Nvidia's forthcoming Rubin AI servers, slated for release later this year, are expected to drive a significant surge in demand for NAND flash memory due to their novel context memory storage (CMX) technology. Analysts predict this surge could substantially impact the global NAND supply chain, potentially prioritizing AI server needs over consumer electronics.
The CMX system integrated into Nvidia's Rubin servers is designed to house a massive 576 SSDs, collectively offering 9,600 TB of storage capacity. Industry estimates forecast that the NAND required for CMX will skyrocket from 35 million TB this year to over 100 million TB by next year. One analyst, Jukan, described this impact on X by stating that Nvidia's CMX will "quite literally soak up NAND supply like a sponge absorbs water," likening the increased demand to the addition of an "Apple-sized source of demand to the NAND market."
This CMX technology functions as an intermediate storage layer between an AI accelerator's high bandwidth memory (HBM) and its conventional storage. Similar to how RAM relates to CPU cache and SSDs, CMX offers faster access than standard long-term storage solutions and can be shared across multiple AI accelerators. Its enhanced accessibility is facilitated by its connection over 'Spectrum-X Ethernet'.
This increased demand presents a significant opportunity for NAND manufacturers, such as Samsung, which is reportedly vying to become a key supplier for this emerging market. However, the development raises concerns for consumers. Experts suggest that the prioritization of large contracts from the AI industry is likely to affect the availability and pricing of NAND and DRAM for the consumer market in the short to medium term. While some industry figures have expressed concerns about consumer pricing, the redirection of supply away from AI servers remains uncertain.