What if your computer could think inside its own memory, like a brain? Today, processors and memory sit in separate places: one handles computation, the other stores information. For demanding applications such as AI models, this creates a major bottleneck. Vast amounts of data must constantly move back and forth, wasting energy and slowing computation. But imagine a chip where computation happens directly where the data is stored. This is the idea behind compute-in-memory architectures and it could change how future AI hardware is built.

Prof. Dr. Laura Bégon-Lours
Assistant Professor in Neuromorphic Electronics with Oxides, Integrated Systems Laboratory, D-ITET, ETH Zurich