NVIDIA's Open-Source AI Models Enhance Scalability in Quantum Computing

[Elif Özaksu]

NVIDIA has introduced the Ising family, open-source AI models designed to tackle fundamental challenges in quantum computing. These models aim to enhance the calibration and error correction processes of quantum processors, making quantum systems more reliable and scalable for practical applications.

### AI-Controlled Layer for Quantum Systems
Quantum computing faces significant scaling hurdles due to the instability of qubits and high error rates. Continuous calibration and real-time error correction are essential for building large and reliable systems.

Ising models deploy AI as a control layer for quantum hardware. This allows for automatic interpretation of quantum measurements and dynamic adjustment of the system. Hybrid architectures are supported, where AI and classical computing resources, such as GPUs, work synchronously with quantum processors.

AI integration enables continuous processing and optimization of data within the digital supply chain of quantum computing, feeding it back into operations.

### Models for Calibration and Error Correction
The Ising family consists of two main sets of models:

• Ising Calibration: A vision-language model that interprets measurement data from quantum processors. This model automates calibration processes, reducing operations that used to take days to just hours.
• Ising Decoding: 3D convolutional neural network models developed for quantum error correction. They offer 2.5 times faster performance and 3 times higher accuracy compared to existing open-source references like PyMatching.

These models are critically important for maintaining qubit coherence and the accuracy of computations in quantum systems.

### Open Source and Flexible Application
Since Ising models are open source, researchers and companies can easily adapt them to their specific hardware architectures and use cases. The models can be run locally, ensuring control over sensitive data while reducing dependency on external infrastructure.

NVIDIA also provides supporting resources such as workflow templates, training datasets, and microservices for model deployment and optimization. This ecosystem enables rapid experimentation and adaptation across different quantum platforms.

### Integration with Hybrid Quantum-Classical Platforms
Ising models are integrated into NVIDIA's quantum computing stack, which supports real-time data communication between quantum processing units (QPUs) and GPUs via NVIDIA's CUDA-Q and NVQLink interconnect technologies.

This integration allows quantum algorithms to be executed in conjunction with classical operations, playing a critical role in moving quantum systems beyond experimental stages and making them ready for application.

### Towards Scalable Quantum Computing
The implementation of AI-powered calibration and error correction models is a significant step towards making quantum computing more practical. The Ising family enhances performance and reliability by supporting the transition from small-scale experimental devices to larger, application-ready quantum systems.