Micron Delivers Industry’s Highest Capacity SOCAMM2 for Low-Power DRAM in the AI Data Center

Core Insights - Micron Technology, Inc. has announced the customer sampling of its 192GB SOCAMM2, a low-power memory solution aimed at enhancing AI data center infrastructure [1][2] - The SOCAMM2 offers 50% more capacity than its predecessor while improving power efficiency by over 20%, which is crucial for optimizing large data center clusters [1][4] - This new memory module significantly reduces time to first token (TTFT) by more than 80% in real-time inference workloads, making it a vital component for AI applications [1][4] Product Features - The 192GB SOCAMM2 utilizes Micron's advanced 1-gamma DRAM process technology, which contributes to its enhanced power efficiency and capacity [2][4] - SOCAMM2 is designed to meet the demands of massive-context AI platforms, providing high data throughput and energy efficiency [4][5] - The modular design of SOCAMM2 improves serviceability and supports future capacity expansion, making it suitable for full-rack AI installations [1][7] Industry Collaboration - Micron has collaborated with NVIDIA for five years to pioneer low-power server memory in data centers, establishing a foundation for SOCAMM2's development [4][6] - The company is actively participating in the JEDEC SOCAMM2 specification definition to promote industry standards for low-power memory adoption in AI data centers [8] Performance Metrics - SOCAMM2 improves power efficiency by more than two-thirds compared to equivalent RDIMMs while occupying one-third of the size, optimizing data center footprint [7] - Customer samples of SOCAMM2 are available in capacities up to 192GB per module and speeds up to 9.6 Gbps, with high-volume production aligned to customer launch schedules [8]