NVIDIA Expands CUDA-Q For Fault-Tolerant Quantum Computing

NVIDIA has expanded its open-source CUDA-Q platform with CUDA-Q Logical, an orchestration layer for fault-tolerant quantum computing. The company announced the update today to help researchers design, test and verify applications for quantum systems using logical qubits.

CUDA-Q Logical lets researchers coordinate algorithms, error-correction methods, hardware architectures and other components. The platform also lets teams switch between configurations and assess performance. 

NVIDIA said the approach targets applications including drug discovery, financial modelling and materials development.

Fault-tolerant quantum computing uses logical qubits to reduce errors found in physical qubits. These systems aim to support larger computations needed for practical applications. However, changes to algorithms or hardware can alter the resources required for an application.

“Quantum computing is maturing into an era of logical qubits,” said Timothy Costa, VP and GM of Quantum at NVIDIA. He said researchers need an open platform that can represent different parts of a fault-tolerant system.

Fermi National Accelerator Laboratory, Infleqtion, IQM Quantum Computers, QCDesign Quantum Motion and Sandia National Laboratories already use CUDA-Q Logical.

Iceberg Quantum used the platform to model a fault-tolerant architecture for Diraq’s qubits (silicon-based spin quantum bits developed by the company Diraq using standard commercial semiconductor manufacturing processes). 

Its modelling showed that 1,000 logical qubits could require 1,50,000 physical qubits. That figure is roughly 10 times lower than Diraq’s earlier estimate.

Fermilab also used CUDA-Q Logical to evaluate physical qubits, runtimes and resource requirements across different approaches. The lab reduced its fault-tolerant algorithm development timeline from five months to three weeks. NVIDIA described this as a sevenfold speed-up.

NVIDIA has also added QUOPS to CUDA-Q. Sandia National Laboratories developed the benchmark to measure progress towards utility-scale quantum computing applications. It provides a hardware-agnostic method to compare progress across fault-tolerant quantum systems.

Sandia has published early QUOPS results for quantum processing units from Google, IBM and Quantinuum. A reference implementation is available through CUDA-Q. Companies including Anyon Computing and Quandela are integrating quantum processors with GPU supercomputing through NVIDIA’s NVQLink architecture.

CUDA-Q Logical is now also available on GitHub, and NVIDIA has made the QUOPS repository available.