In the evolving market landscape of high-bandwidth memory (HBM) technologies, Samsung Electronics Co. Ltd. and SK Hynix have emerged as the primary suppliers for Nvidia Corporation's upcoming HBM4 memory chips. This development signals a narrowing of the competitive field as the industry focuses on supplying memory components for Nvidia's new AI accelerator series.
Nvidia’s next-generation AI chip, codenamed "Vera Rubin," is slated to be the initial platform utilizing sixth-generation HBM4 technologies. According to recent analysis, Nvidia intends to allocate its HBM4 sourcing almost exclusively between SK Hynix and Samsung, effectively excluding Micron Technology from this supply chain segment.
Specifically, projections now indicate SK Hynix will provide approximately 70% of the HBM4 memory for the Vera Rubin series, with Samsung supplying the remaining 30%. This marks a significant departure from prior expectations that included Micron as a key participant. The exclusion of Micron is largely attributed to its inability to meet Nvidia's increasingly rigorous technical criteria for HBM4 components.
Industry observers note that Nvidia has raised its performance benchmarks for memory data transfer speeds beyond 11 gigabits per second. These heightened requirements have reportedly presented a barrier for Micron, leading to its removal from the supply roster for the Vera Rubin AI chips.
Amid this recalibration, Samsung is advancing its manufacturing capabilities, preparing for mass production of HBM4 memory. Reports indicate that Samsung aims to commence shipments of these advanced chips to Nvidia by late February, aligning with the launch timeline of the Vera Rubin accelerators designed for AI infrastructure applications.
This progress positions Samsung to strengthen its foothold in Nvidia’s supply ecosystem, closing the gap with SK Hynix, which has traditionally held a leading role. Market response to Samsung’s developments has been positive, with its stock price rising. This uptick is partly driven by broader market optimism fueled by robust demand for AI-driven memory chips from major cloud service providers and technology companies.
Investors are also optimistic about the broader data center expenditure landscape, with hyperscalers forecasted to direct around $650 billion in spending during the current year. Nvidia’s shares reflected these sentiments with a near 8% increase recently.
Conversely, Micron Technology’s stock experienced a decline of approximately 2.23%, trading around $385.88 during pre-market hours, reflecting market concerns over its diminished role in Nvidia’s strategic AI memory sourcing.
Looking ahead, competition among memory chip producers remains intense as the demand for advanced AI infrastructure components continues to escalate. Samsung and SK Hynix’s positions as primary suppliers for Nvidia’s HBM4 chips highlight the importance of meeting stringent technical specifications and the capacity to ramp up production in alignment with market needs.