IBM has published a detailed reference architecture for quantum-centric supercomputing, providing a blueprint for integrating quantum processing units (QPUs) into existing high-performance computing (HPC) infrastructure without disruptive changes to current systems.
The release marks a significant step toward realising the vision articulated by physicist Richard Feynman, who argued decades ago that accurately simulating nature would require quantum-mechanical computation.
The architecture describes how quantum and classical systems, including CPUs, GPUs, and QPUs, can work together across multiple layers, from application and middleware tools such as Qiskit and CUDA through to resource management systems that orchestrate workloads in real time.
New algorithms such as Sample-based Krylov Quantum Diagonalisation (SKQD) are already demonstrating cases where quantum-centric workflows outperform leading classical-only methods, including in molecular ground-state energy calculations, where classical techniques failed to converge.
Real-world research applications are already emerging.
Scientists at the Cleveland Clinic Foundation used quantum-centric methods to simulate a 300-atom protein, the largest molecular simulation to date. In contrast, a team spanning IBM, Oxford, ETH Zurich, and other institutions used quantum algorithms to study a newly engineered ‘half-Möbius’ molecule whose electronic structure cannot easily be modelled classically.
IBM describes the trajectory as pointing toward a near future in which quantum computing can predict molecular properties that scientists can then bring to life in the laboratory.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
