Australian researchers have validated a quantum machine learning model for semiconductor fabrication on experimental data in what they described as a world-first that could reshape how future chips are designed.

A research team led by CSIRO demonstrated that quantum machine learning can outperform classical artificial intelligence in modelling Ohmic contact resistance, a critical process step in modern semiconductor device fabrication.

Published in the journal Advanced Science, the study presents what the researchers said was the first experimental validation of a quantum kernel method applied to real semiconductor process data.

The team developed a Quantum Kernel-Aligned Regressor architecture, which integrates a Pauli-Z quantum feature map with a learnable quantum kernel alignment layer.

Despite being trained on just 159 experimental gallium nitride samples, the quantum model outperformed seven classical machine learning baselines, achieving what the researchers described as state-of-the-art predictive accuracy.

Dr Zeheng Wang, lead author and quantum researcher at CSIRO, said the semiconductor industry was increasingly constrained by data scarcity and rising process complexity.

“Our results show that quantum models, when carefully designed, can capture patterns that classical models may miss, especially in high-dimensional, small-data regimes,” Dr Wang said.

The study also explored the model’s robustness under simulated quantum noise. Even at noise levels exceeding those typically observed on today’s quantum devices, the model retained predictive capability.

Professor Muhammad Usman, quantum system team leader at CSIRO, said the work offered a proof-of-concept for deploying quantum-enhanced modelling directly in semiconductor workflows.

Picture: credit CSIRO