Nvidia expands quantum platform with fault-tolerant layer

Nvidia has enhanced its open source CUDA-Q platform by introducing CUDA-Q Logical, a tool designed to facilitate the development of fault-tolerant quantum computing applications, as reported by a press release. This new orchestration layer empowers researchers to create and evaluate quantum computing systems utilizing logical qubits, a crucial step in mitigating errors in physical qubits. The technology is aimed at sectors such as drug discovery, financial modeling, and materials development.

Fermi National Accelerator Laboratory used CUDA-Q Logical, significantly reducing the time required to develop fault-tolerant architecture, cutting from five months to just three weeks. The lab confirmed earlier findings and assessed resource requirements across various error-correction methods. According to Anna Grassellino, chief technology officer at Fermilab, the team utilized CUDA-Q Logical to examine different resource combinations within a short period - three weeks, as opposed to the usual five months for constructing specialized infrastructure.

Similarly, Iceberg Quantum employed the platform to model fault-tolerant architecture for Diraq's qubits, ultimately determining that only 150,000 physical qubits are needed to generate 1,000 logical qubits, a substantial reduction compared to earlier estimates. Sandia National Laboratories crafted QUOPS, a benchmark designed to assess the readiness of fault-tolerant quantum hardware.

This benchmark has now been incorporated into CUDA-Q, and initial QUOPS benchmarks for quantum processing units from Google, IBM, and Quantinuum were released ahead of IEEE Quantum Week.

Among the organizations currently leveraging CUDA-Q Logical are Infleqtion, IQM Quantum Computers, QCDesign Quantum Motion, and Quantum Motion. Additionally, Diraq, Anyon Computing, Quandela, and Quantum Machines are integrating their systems with Nvidia's quantum computing technologies. CUDA-Q Logical is accessible via GitHub.