New wafer-scale memristor integration advances prospects for brain-inspired AI chips

by Riko Seibo
Tokyo, Japan (SPX) Nov 12, 2025

A research team led by Professor Sanghyeon Choi from DGIST has achieved the mass-integration of memristors at wafer scale, marking a major development in semiconductor technology for artificial intelligence systems modeled after the human brain.

The human brain contains approximately 100 billion neurons and 100 trillion synapses, permitting vast information storage and processing within a compact space. Research efforts in next-generation AI aim to replicate this efficient structure in "brain-like AI chips," yet current semiconductor devices require intricate circuitry and significant power, resulting in low efficiency compared to biological systems.

Memristors, which store the extent of current passed through them, combine memory and computation functions in a single semiconductor device. Their simple architecture supports much higher circuit density than conventional designs. Specifically, the crossbar arrangement enables dozens of times more information to be stored in the same area compared to SRAM chips.

Efforts to integrate memristors have previously been limited by manufacturing complexity, low product yield, voltage losses, and current leakage. These constraints have prevented full wafer-scale production.

Professor Choi's team collaborated with Dr. Dmitri Strukov's group from UC Santa Barbara, developing a co-design strategy spanning materials, components, circuitry, and algorithms. The result was a memristor crossbar circuit that reached roughly 95% product completion rate across a four-inch wafer without requiring complex processes.

The researchers also demonstrated a three-dimensional vertical stacking structure, supporting the expansion of memristor circuits for larger-scale AI computation.

Testing included spiking neural networks based on the proposed integration technology, showing notable operational efficiency and stable performance during actual AI computational tasks.

Professor Choi stated, "This study proposed a method for improving memristor integration technology, which had been limited in the past" and added, "We are expecting it to lead to the development of a next-generation semiconductor platform in the future."

Research Report:Wafer-scale Fabrication of Memristive Passive Crossbar Circuits for Brain-scale Neuromorphic Computing