memQ unveils CUDA-Q based compiler for linked QPUs

memQ has outlined a development plan for an Extensible Distributed Quantum Compiler designed to split a single quantum workload across multiple quantum processing units (QPUs) connected within a system or over a network.

The compiler, called xDQC, is built on Nvidia's CUDA-Q platform. It aims to distribute circuits and operations across connected processors based on available qubits and workload type.

Distributed quantum computing is attracting interest as researchers and developers look beyond single, monolithic machines. The idea is to link multiple processors and treat network connections as part of the computation, rather than a simple transport layer.

memQ describes xDQC as a hardware- and network-aware orchestration layer. It treats links between quantum processors as components that can be optimised during compilation, which it says can improve throughput on complex workloads compared with running the same job on a single device.

Workload distribution

xDQC's core feature is its ability to split a single workload across multiple quantum processors. It profiles the workload against available resources, then evaluates different routing and task-assignment options across processors. The workflow selects a plan that balances performance with resource use.

A simulation stage generates recommendations using hardware-aware noise models that account for interconnect conditions. memQ describes this as a digital twin of distributed quantum processors on a physical network. After selecting a plan, xDQC assigns tasks to processors for execution and then recombines the results into a single output.

The design assumes future systems may combine different qubit modalities and include hardware from multiple vendors-an increasingly common theme in efforts to scale beyond today's devices.

Andre Konig, CEO of Global Quantum Intelligence, tied the approach to a shift in how work is allocated across machines.

"We see the emergence of a 'right qubit for the right task' paradigm which leverages systems of different qubit modalities - and possibly different vendors - as quantum workloads increase," said Andre Konig, CEO, Global Quantum Intelligence.

CUDA-Q foundation

Nvidia's CUDA-Q serves as the underlying platform for the compiler. The platform targets hybrid quantum-classical development and simulation. memQ cited backend flexibility and GPU-based simulation as key reasons for choosing it.

Sean Sullivan, CTO of memQ, described the company's view of scaling quantum systems and the role of networked architectures.

"The industry approach to 'scale' is shifting from monolithic architectures - which will find a hard 'ceiling' - to modular, distributed computing. And the missing piece in scaling isn't just adding more qubits, it's leveraging the complex networks that connect them to unlock new applications. We're building a full-stack simulation toolkit that lets researchers co-design hardware and architecture for distributed quantum systems at scale," said Sean Sullivan, CTO of memQ.

He added that the software will be released as open source to broaden access for research and experimentation. "We chose CUDA-Q as the foundation for this solution due to its open ecosystem, backend flexibility, and GPU-accelerated simulation capabilities that allow us to profile key dynamics such as modality, circuit type, topology, and resource loads in a comprehensive way. By making it open source, we're opening the ability to co-design at scale to the entire community," Sullivan said.

Nvidia framed the work as part of broader efforts to connect quantum processors within larger, mixed computing environments.

"CUDA-Q is built to support developing workloads for at-scale hybrid quantum-classical systems," said Sam Stanwyck, Director of Quantum Product at NVIDIA. "memQ's use of CUDA-Q to provide access to QPU-to-QPU interconnected systems is a key step towards scaling and integrating quantum processors to work with tomorrow's supercomputers."

Broader stack

xDQC sits alongside memQ's xQNA portfolio, which includes quantum network interface controllers, quantum memory modules, and quantum control systems. The company's broader focus is quantum networking and connectivity across optical links between quantum computers.

memQ was founded in 2021 as a spin-out from the University of Chicago. It has focused on connectivity across local, campus, metropolitan, and wide-area environments, with an emphasis on interoperable links across different types of quantum hardware.

A preview of the CUDA-Q-based xDQC is expected in the first half of 2026.