Quantum Machines links GPUs with quantum control stack

Quantum Machines has launched a framework designed to link classical processors more tightly with quantum control systems, backed by Nvidia, AMD and Riverlane. The Open Acceleration Stack lets researchers connect different types of classical compute to a quantum control stack with microsecond-level latency.

The launch reflects a wider push in quantum computing to tighten integration between quantum processors and conventional computing. Developers increasingly rely on classical processing in quantum experiments and early applications, particularly for calibration and quantum error-correction workflows that must run alongside a quantum program.

Hybrid control

The Open Acceleration Stack extends Quantum Machines' Orchestration Platform, which combines hardware and software for running and operating quantum processors. Quantum Machines positions the framework as a way to add an external classical processor, described as an XPU, directly into the control system.

The approach centres on an interface card called OPNIC, which links to Nvidia's NVQLink. It also uses Quantum Machines' Pulse Processing Unit (PPU), which runs the quantum program inside the control stack.

In practical terms, the Open Acceleration Stack is intended to connect the different compute elements used in quantum labs, including CPUs, GPUs, FPGAs and application-specific integrated circuits. The goal is a low-latency, high-bandwidth connection between those devices and the quantum control layer.

Some quantum computing tasks place heavy demands on timing and synchronisation, including decoding steps for real-time quantum error correction and AI-driven methods for qubit and circuit calibration. These workflows often require rapid back-and-forth between the quantum system and a classical processor.

Partner support

The initial configuration includes CPU-GPU-QPU integration using NVQLink, which Quantum Machines says it co-developed with Nvidia. The design is open and modular, allowing users to mix different classical processors within the same control environment.

The aim is to match the variety of hardware used in quantum computing today, where many labs combine equipment from multiple suppliers. Quantum Machines says the framework lets users choose processor types based on performance and cost constraints.

Yonatan Cohen, CTO and co-founder of Quantum Machines, said the announcement reflects a broader shift in the sector.

"The Open Acceleration Stack reflects the industry's shift from quantum computing demonstration to scaling and integration. It meets the needs of two critical areas of quantum development: real-time error correction and advanced qubit calibration, and provides the framework to scale both hardware and software with user experience and performance in mind. We are very excited to engage with our customers and the entire quantum community around these tools and their utilisation to advance the field and bring useful quantum computation closer."

Sam Stanwyck, Director of Quantum Product at NVIDIA, commented, "GPU computing is integrating deeply with quantum processors to scale logical qubits. With NVIDIA NVQLink, Quantum Machines is delivering the low-latency, high-bandwidth performance to power real-time quantum error correction and calibration - bringing practical, large-scale quantum-GPU supercomputing closer to reality."

Madhu Rangarajan, Corporate Vice President, Computing, Enterprise and AI technologies, AMD, said, "The path to scalable quantum computing depends on heterogeneous architectures that combine quantum processors with high-performance classical compute. The Open Acceleration Stack enables integration of AMD CPUs and adaptive computing technologies into the quantum control layer, supporting demanding workloads such as real-time error correction and complex hybrid algorithms. By bringing leadership high-performance compute into quantum control systems, AMD is helping enable the next phase of scalable quantum research and deployment."

Steve Brierley, Founder and CEO of Riverlane, said, "Fault-tolerant quantum computing depends on fast, reliable quantum error correction running in real time. Delivering that capability requires specialised classical infrastructure tightly integrated with the quantum control stack to produce reliable logical qubits at scale. Open frameworks such as Quantum Machines' Open Acceleration Stack help enable this level of integration, representing an important step toward scalable, error-corrected quantum computers."

Demonstrations planned

Quantum Machines will show live demonstrations of the Open Acceleration Stack at the APS Global Physics Summit in Denver. One is expected to show fault-tolerant quantum phase estimation using an OPX1000-based system. Another is planned to connect remotely to qubits at the Illinois Quantum and Microelectronics Centre and demonstrate real-time qubit calibration.