What will it take to make artificial intelligence more energy efficient, secure, and sustainable? Why are researchers looking beyond conventional computing architectures, and how could emerging memory technologies help overcome the growing power demands of AI?
In the latest episode of the NeuMat Podcast, recorded during the NeuMat Winter School 2026, host Dip Das is joined by Prof. Regina Dittmann (Peter Grünberg Institute, Forschungszentrum Jülich), Dr Markus Hellenbrand (University of Cambridge) and Dr Alex Serb (University of Edinburgh) for a wide-ranging discussion on the future of neuromorphic computing.
The conversation explores one of the biggest challenges facing modern AI: energy consumption. As large language models and data centres continue to grow in scale, the speakers discuss why simply expanding existing computing infrastructure may not be sustainable, and how new hardware paradigms could dramatically improve efficiency by bringing memory and computation closer together.
A central theme of the episode is resistive RAM (ReRAM) and memristive devices, technologies that are enabling researchers to rethink conventional computing architectures. The panel examines how defects, once considered something to eliminate from electronic materials, have become a key resource for creating novel memory devices. This shift in thinking has opened up exciting opportunities for interdisciplinary collaboration between materials scientists, device physicists, engineers, and computer scientists.
The discussion also covers the challenges of translating laboratory discoveries into practical technologies. Topics include analogue in-memory computing, reliability and reproducibility of multi-state memory devices, CMOS compatibility, materials selection, system-level integration, and the need for co-design across the entire technology stack, from materials and devices through to circuits and architectures.
Beyond technology, the speakers reflect on the broader societal implications of AI. They discuss data privacy, edge computing, European technological sovereignty, and the importance of developing AI hardware that is both energy efficient and aligned with societal values. The episode also highlights the critical role of interdisciplinary research networks such as NeuMat in bringing together diverse expertise to tackle these complex challenges.
The panel concludes by offering advice for early-career researchers entering the field, identifying some of the most exciting open questions in neuromorphic materials and devices, and discussing the opportunities that still exist for major breakthroughs in switching mechanisms, memory technologies, and next-generation AI hardware.
Whether you are interested in materials science, electronic devices, computer engineering, AI, or research policy, this episode provides a fascinating glimpse into the opportunities and challenges shaping the future of neuromorphic computing.
Listen now and discover how advances in materials, memory technologies and hardware design could help drive the next revolution in artificial intelligence.
