From Bits to Spins
A New Approach to Computing
- von Juliana Fischer
- 20.03.2026
The rapid rise of AI and related technologies is driving a sharp increase in energy demand. Applications such as ChatGPT and complex optimisation algorithms are pushing up energy use in data centres worldwide. Researchers are therefore exploring new approaches to make computer hardware more efficient. One promising direction is spin-based computing. In a new review article published in Nature Reviews Physics, researchers – including Prof. Dr Karin Everschor-Sitte from University of Duisburg-Essen – outline how spintronic devices could form the basis of novel computer architectures.
Today’s computers rely on binary code made up of zeros and ones – for example, sequences such as 01010000011. These are based on transistors that translate electrical signals into digital states. However, as the demands of data-intensive applications continue to grow, this approach is increasingly running up against physical and energy limits.
Spintronics takes a different approach: it utilises not only the electrical charge of electrons, but also their magnetic property, known as spin. This additional degree of freedom opens up new ways of storing and processing information.
Magnetic materials have properties that make them particularly attractive for modern computing approaches. They are non-volatile, respond quickly and exhibit complex dynamic behaviour such as non-linearity, randomness and temporal feedback. These characteristics can be harnessed for neuromorphic and probabilistic computing methods – in other words, for algorithms inspired by biological nervous systems or designed to operate under uncertainty.
The review outlines a range of potential building blocks. These include spintronic neurons and synapses, probabilistic bits (p-bits), and larger computing architectures such as magnetic reservoir computing and so-called Ising machines. These are particularly well suited to solving complex optimisation problems. “We are particularly investigating how reservoir computing can be realised using magnetic structures such as so-called skyrmions,” explains Prof. Dr Karin Everschor-Sitte, physicist at University of Duisburg-Essen. “Another key aspect of our work is developing new metrics that allow us to reliably assess the performance of such systems.”
One advantage of this approach is that many spintronic components can be integrated with existing semiconductor processes. Magnetic tunnel junctions are already used in commercial memory technologies and can be incorporated into conventional CMOS fabrication processes.
At the same time, research and development still face challenges. These include the fine-tuning of materials, devices and algorithms, as well as the development of suitable benchmarks to compare the performance of new hardware with established systems.
In the long term, experts do not expect spin-based technologies to replace conventional computers, but to complement them. Hybrid approaches that combine different physical computing principles appear particularly promising. Spin-based systems could therefore help to meet the growing computational demands of modern data-driven applications in an energy-efficient way.
Further Information:
Read the Publication in Nature Reviews Physics here:
Prof. Dr. Karin Everschor-Sitte, University of Duisburg-Essen, Faculty of Physics, Tel. 0203/37 9-472, karin.everschor-sitte@uni-due.de