Micron Becomes First to Ship Both HBM3E and SOCAMM Memory for NVIDIA AI Platforms

Micron Technology has officially announced that it is the first and only memory supplier in the world to simultaneously ship HBM3E and SOCAMM (Small Outline Compression Attached Memory Module) products—two next-generation memory solutions purpose-built for AI data center servers.

Micron Technology has officially announced that it is the first and only memory supplier in the world to simultaneously ship HBM3E and SOCAMM (Small Outline Compression Attached Memory Module) products—two next-generation memory solutions purpose-built for AI data center servers.

Powering NVIDIA’s Cutting-Edge Grace Blackwell AI Superchips

The SOCAMM memory module, co-developed with NVIDIA, is a compact, modular LPDDR5X solution specifically designed to support the NVIDIA GB300 Grace Blackwell Ultra superchip. Compared to traditional RDIMM modules, SOCAMM offers:

Higher bandwidth

Lower power consumption

Smaller physical footprint

This makes SOCAMM particularly well-suited for dense data center configurations and compute-intensive AI workloads.

HBM3E Memory Speeds Up the Most Demanding AI Applications

Micron’s HBM3E 12-High (12H) 36GB stack is optimized for NVIDIA’s HGX B300 NVL16 and GB300 NVL72 systems, delivering:

Over 1.2 TB/s memory bandwidth

>9.2 Gb/s pin speed

Industry-leading energy efficiency

Meanwhile, the HBM3E 8-High (8H) 24GB variant powers NVIDIA’s HGX B200 and GB200 NVL72 platforms. The deployment of both HBM3E versions across NVIDIA’s Hopper and Blackwell architectures underscores Micron’s critical role in accelerating AI workloads.

Micron’s Vision: Memory at the Heart of AI Transformation

According to Raj Narasimhan, Senior VP and GM of Micron’s Compute and Networking Business Unit:

“AI is driving a paradigm shift in compute, and memory is at the core of this transformation. Micron’s HBM and low-power memory solutions boost the computational efficiency of GPUs, helping to break performance barriers.”

SOCAMM Enters Mass Production, Boosting Micron’s AI Market Momentum

Micron confirmed that SOCAMM is now in mass production, delivering:

Unmatched power efficiency

Superior serviceability

Maximum density in a compact form

Optimal performance for AI servers and data-intensive applications

Designed to meet the growing demands of generative AI, SOCAMM is poised to become a key enabler of the next era of data center innovation.

Strong Financial Momentum in Data Center Memory

In its FY2025 Q2 earnings, Micron reported:

$8.05 billion in revenue, up $2.23 billion YoY

Significant performance uplift driven by 2x growth in data center revenue

CEO Sanjay Mehrotra expects a record-breaking Q3, citing strong demand for DRAM and NAND in both enterprise and consumer markets

With SOCAMM and HBM3E products now shipping, Micron is strategically positioned to capitalize on the booming AI infrastructure demand and lead the memory market for next-gen compute.

Related Articles