INESC MN, a research unit of Instituto Superior Técnico, is one of the partners in the European project “MultiSpin.AI: n-ary spintronics-based edge computing co-processor for artificial intelligence”, which has secured funding of over €3M from the European Innovation Council (EIC) through the Pathfinder Open program. The project started in February 2024 and will last for 3 years.

The research consortium, led by Prof. Lior Klein from the Department of Physics at Bar-Ilan University, brings together researchers from INESC-MN (Portugal), Bar-Ilan University (Israel), and UCLouvain – Université catholique de Louvain (Belgium), as well as industrial partners such as SpinEdge LLC (Israel) and Interactive Fully Electrical Vehicles (Italy).

In Lisbon, the team is led by Prof. Susana Freitas, a researcher at INESC MN and professor at the Department of Physics of Instituto Superior Técnico, whose widely distinguished work in thin films and spintronics will be essential for the design and manufacturing of magnetic nanodevices in the project.

The Pathfinder Open program promotes high-risk, high-reward scientific innovations and the development of technologies still in the early stages. This is the case of the innovative device that the project proposes to develop, applying spintronics to the field of artificial intelligence: processing units based on magnetic tunnel junctions that exhibit multiple distinguishable magnetic states, allowing for more than the usual 2 digital states and, therefore, achieving significant gains in terms of computing speed and energy efficiency. This hardware will be integrated into information processing through neuromorphic computing. The impact will be significant in artificial intelligence applications that require high energy consumption and increased computational resources, thus benefiting from the optimization of processing capacity (for example, in areas that require “on-site” computing, such as autonomous mobility or robotics).

With the project, the development of an initial prototype and the demonstration of the proof of concept are expected, through the integration of magnetic devices into a processing array and the construction of artificial intelligence algorithms that validate it.