Key insights
- SK Hynix has begun mass production of its SOCAMM2 memory module designed for Nvidia's Vera Rubin AI chip. This advancement addresses memory bottlenecks in AI model training and inference. While benefiting SK Hynix, it also signals continued growth and innovation in the AI server market, indirectly supporting Nvidia's future product line and potentially boosting sentiment in the semiconductor sector.
Investing.com-- SK Hynix Inc (KS:000660) said on Monday it had begun mass production of a next-generation memory module designed for Nvidia’s next-generation Vera Rubin artificial intelligence chip.
The company said it began mass production of the 192GB SOCAMM2, a next-generation memory module aimed at AI server functionality while using less power.
Get more updates on the top AI companies by subscribing to InvestingPro
SOCAMM2 products are designed for Nvidia Vera Rubin, SK Hynix said in a statement. The company said the new products will “fundamentally resolve the memory bottlenecks encountered during the training and inference of large language model(s).”
SK Hynix is one of the world’s largest memory chip makers, and is a key supplier of advanced memory to Nvidia.
Shares of the memory chip maker rose 2.1% on Monday, helping drive a 1% rise in the KOSPI index. Rival Samsung Electronics Co Ltd (KS:005930) shed 1%.
Both companies have benefited greatly from a surge in memory chip demand in the AI industry. Advanced memory is a crucial component of AI servers, given the high computing requirements of advanced AI models.
Nvidia had unveiled Vera Rubin earlier this year, with the line set to succeed its Blackwell processor line. Initial shipments are expected to begin in the second half of 2026, although recent reports flagged some potential delays due to capacity constraints at major Nvidia suppliers.