Hybrid Classical–Quantum Processing for Real-Time Calibration of Transmon Qubits

Utility-scale quantum processors, comprising hundreds of thousands to millions of qubits, will need to operate reliably over algorithmic runtimes ranging from minutes to weeks [1]. Under such conditions, real-time calibration becomes critical, as parameter drifts are unavoidable. Previous work has shown that FPGA-embedded feedback loops can correct frequency drifts and improve coherence in spin [2] and superconducting qubits [3]. In this work, we present one of the first demonstrations of a low-latency hybrid classical–quantum link used to improve the performance of a superconducting qubit system. Flux-tunable transmons are controlled by a quantum controller connected directly to a GPU–CPU accelerator through a low-latency interface. We implement both direct and Bayesian estimation routines that allow the classical processor to continuously track system parameters and dynamically correct drifting quantum gates in real time. By leveraging the controller’s capability for in-line pulse updates and interleaved quantum–classical execution, running short calibration sequences alongside user algorithms without recompilation, we achieve concurrent estimation and correction of multiple gate-critical parameters. Our results demonstrate that tight integration between high-performance classical computation and quantum control enables efficient real-time parameter adaptation, enhancing gate fidelity and coherence. This approach provides a scalable path toward maintaining quantum performance in large, continuously operating processors.

[1] Mohseni, M. et al. How to Build a Quantum Supercomputer: Scaling from Hundreds to Millions of Qubits. arXiv:2411.10406 (2025).

[2] F. Berritta et al., PRX Quantum 6, 030335 (2025).

[3] Vepsäläinen, A. et al. Improving qubit coherence using closed-loop feedback. Nat. Commun. 13 (2022).