The countless switching operations occurring every second in the integrated circuits of our devices consume vast amounts of energy worldwide, a demand expected to rise sharply as AI continues to scale. Recent advances in materials with non-trivial electronic topology have opened new pathways for rapid, low-energy switching that could revolutionize data storage and processing beyond conventional CMOS logic.

Our project, LEAP (Low-Energy, AI-Informed Phase Transitions), aims to develop the first agentic AI system that incorporates theory, experiment, and literature to accelerate the discovery of novel switching mechanisms in topological materials. We are partnering with AI experts, theorists, and experimentalists across the UC system, including 5 UC campuses and 3 National Labs. By employing physics-informed AI to accelerate next-generation topological platforms, LEAP seeks to enable transformative reductions in power consumption and pave the way toward truly scalable, energy-efficient “AI-at-scale”.