Samsung and AMD MoU for HBM4 Memory, DDR5 Spurs Nvidia GPU Supply Concerns
Samsung and AMD signed an MoU to supply HBM4 memory for Instinct MI455X accelerators and optimized DDR5 for sixth-generation EPYC processors, and explore a foundry partnership. The deal, unveiled during Nvidia's GTC when Jensen Huang praised HBM4, could heighten competition for memory and tighten Nvidia’s AI GPU supply.
1. MoU Details
Samsung Electronics and AMD agreed to a memorandum of understanding covering Samsung’s next-generation HBM4 memory for AMD’s Instinct MI455X AI accelerators and optimized DDR5 modules for AMD’s sixth-generation EPYC processors. The pact also initiates talks for Samsung to provide contract chip manufacturing services for AMD’s future product lines.
2. Timing at Nvidia GTC
The announcement coincided with Nvidia’s annual GTC event, where CEO Jensen Huang highlighted Samsung’s HBM4 capabilities and unveiled a separate foundry partnership with the Korean firm. This alignment underscores Samsung’s strategic push to become a key supplier across major AI chipmakers.
3. Impact on Nvidia GPU Supply
By locking in long-term HBM4 and DDR5 commitments with AMD, Samsung may need to allocate more advanced memory wafers to AMD customers, potentially limiting available capacity for Nvidia’s AI GPU production. Tighter memory availability could influence Nvidia’s manufacturing timelines and pricing for upcoming GPU launches.
4. Semiconductor Market Context
Samsung holds about 22% of the global HBM market, trailing market leader SK Hynix at 57%. AMD recently secured up to $60 billion in AI chip deals with major cloud providers, while Samsung seeks to narrow the gap in high-bandwidth memory as AI-driven demand accelerates.