Micron has begun sampling its 256 GB DDR5 registered dual in-line memory modules to key server ecosystem partners. The modules achieve speeds up to 9,200 MT/s, representing more than 40% improvement over current volume production modules. Built on Micron's advanced 1-gamma technology, these RDIMMs target AI and high-performance computing infrastructure requiring massive memory capacity.

The modules employ 3D stacking (3DS) with multiple memory dies connected by through-silicon vias (TSVs). According to Micron, a single 256 GB module can reduce operating power by over 40% compared to using two 128 GB modules, delivering significant efficiency gains for data centers building AI infrastructure at scale.

The rapid proliferation of large language models, agentic AI, and high-core-count CPU workloads is driving urgent demand for greater server memory capacity. Raj Narasimhan, senior vice president of Micron's Cloud Memory Business Unit, stated that capacity, bandwidth, and power are the defining drivers of AI efficiency. The company is collaborating with key ecosystem partners to validate the modules across current and next-generation server platforms.