Samsung Nears AMD 2nm AI Chip Deal
Samsung Electronics is moving closer to securing a major foundry agreement with AMD for next-generation AI processors using Samsung’s second-generation 2nm process technology. The negotiations began in late 2025, while both companies formalized discussions through a memorandum of understanding signed in March 2026.
Under the agreement, Samsung will supply HBM4 memory for AMD’s upcoming Instinct MI455X GPU. Additionally, the company will provide advanced DDR5 solutions for AMD’s sixth-generation EPYC Venice processors.
The partnership also includes discussions surrounding foundry services for future AMD products. Consequently, Samsung continues expanding its role across both semiconductor manufacturing and AI infrastructure markets.
Samsung Expands Foundry Business
Foundry discussions between the companies have continued since late 2025. Earlier reports indicated Samsung’s Foundry Business Unit was evaluating production of AMD’s EPYC Venice server CPUs on the SF2P 2nm node.
Samsung planned to validate the process through multi-project wafer testing before moving toward large-scale manufacturing. As a result, industry observers increasingly view commercial production as highly probable.
Meanwhile, Samsung has aggressively expanded its 2nm customer base. The company recently confirmed ongoing discussions with several major technology firms regarding additional 2nm logic chip contracts.
Samsung also secured a multibillion-dollar 2nm semiconductor order from Tesla. However, the SF2P process reportedly maintains yields near 55%, slightly below the standard threshold generally required for stable mass production.
If Samsung satisfies AMD’s performance targets, AMD could eventually adopt a dual-foundry strategy involving both Samsung and TSMC for future consumer processors.
HBM4 Competition Intensifies
The collaboration further strengthens Samsung’s position in the high-bandwidth memory market. Currently, SK Hynix controls roughly 55% of the HBM sector, while Samsung holds approximately 25%.
Earlier this year, Samsung became the first company to begin mass production and shipment of HBM4 chips. These memory products are expected to play a major role in future AI computing systems, including Nvidia’s upcoming Vera Rubin architecture.
Additionally, growing demand for AI processors continues to reshape the semiconductor industry. Therefore, partnerships involving advanced foundry manufacturing and HBM technologies are becoming increasingly critical to long-term AI infrastructure development.
