Toward Computing’s Extremes: Low-Power Integrated Circuits for Edge AI and Quantum Computing

With the end of Dennard scaling and the slowing of Moore’s law, improvements in general purpose computing have plateaued. In response, new paradigms are emerging at the extremes of the computing spectrum, driving new opportunities for technological and societal advancement. At the low-power end, edge AI hardware enables ubiquitous, privacy-preserving intelligence with low latency and scalability. At the high-performance end, quantum computing is gaining practicality, offering solutions to previously intractable problems in physical sciences, biology, and optimization. Despite their differences, both domains share a common, critical challenge: the need for integrated circuits (ICs) that are simultaneously energy-efficient and compact. Edge nodes must operate under tight energy and form-factor constraints, while quantum-classical interfaces must meet stringent thermal and spatial limits at cryogenic temperatures.

In this talk, I will present cross-stack designs of low-power ICs that address these challenges at both ends of the computing spectrum. The first thrust focuses on a series of energy-efficient edge AI processors that co-optimize architecture, circuit, and embedded non-volatile memory technology. The second highlights scalable cryogenic CMOS co-processing hardware for fault-tolerant quantum computing (FTQC), with emphasis on quantum error correction decoding and qubit readout. I will conclude with my vision for advancing edge intelligence and large-scale FTQC through IC innovations, and how AI-driven design automation can accelerate progress in both classical and quantum domains.

Speaker: Qirui Zhang, University of Michigan