Micron Technology Develops AI-Specific Edge Inference Memory

Micron Technology is advancing its development of Edge Inference Memory (EIM) tailored for physical artificial intelligence applications, as announced by Vice President Li Zhe at the Korea Semiconductor Exhibition SEDEX 2026 in Seoul, South Korea. The company aims to begin mass production of this innovative memory technology in the second half of 2028, targeting low-power DRAM needs for sectors such as autonomous driving and humanoid robotics.

EIM is designed to enhance the efficiency of AI processing by integrating LPDDR memory with logic chips, facilitating rapid inference data processing. While the complete computing architecture remains undisclosed, Micron is actively engaging with major clients to explore the potential applications of EIM. Li Zhe highlighted that current robotic AI models operate with 1 to 2 billion parameters and have low bandwidth requirements, but as more sophisticated visual language models are developed, the need for High Bandwidth Memory (HBM) will arise for humanoid robots.

Micron's projections indicate that future humanoid robots will necessitate around 200 GB of DRAM and 4 TB of storage, significantly more than the current requirements of service robots, which stand at approximately 12 GB of DRAM and 256 GB of storage. CEO Sanjay Mehrotra noted that the storage capacity for a single humanoid robot could be ten times greater than that of a typical L2+ passenger vehicle. The demand for DRAM in the humanoid robotics sector is expected to surge, with Counterpoint Research forecasting an increase from about 8,300 Tbit this year to approximately 172,000 Tbit by 2030, driven by the need for extensive data synchronization and local model execution for rapid control.

© 2026 KLEA News. All Rights Reserved. This article is provided for informational purposes only. It is not offered or intended to be used as legal, tax, investment, financial, or other advice.

Source: KLEA News