TetraMem's 22nm AI chip MLX200 passes validation, targets low-power edge devices

Silicon Valley AI chip startup TetraMem says its 22nm SoC MLX200 has passed validation on TSMC's process, with evaluation kits expected in the second half of 2026. The chip targets low-power, low-latency applications like wearables, edge IoT, sensors, and embedded systems.

MLX200 uses 'analog in-memory computing' with RRAM arrays and mixed-signal engines to run vector-matrix multiplication—key for neural networks—directly in memory. This cuts data travel distances, enabling efficient, local computation for better energy use and throughput.