[Edaily Reporter SOYEON KIM ] Following NVIDIA, AMD and Qualcomm are now adopting SOCAMM—a low-power DRAM (LPDDR)-based memory expansion solution—leading to the expansion of the related market. As the application of SOCAMM, developed at NVIDIA’s request, broadens across the global big tech sector, LPDDR from SamsungElectronics and SK hynix is expected to emerge as a key driver of next-generation earnings.

According to industry sources on the 27th, SOCAMM2 has entered mass production, optimized for NVIDIA’s next-generation artificial intelligence (AI) platform, “Vera Rubin,” scheduled for release in the second half of this year. AMD announced in April that it would incorporate SOCAMM2 into its next-generation EPYC server central processing unit (CPU), “Verano,” and Qualcomm is also expected to adopt SOCAMM2 for its next-generation AI accelerators and solutions.

SOCAMM is a module that reconfigures low-power memory—previously used primarily in mobile devices such as smartphones—to suit server environments. It is gaining attention as an alternative to alleviate memory bottlenecks in next-generation AI servers and is referred to as the “second-generation High Bandwidth Memory (HBM).” It is characterized by its ability to achieve a smaller form factor than existing server memory modules while also improving power efficiency.

As the focus of the AI market has recently shifted from training to inference, memory bottlenecks are becoming increasingly prominent. This is because, as AI models become more sophisticated and data center memory demand surges, it is becoming difficult to simultaneously meet various requirements—such as performance, power efficiency, and form factor—using only HBM and server-grade DRAM.

SoCam focuses on reconfiguring LPDDR—a low-power DRAM—into server modules to achieve the module scalability and ease of replacement comparable to DDR5 RDIMMs, while maintaining LPDDR’s high power efficiency. In particular, SoCam2 is regarded as a new memory tier positioned between HBM and DDR5, offering low power consumption, high bandwidth, and ease of module replacement.

The market is expanding as global tech giants are turning their attention to SoCam as one of several alternatives, given HBM’s high prices and supply constraints. SoCam, which originated from NVIDIA’s requirements, is now receiving interest from global tech giants such as AMD and Qualcomm. Since the Joint Electron Device Engineering Council (JEDEC) is also pushing for the standardization of LPDDR6-based SoCam, its scope of application is likely to expand even further in the future.

As memory demand surges, particularly in AI data centers, the LPDDR market is also growing rapidly. According to market research firm Verified Market Report, the low-power DRAM market is projected to grow at an average annual rate of 8%, expanding from $26.85 billion (approximately 37 trillion won) in 2025 to $32.6 billion (approximately 45 trillion won) by 2030.

LPDDR, which was primarily used in mobile devices such as smartphones, is effectively expanding the scope of the memory market by securing a new source of demand in AI data centers. An industry official stated, “As AI memory demand, which had been concentrated on HBM, expands to include low-power, high-capacity memory expansion solutions, LPDDR is also emerging as a new growth driver.”