Academia Sinica Logo

Quantum Processing Units (QPUs) are the core of superconducting quantum computers. Academia Sinica is actively advancing scalable and reliable fabrication technologies for superconducting QPUs. A key step toward this goal is the implementation of high-throughput, iterative testing on large-scale QPUs. To overcome the limitations of traditional calibration methods, Academia Sinica has recently collaborated with NVIDIA to leverage the NVIDIA Ising family of AI models, developing AI-driven automated calibration technologies. This approach significantly improves testing efficiency and system scalability, laying the foundation for large-scale quantum chip development.

President James C. Liao noted that traditional QPU calibration has relied heavily on human experience and iterative manual adjustments, which are inefficient and difficult to scale to large systems. The NVIDIA Ising model introduces an innovative approach by interactively inferring calibration operations directly from experimental QPU data. By adopting this technology, Academia Sinica integrates quantum computing with Graphics Processing Units (GPUs), accelerating the development process for QPU fabrication and testing, and effectively narrowing the gap between current hardware capabilities and practical quantum applications.

Establishing a New Paradigm for Superconducting Quantum Chip Testing and Strengthening Academia Sinica’s QPU Platform

As highlighted by 2025 Nobel Prize laureate John Martinis, Taiwan holds a key advantage in global quantum technology development due to its advanced semiconductor manufacturing capabilities. As superconducting quantum computing continues to advance, there is a growing demand for QPUs that combine high scalability, precision, and manufacturability.

NVIDIA Ising is an open source family of models, training tools, and frameworks specifically designed for quantum computing applications. According to Dr. Chii-Dong Chen, Executive Officer of the Thematic Center for Quantum Computing at Academia Sinica’s Research Center for Critical Issues, the system has been integrated into the institute’s quantum computing platform. Through an automated multi-agent collaborative architecture, it provides several key capabilities.

The system supports closed-loop parameter optimization, continuously performing “execute–measure–adjust” cycles to automatically fine-tune experimental parameters and detect qubit frequencies. With a single natural language command, it can complete readout calibration on a quantum chip containing five qubits and four couplers, while automatically generating an experimental workflow. This demonstrates strong potential for application to the center’s recently developed 20-qubit chip.

Building Quantum Testing Infrastructure to Drive Ecosystem Development

Through its collaboration with NVIDIA, Academia Sinica is establishing a quantum testing infrastructure that is highly reliable, efficient, and scalable. This platform will be made available to domestic academic and industrial communities, promoting research, development, and application exchange in quantum technologies, further strengthening Taiwan’s overall quantum technology ecosystem.