TetraMem announced successful tape-out and validation of its MLX200 platform, a 22nm multi-level RRAM-based analog in-memory computing SoC. This Silicon Valley-based semiconductor company has achieved a critical milestone by developing technology that performs computation directly within memory arrays, addressing growing challenges in AI systems related to data movement and power consumption.

The multi-level RRAM technology demonstrated at TSMC's 22nm process offers CMOS compatibility with minimal additional complexity, low-voltage operation, and strong retention characteristics. Early silicon results show consistent functionality across arrays, validating the approach for both embedded non-volatile memory and compute-in-memory applications. This builds on TetraMem's earlier MX100 platform work at 65nm.

The MLX200 and MLX201 platforms target power-sensitive edge AI applications including voice processing, wearable devices, and IoT systems. Evaluated sampling is expected to begin in the second half of 2026, with memory IP available for licensing. Dr. Glenn Ge, TetraMem's CEO, emphasized this approach provides a practical path to improving energy efficiency for next-generation AI systems.