufacture for AMD remain a subject of speculation. Two prominent possibilities emerge: lower-tier APUs and the Input/Output Die (IOD) for Zen 6 processors. Budget-friendly CPUs often utilize older, more cost-effective process nodes, making them suitable candidates for Samsung's 4LPP system. Furthermore, IODs, which integrate a considerable amount of analog circuitry for functions like inter-chiplet communication, memory interfaces, and peripheral connectivity, do not benefit as significantly from process node scaling as core logic. Therefore, producing these components on a slightly larger, more mature node at Samsung could be a cost-efficient solution.
Beyond capacity concerns, cost efficiency is another compelling driver for AMD to explore a partnership with Samsung. In the current economic climate, marked by a global memory crisis and escalating prices for components like DRAM and NAND flash, the PC industry is facing headwinds, with declining shipments across most sectors except data centers. As a result, chip manufacturers are actively seeking ways to reduce production costs. If Samsung can offer competitive pricing for its foundry services, it presents a clear financial incentive for AMD to consider shifting some of its chip production away from TSMC, especially for segments where extreme cutting-edge technology is not the paramount requirement.
