NeuMat Spotlight: Frontiers in Science Deep Dive – Breaking the ‘Memory Wall’ with Next‑Generation AI Hardware
Featuring Prof. Kaushik Roy, Purdue University
NeuMat is pleased to highlight an upcoming Frontiers in Science Deep Dive webinar that will be of strong interest to our community working across neuromorphic materials, device engineering, and next‑generation AI hardware.
📅 Date: Thursday 12 February 2026
⏰ Time: 3:00pm GMT (16:00 CET)
🎙 Speaker: Prof. Kaushik Roy (Purdue University), with collaborators from Purdue and Georgia Tech
🔗 Event link: https://events.frontiersin.org/next-gen-ai-hardware/uni
📄 Lead article: Breaking the memory wall: next‑generation artificial intelligence hardware (Frontiers in Science, 2025)
▶ Example Deep Dive: Prof Martin Siegert’s prior session (YouTube) [frontiersin.org] [youtube.com]
Why This Matters for NeuMat Researchers & Students
As AI models scale, energy and latency bottlenecks in conventional von Neumann architectures—especially the growing “memory wall”—have become fundamental limiting factors. The featured Frontiers in Science lead article by Prof. Kaushik Roy and colleagues offers a comprehensive technical roadmap for overcoming these limits through compute‑in‑memory (CIM) and neuromorphic hardware, presenting solutions directly aligned with NeuMat’s mission.
According to the article, traditional hardware struggles because computation and memory are physically separated, causing heavy data‑movement overheads. CIM architectures integrate computation directly where data are stored, dramatically improving efficiency and supporting AI algorithms such as spiking neural networks. These approaches promise significant gains for edge AI, robotics, autonomous systems, and other power‑constrained use cases. [frontiersin.org]
A complementary Frontiers in Science announcement further emphasises how brain‑inspired computing paradigms could reduce data‑centre energy consumption and enable real‑time intelligence in compact systems such as medical devices, autonomous vehicles and drones. The webinar will expand on these themes while highlighting future research directions across materials, devices, circuits, algorithm–hardware co‑design, and systems integration. [eurekalert.org]
What to Expect in the Webinar
The 90‑minute interactive Deep Dive will include:
- An expert‑led walkthrough of the compute‑in‑memory paradigm and its potential to break AI’s memory bottlenecks
- Discussion of neuromorphic architectures, including stochastic and event‑driven approaches inspired by biological computation
- Real‑world application domains such as healthcare, transportation, robotics, and autonomous navigation
- A live Q\&A session with Prof. Roy and colleagues from Purdue and Georgia Tech
- Insights into the broader Frontiers in Science Next‑Generation AI Hardware article hub and related publications shaping the field (e.g., the 2025 lead article and expert commentaries). [frontiersin.org]
Why You Should Attend
Whether you work on materials, device architectures, circuits, or algorithmic frameworks, this Deep Dive directly intersects with NeuMat’s research aims. Students will benefit from seeing how front‑line research in neuromorphic engineering is being translated into system‑level architectures, and how the field is moving towards unified, energy‑efficient designs.
NeuMat particularly encourages participation from early‑career researchers interested in:
- Emerging non‑volatile memory technologies
- Integrated materials-to-systems design
- Neuromorphic and CIM hardware for future AI
- Cross‑disciplinary perspectives shaping next‑gen architectures
