Nvidia Rubin memory costs 435%↑, Samsung and SK’s negotiating power in focus
A Morgan Stanley analysis estimates that memory costs related to Nvidia’s next-generation AI platform, Vera Rubin, will 435%. The HBM4 supply landscape and the supply negotiating power of Samsung Electronics and SK hynix are key variables.
According to Morgan Stanley data, memory costs related to Nvidia’s next-generation AI platform, Vera Rubin, are estimated to rise sharply. The data was based on an analysis of component costs for VR200 NVL72 racks. The analysis indicated that memory costs would increase by 1%435.
The data said GPUs still account for the largest share of AI server rack costs. However, it pointed out that memory is the item driving the biggest increase in costs during a generational transition. The transition to a new memory generation was cited as a potential source of increased demand for HBM4.
Morgan Stanley’s data drew attention to the HBM4 supply landscape. It said memory prices and supply stability could vary depending on suppliers and production capacity. This was identified as a key variable in cost changes.
The report200 NVL72 indicates that memory accounts for a significantly larger share of the components in the rack configuration. It states that the specific cost estimate reflects memory unit prices and the amount installed. The figure is an estimate based on Morgan Stanley’s modeling.
The report noted the potential for Samsung Electronics and SK Hynix to benefit. Their HBM4 production capacity and supply contract status were identified as areas of interest. It emphasized that changes in bargaining power could affect each company’s revenue mix.
Sources: 연합인포맥스
Bullstory synthesized this report from the linked sources. AI may have assisted the writing process. Check the original reports for factual confirmation.