Toyota-backed AISIN unveils MRAM-embedded edge AI chip

AISIN, in which Toyota is the major shareholder, and Tohoku University showed a 16nm SoC with built-in STT-MRAM that stores both AI model weights and the OS. It reaches kernel startup in 65.7 ms, 38 times faster than with external memory, and cuts model weight data by 53 percent while keeping 16-bit precision.

2 effects · 4 potentially affected companies · The full cascade analysis is part of the Pro plan. Create free account

AISIN and Tohoku University demonstrated a 16nm SoC with embedded STT-MRAM that holds model weights and the OS, avoiding external DRAM traffic. The pitch is faster boot, lower latency and less exposure to expensive DRAM for automotive and physical-AI inference. It is a research-stage demonstration, and AISIN has given no production timeline. It pushes on-device inference design toward in-package non-volatile memory, which competes at the margin with the Nvidia and Qualcomm automotive platforms. Near-term revenue effects are negligible, but it shows automotive suppliers want their own edge-AI silicon.