RIKEN and IBM Demonstrate Quantum-Centric Supercomputing at Scale

RIKEN and IBM Demonstrate Quantum-Centric Supercomputing at Scale RIKEN and IBM have successfully executed a closed-loop hybrid workflow utilizing the full capacity of the Fugaku supercomputer and an on-premises IBM Quantum Heron processor. This demonstration represents a milestone in Quantum-Centric Supercomputing (QCSC), where the entirety of a pre-exascale classical system—comprising 158,976 nodes and over 7.6 million cores—worked in continuous tandem with quantum hardware. The joint team used this integrated infrastructure to calculate the electronic structure of complex iron-sulfur molecules, achieving the largest and most accurate chemistry simulation ever performed on a quantum computer to date. The technical core of the experiment centered on a closed-loop orchestration system designed to move beyond the high latency of traditional sequential processing. In this model, data is fed back and forth between the supercomputer and the QPU in an unbroken workflow, minimizing idle time for both resources. Researchers developed a specialized task assignment system to manage the interdependency of the classical and quantum tasks. This system ensured that the billion-dollar classical infrastructure of Fugaku remained active while the Heron processor performed its specific computational steps, a necessity for the economic viability of HPC-quantum integration.

The team employed the Sample-based Quantum Diagonalization (SQD) algorithm to solve the electronic structure problem. SQD is part of a new class of hybrid algorithms that partition problems into segments optimized for different compute resources. In this case, the IBM Quantum Heron processor was used to sample the enormous configuration space of the molecule’s electrons, identifying the most relevant quantum states. The Fugaku supercomputer then used this sampled data to perform the final diagonalization and drive toward a solution, effectively unlatching the most complex part of the problem with quantum resources while the classical system handled the high-volume data processing. The simulation targeted iron-sulfur molecules [Fe₂S₂(SH)₄]²⁻, which are notoriously difficult to model due to their complex electronic distributions. The accuracy achieved by the Fugaku-Heron loop surpassed previous quantum attempts and proved comparable to the most advanced classical approximation methods. This high level of precision demonstrates that QCSC can already tackle molecular problems that are at the edge of or beyond the reach of exact classical methods, providing a practical framework for applications in materials science and drug discovery. Looking forward, the researchers confirmed that the task assignment system is compatible with various cloud-based HPC environments, suggesting that the model is scalable beyond specialized laboratory setups. The next phase of the RIKEN-IBM roadmap involves integrating GPUs to further accelerate the hybrid loop and reduce execution time. By optimizing the interaction between CPUs, GPUs, and QPUs, the team aims to establish a definitive path toward quantum advantage for industrial-scale chemical and physical simulations within the current year. For full technical details on the SQD algorithm and the Fugaku-Heron integration, consult the official IBM Research blog here. March 11, 2026 Mohamed Abdel-Kareem2026-03-11T17:49:27-07:00 Leave A Comment Cancel replyComment Type in the text displayed above Δ This site uses Akismet to reduce spam. Learn how your comment data is processed.