NVIDIA Adds Nanya LPDDR5X Memory To Vera Rubin

Wccftech, citing a report from UDN, reports that Nanya Technology has been selected as a supplier of LPDDR5X memory for NVIDIA's Vera Rubin AI server platform and that Nanya is the first Taiwanese memory manufacturer reported to enter NVIDIA's AI server main memory supply chain. Wccftech reports that each Vera Rubin Superchip will include 1.5 TB of LPDDR5X memory operating at approximately 1.2 TB/s, which the article states represents a 3x increase in capacity and about a 50% bandwidth gain versus the Grace Blackwell (GB300) generation.

Editorial analysis - technical context: Public reporting distinguishes two memory tiers on Rubin platforms, with HBM4 used for GPUs where ultra-high bandwidth is required and LPDDR5X used for CPUs where higher density and power efficiency matter. Industry coverage notes LPDDR5X offers higher density and lower power draw compared with some alternatives, while HBM4 provides far higher peak bandwidth but is produced by a smaller set of suppliers.

Supply-chain diversification for AI memory is a recurring pattern as large AI platforms scale. Observers have documented that qualifying additional suppliers can ease single-vendor risk and expand production capacity during aggressive ramp cycles. The reported selection of a Taiwanese firm for LPDDR5X qualifies as notable because Taiwan has been less represented among AI main-memory suppliers historically, per the UDN-cited coverage referenced by Wccftech.

Editorial analysis: Practitioners and procurement teams should monitor three indicators in coming months: supplier qualification timelines and sample-to-production yields; shipment volumes and price trends for LPDDR5X versus prior-generation server DRAM; and how system-level configurations allocate LPDDR5X and HBM4 across Rubin-node variants. Aggregator snippets also reference TSMC involvement in process support for Taiwanese memory makers; following TSMC and Nanya announcements could clarify the manufacturing and qualification role they play.