European Investor
20 Apr 2026, 10:06
SK hynix announced the mass production of its 192GB SOCAMM2 memory module, positioning it as a next-generation standard for AI server performance.
The SOCAMM2 module is based on LPDDR5X DRAM built on the advanced 1cnm process and is specifically designed for AI server environments, including NVIDIA’s Vera Rubin platform. It adapts low-power mobile memory technology for high-performance data center use, combining compact form factor with scalability.
The company highlighted that the new module delivers more than double the bandwidth and over 75% better power efficiency compared to conventional RDIMM solutions, addressing critical energy and performance constraints in AI infrastructure.
SK hynix expects SOCAMM2 to significantly reduce memory bottlenecks in large language model (LLM) training and inference, particularly for models with hundreds of billions of parameters. As AI workloads increasingly shift toward training, the demand for high-capacity, energy-efficient memory solutions is rising.
With early stabilization of mass production and close collaboration with global cloud and AI partners, including NVIDIA, SK hynix aims to strengthen its position as a leading provider of AI-focused memory solutions.
The SOCAMM2 module is based on LPDDR5X DRAM built on the advanced 1cnm process and is specifically designed for AI server environments, including NVIDIA’s Vera Rubin platform. It adapts low-power mobile memory technology for high-performance data center use, combining compact form factor with scalability.
The company highlighted that the new module delivers more than double the bandwidth and over 75% better power efficiency compared to conventional RDIMM solutions, addressing critical energy and performance constraints in AI infrastructure.
SK hynix expects SOCAMM2 to significantly reduce memory bottlenecks in large language model (LLM) training and inference, particularly for models with hundreds of billions of parameters. As AI workloads increasingly shift toward training, the demand for high-capacity, energy-efficient memory solutions is rising.
With early stabilization of mass production and close collaboration with global cloud and AI partners, including NVIDIA, SK hynix aims to strengthen its position as a leading provider of AI-focused memory solutions.