October 2025

Researchers from at Southern University of Science and Technology, Xi'an Jiaotong University and other institutes recently reported a spintronic compute-in-memory (CIM) macro designed to improve computational efficiency in artificial intelligence hardware. The device is a 64-kb non-volatile digital CIM macro fabricated using 40-nm spin-transfer torque magnetic random-access memory (STT-MRAM) technology, which stores information through the magnetic orientation of nanometer-scale layers.

Conventional computing architectures separate memory and processing units, requiring frequent data transfer that increases latency and energy consumption. CIM designs address this limitation by integrating storage and computation, though most prior implementations have relied on analog operations that constrain accuracy, scalability, and robustness. The newly developed digital CIM architecture addresses these limitations by combining the endurance and non-volatility of STT-MRAM with digitally controlled computation.

Researchers from the Institut Català de Nanociència i Nanotecnologia (ICN2), CSIC and BIST have reported a theoretical framework and numerical confirmation for fully efficient spin-charge interconversion in graphene. Efficient conversion of charge current into spin current is a central objective in spintronics, and the intrinsic properties of graphene make it an attractive platform to explore this phenomenon.

Joaquín Medina Dueñas, Santiago Giménez de Castro, Jose H. Garcia, and Stephan Roche from ICN2 demonstrate that a complete conversion can be achieved by controlling the coupling between spin and pseudospin degrees of freedom. The study shows that a combined spin-pseudospin operator remains conserved in graphene, enabling fully efficient spin-charge conversion through the Rashba-Edelstein effect. The results also reveal the presence of a spin Hall effect that is resilient to disorder, indicating a stable mechanism for spin transport in realistic graphene systems.

Here are some recent and popular spintronics industry and research news that you may find of interest:

• AI framework accelerates discovery of antiferromagnets for next‑gen spintronics
• Wrinkles in 2D materials could enable efficient spintronic devices
• Breakthrough method uncovers hidden magnetic signals in non-magnetic metals
• Novel way to manipulate skyrmions will enable better memory and sensing devices
• Researchers observe spin currents in graphene without magnetic fields
• Researchers observe a new form of magnetism
• Cholesterol-based metal–organic supramolecular materials could boost spintronic devices
• Researchers show that light can interact with single-atom layers
• New antiferromagnetic spintronics project receives funding of nearly $4 million
• TDK announces the world's first "Spin Photo Detector" with 10X data transmission speeds
• New EU-funded project applies spintronics to the field of artificial intelligence
• Intel unveils a highly promising MESO spintronics device architecture
• Researchers use perovskite materials and light to control electron spins

An international research team that included researchers from Chalmers University of Technology, Kyushu University and DGIST has developed a new fabrication method for energy-efficient magnetic random-access memory (MRAM). The new method relies on a material called thulium iron garnet (TmIG) which has been attracting global attention for its ability to enable high-speed, low-power information rewriting at room temperature. 

The team hopes these new findings will lead to significant improvements in the speed and power efficiency of high-computing hardware, such as those used to power generative AI.

Researchers from China's Hangzhou Dianzi University have developed an artificial intelligence-driven framework that could accelerate the discovery of antiferromagnetic materials for spintronics. Antiferromagnets (AFMs) are prized in advanced electronics because their alternating spin orientations cancel stray magnetic fields, creating fast, stable, and densely packable devices. However, their complex magnetic interactions and vast chemical possibilities have made systematic design extremely challenging.

The team’s approach combines a crystal diffusion variational autoencoder with data augmentation (CDVAE-DA), crystal graph convolutional neural networks (CGCNNs), a genetic algorithm (GA), and density functional theory (DFT) validation into a single, integrated pipeline. CDVAE-DA learns from tens of thousands of known crystal structures and then fine tunes its predictions on an AFM-specific dataset, producing novel, chemically valid candidates with over 90% composition accuracy. These structures are rapidly screened by CGCNN models, which assess three key properties: formation energy, total magnetic moment, and electronic band gap. Candidates meeting AFM-friendly criteria—stable energy, low net magnetization, and a targeted band gap range—are passed to the optimization stage.