Weebit Matrix brings ReRAM-based in-memory AI research to real silicon

What is changing?

On November 4, Weebit Nano launched the Matrix In-Memory Compute Research Program. Available to qualified academic and industry researchers, each package includes chips, documentation and a development board with a socket. It is intended for prototyping workloads such as AI inference and neuromorphic computing on real ReRAM hardware.

The Matrix chip uses a 32×32 crossbar array to support multiply-accumulate operations, vector-matrix multiplication, convolution and spiking neural-network experiments. Applications are evaluated against the program’s research focus; the platform is for research and prototyping and does not grant commercial-use rights.

Weebit says in-memory computing could reduce data movement between processors and memory. Its power and latency claims still require independent validation, and the official announcement acknowledges that technical challenges remain before a commercial IMC product.