AMD builds open software stack for quantum-supercomputer links

AMD is demonstrating an initial, testable connection between independently developed quantum and high-performance computing hardware at Supercomputing 2026, the company says. The company is building an open software stack, openQSE, with an ecosystem of partners to integrate these systems using vendor-neutral interfaces. AMD states that it will take classical and quantum systems working together on open, vendor-neutral standards. This collaborative effort aims to ground future specifications in practical implementation experience across workflows, scheduling and compilation, rather than defining a standard immediately.

OpenQSE: Bridging Classical HPC and Quantum Resources

This initial implementation focuses on establishing explicit interfaces between layers, allowing for rigorous testing of integration points within a hybrid computing stack. The demonstration will not be a final product, but a practical exercise of boundaries across workflows, scheduling, compilation, resource interfaces, and device-facing runtimes.

The openQSE project prioritizes grounding future specifications in practical experience rather than attempting to define a final standard immediately; the team intends to identify and address integration challenges through real-world testing. This approach allows for the evaluation of both staged and co-scheduled execution models, each designed to test different interface boundaries.

Staged execution decouples classical and quantum workloads in time, emphasizing orchestration and data movement, while co-scheduled execution reserves resources for simultaneous operation, focusing on reservation binding and synchronized execution. AMD is positioning its broad portfolio, including CPUs, GPUs, FPGAs, networking and the ROCm software stack, as a foundational platform for bridging classical and quantum computing. The reference implementation uses a layered execution path, connecting application workflows to quantum resources through these explicitly testable interfaces.

This model supports workflows where classical and quantum tasks can be separated in time, maintaining a consistent application execution plan. Once capacity is granted, the runtime can compile quantum work and submit it through the system, continuing classical processing while the quantum side is dispatched through scheduler and interface layers.

Staged and Co-Scheduled Hybrid Workflow Orchestration

The demonstrated implementation at Supercomputing 2026 exercises both staged and co-scheduled hybrid workflows, revealing distinct challenges in integrating classical and quantum resources. In the staged model, a workflow engine manages the entire application, dictating the sequence of preprocessing, quantum computation, and postprocessing, alongside the transfer of data between them. Classical preprocessing, currently compatible with Slurm and targeted for Flux integration, generates intermediate outputs before quantum admission evaluates resource availability for the next stage.

Quantum stages then execute within reserved time slots, returning results for subsequent classical postprocessing, which records the final outputs. This approach emphasizes orchestration and data movement, reserving and releasing resources per stage rather than maintaining a continuous execution window. The co-scheduled flow tests different boundaries, specifically reservation binding and synchronized execution across classical and quantum resources.

This dual approach is not merely academic. It reflects the reality that much of a hybrid workflow remains classical, with GPUs accelerating numerically intensive computations in preprocessing, postprocessing, and even within error-correction loops. The SC26 demonstration aims to prove interoperability between independently developed components, workflow engines, compilers, runtimes and resource managers, within a shared execution model. The team intends to demonstrate that vendor-neutral abstractions, such as QRI, can support integrations with diverse backends and that the architecture can accommodate low-latency fault-tolerant quantum computation and error correction paths without compromising abstraction boundaries, according to Supercomputing 2026.

The project team states that it is a natural proving ground for co-designing, testing and validating openQSE implementations, emphasizing the practical value of this initial integration milestone. The demonstration’s value lies in exercising boundaries, not claiming completeness.

openQSE’s Core Technical Building Blocks for Portability

Workflow systems like Tierkreis, StreamFlow, and Prefect form the application-facing layer within openQSE, responsible for both defining workflow logic and managing data transfer between classical and quantum computing stages. Currently, Slurm manages classical resources, with plans to integrate Flux as an additional option, while QFw extends control to quantum resources through components like qfw-slurm, which links Slurm requests to quantum-aware reservations. Separating workload-manager adapters from admission and dispatch logic allows for support of multiple workflow managers while clearly defining responsibility for policy enforcement, task placement and launch procedures.

The compiler layer is designed to ensure portability across different target resources, using tools such as MQT Core, Qiskit and PennyLane to prepare quantum programs for execution. These tools operate above provider-specific topology and control details, a key element in achieving vendor neutrality. Applications and compilers can express intent and constraints without being tied to specific runtime wiring.

AMD, IBM, and ORNL Co-Design for Quantum Error Correction

AMD, IBM, and the Quantum Science Center are actively co-designing a platform specifically for quantum error correction, moving beyond software integration into combined hardware testing. This effort uses the Umbra high-performance computing system, built with AMD Instinct MI355X GPUs and 5th Gen AMD EPYC CPUs, and pairs it directly with IBM quantum hardware, establishing a dedicated environment for rigorous validation. The collaborative approach aims to identify ambiguous areas and refine interface definitions through practical application, rather than solely relying on theoretical standards development.

The Supercomputing 2026 milestone extends beyond a simple demonstration, functioning as a structured method for assessing the clarity of interfaces between quantum and classical systems, the firm reports. This multi-organizational effort is designed to validate where standards require sharper definitions, and where ownership of specific components is unclear, according to the project team.

Stay current

See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.

Avatar of Ivy Delaney

Ivy Delaney

Ivy Delaney has been working with neural networks and machine learning since the mid-nineties, back when a couple of hidden layers and a long afternoon of training counted as ambitious. She has watched the field go from academic curiosity to the thing quietly running underneath everything, and she brings that long view to quantum computing. For Quantum Zeitgeist she covers the ground where the two fields meet. That means quantum machine learning and the variational algorithms it leans on, and it also means the less glamorous but more interesting story of classical machine learning already doing real work inside quantum machines, decoding error-correcting codes, calibrating noisy hardware and learning the error models that simulators depend on. She writes about the hardware those algorithms have to run on too, and about the post-quantum cryptography scramble that the same hardware has set off. Her stories typically start with the paper, whether that is peer-reviewed work, conference proceedings or an arXiv preprint, with the source linked so you can hold a claim up against the research it came from. She is unimpressed by benchmarks that will not say what they beat, and by demonstrations that only work in the press release.

Latest Posts by Ivy Delaney: