ReportGem ReportGem 中文

REAL-TIME GLOBAL RESEARCH

Global Semis: SOCAMM2 - Innovative but niche, limited impact on memory interface chip TAM

Published: 2026-06-18Institution: BernsteinCompany / ticker: 6809.HK,688008.CH,6723.JPPages: 21Original language: EnglishEvidence page: 1

Research evidence excerpt

Global Semis: SOCAMM2 - Innovative but niche, limited impact on memory interface chip TAM

18 June 2026

China and Japan Semiconductors

Global Semis: SOCAMM2 - Innovative but niche, limited impact

on memory interface chip TAM

Nvidia recently released their latest ARM-based server CPU called Vera, in which they used Qingyuan Lin, Ph.D.

+852 2123 2654 a new memory package format called SOCAMM2. As this new package uses much less

qingyuan.lin@bernsteinsg.com memory interface chip content, some investors are concerned that this will reduce the

TAM for Montage/Renesas. We believe the SOCAMM2 format will likely remain niche to

David Dai, CFA Nvidia and thus only bring HSD to LT% impact to the memory interface chip TAM, reiterate

+852 2918 5704

david.dai@bernsteinsg.com Outperform on Montage and Renesas.

Francis Ma SOCAMM is a DRAM package customized for NVIDIA rather than a disruptive new

+852 2123 2626 standard. In Nvidia’s Grace CPU design, they use LPDDR5x instead of DDR5 to reduce

francis.ma@bernsteinsg.com the power consumption. From Grace to Vera, the shift from soldered LPDDR5x to modular

SOCAMM2 provides modularity flexibility and expands the total capacity/bandwidth.

Kai Zhang However, SOCAMM2 advantages are tightly aligned with NVIDIA’s rack-level optimization, +852 2123 2665

kai.zhang@bernsteinsg.com where maximizing performance-per-watt is paramount. Critically, these benefits come

with trade-offs in capacity, bandwidth, and ecosystem maturity compared with MRDIMM,

therefore we expect this solution will remain niche mainly to Nvidia, which we estimated to

only have HSD to LT% in volume in the next few years.

Other high-end ARM-based CPU and x86 server CPU likely will remain using

MRDIMM if they want to enjoy better capacity and bandwidth. NVIDIA’s adoption

The English excerpt is extracted automatically from the cited source page and may contain layout or recognition errors. It is never batch translated.

Open report viewer