Energy Efficiency Up to 2,000 Times Higher: Are Neuromorphic Chips the Future of AI?

The growth of artificial intelligence has highlighted an important issue with this technology: the enormous amount of energy it requires for data processing. To address this problem, a group of physicists from Loughborough University has proposed a radical solution that aims to change the very architecture of computing systems.

The team, led by Dr. Pavel Borisov, has developed a neuromorphic chip based on a thin film of niobium oxide with irregularly distributed nanopores. This structure creates a network of multiple electrical pathways that artificially replicates the complexity of neuronal connections in the human brain.

Unlike traditional systems, which execute algorithms through software on CPUs or GPUs, this device exploits the physical processes of the material to directly process signals that change over time. The result is a drastic reduction in energy requirements: in some scenarios, the efficiency is up to 2000 times higher than conventional approaches.

From a technical standpoint, the chip falls into the category of memristors, electronic components capable of