Micron Developing Edge AI Inference Memory for Physical AI Applications

Micron ($MU) is developing edge inference memory (EIM) for physical AI applications. The technology targets fields requiring low-power DRAM, such as autonomous driving and humanoid robots. ChainCatcher reported that Micron's Vice President Lee Jech (리철) disclosed this development during the SEDEX 2026 semiconductor exhibition in Seoul's COEX. Micron is researching how to apply EIM to LPDDR6 and discussing related products with key customers, according to ChainCatcher. Industry analysts expect mass production of the product to begin in the second half of 2028, as reported by ChainCatcher. Chinese media Kezhang Banyibao (科创板日报) noted that EIM is positioned between AI chips and memory, stacking LPDDR on logic chips to rapidly process inference data. Specific computing architecture details remain undisclosed, per the report. Vice President Lee explained that current robot AI models operate at 1-2 billion parameters, which does not require high bandwidth. However, more advanced vision-language models may outgrow LPDDR's capabilities, ChainCatcher added.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 마이크론, 물리 AI용 엣지 추론 메모리 개발 중…LPDDR6 적용 연구