Micron readies 'Edge Inference Memory' for humanoids and says future robots may need 200 GB of DRAM

Micron is developing Edge Inference Memory (EIM), low-power DRAM stacked on a logic die to cut latency and raise bandwidth for physical-AI devices, Micron director Lee Cheol said at a SEDEX 2026 session on humanoid chips in Seoul on Oct 16, TheElec reported; Micron has not issued a release, so the plan is unconfirmed.

Lee said Micron is applying EIM to LPDDR6 and expects high-performance humanoids to need about 200 GB of DRAM and 4 TB of storage, against 12 GB and 256 GB in today's service robots; TheElec cited industry expectations of mass production in the second half of 2028.

The timing lines up with Intel's plan, reported from the same Seoul event and covered by MEKAWIRE, to launch its roughly 700-TOPS 'Razer Lake AX' robotics chip in 2028.