Artificial Intelligence (AI) is changing the world—not only how we interact with technology, but also how we conduct scientific research. Yet behind AI’s impressive capabilities lies a problem: the energy and hardware demands of current AI systems are enormous. To address this, researchers are exploring entirely new ways to build computers that work more like the human brain—and do so with much greater energy efficiency.
One promising path is called magnonics. Instead of using electricity to carry and process data, magnonics uses tiny waves—called spin waves—that travel through magnetic materials. These waves can encode and manipulate information using their phase and amplitude, much like neurons in the brain use electrical signals. They operate at very high frequencies (in the gigahertz range), enabling ultra-fast and compact devices.
This international project is coordinated on the Austrian side by Univ.-Prof. Dr. Andrii Chumak, with Univ.-Prof. Dr. Dieter Süss from the Faculty of Physics as co-principal investigator. PD Dr. Claas Abert from the Physics of Functional Materials group will play a key role in advancing the simulation work. The Hungarian partner group is based at the Pázmány Péter Catholic University in Budapest and is led by Assoc. Prof. Dr. Gyorgy Csaba. The project also benefits from a strong international network, with valuable contributions from Prof. Dr. Markus Becherer (TU Munich, Germany), Dr Michal Urbánek (CEITEC, Brno, Czech Republic) and PD DS Dr Roman Verba (V. G. Baryakhtar Institute of Magnetism of the NAS of Ukraine).
The MagNeuro project aims to take a big step forward in this field by building the first truly scalable magnonic neural networks. The idea is to create small processing units, designed by artificial intelligence through a method called inverse design, which are capable of recognizing patterns—like vowel sounds in speech—by using wave interference. These building blocks are then linked together using nanoscale amplifiers that boost the wave signals, making it possible to combine them into larger networks, much like how layers of neurons form a brain. In the framework of the project, new nano-scale localized parametric amplifiers will also be developed. These are essential for amplifying weak signals and enabling the cascading of individual computing units into an integrated, multi-layered network.
MagNeuro will pave the way for an entirely new class of computing hardware: faster, smaller, and far more energy-efficient than today’s silicon-based technology. This research could have major impacts on how we build future AI systems, process information in telecommunications, and design devices for everything from speech recognition to data security.
- Scalable Magnonic Neural Networks ( FWF MagNeuro)
Grant DOI 10.55776/PIN1434524
Duration: May 1, 2025 – April 30, 2029
