E.ON, Qoro and NQCC split a 1,710-qubit grid problem across IBM Heron processors
A preprint from E.ON, Qoro Quantum and the NQCC partitions a 1,710-variable grid optimization into small QAOA jobs and packs them onto IBM Heron chips.
Energy company E.ON, UK software start-up Qoro Quantum and the UK's National Quantum Computing Centre (NQCC) have shown how an optimization problem that would need 1,710 qubits can be broken into pieces small enough for today's IBM processors and run in parallel. The work is described in a preprint, Scaling Quantum Optimization to the Thousand-Qubit Scale with Distributed Quantum Sampling, whose authors include Qoro co-founders Dan Holme and Stephen DiAdamo.
What was announced
The team tackled the Minimum Birkhoff Decomposition, which seeks to express a network's demand as a sparse weighted sum of matchings, sets of edges that share no vertex. The paper cites network scheduling and energy trading as applications. The test case was PEGASE-1354, a benchmark representing a 1,354-bus transmission grid with 1,710 lines; encoded directly, each line becomes a qubit.
According to Quantum Computing Report, the project ran under the NQCC's SparQ programme, which supports proof-of-concept work on quantum applications, with technical oversight from the STFC Hartree Centre.
The technical details
The pipeline combines single-layer QAOA sampling with a classical Extended Fully-Corrective Frank-Wolfe optimizer. Spectral graph partitioning cuts the problem into subproblems; QCR reports 41 subgraphs of at most 50 qubits each, executed on 156-qubit IBM Heron processors. Samples from the pieces are stitched together and a greedy classical repair step removes conflicting edges so every candidate is a valid matching.
"Spatial circuit packing", running several independent circuits side by side on one chip, cut hardware execution time by roughly a factor of three, according to the paper. QCR reports that four independent circuits, 144 qubits in total, were placed on a single 156-qubit device. On the benchmark, the authors say repaired QAOA samples reached lower residual decomposition error than simulated annealing and uniform random sampling.
The authors also checked whether the quantum step was doing real work. Shrinking partitions to 20 qubits hurt convergence, and in matrix product state simulations, run with Qoro's Maestro engine per QCR, truncating quantum correlations lowered candidate quality. They argue this supports keeping a substantial quantum sampling task in each partition as problems grow.
Why it matters
Many utility-scale optimization problems are far larger than any single quantum processor. This study offers a concrete recipe, partition, pack, repair, for running thousand-variable constrained problems on current hardware, and a benchmark grounded in a real European grid model.
What to watch
Beating simulated annealing under the same repair step is not the same as beating the best classical solvers, and the paper does not claim quantum advantage. QCR notes that grid connectivity forced dense qubit routing, with two-qubit depths reaching 169 even on 50-qubit pieces, so hardware noise remains the limiting factor.
Sources
Facts on this page come from the sources below, retrieved by DesignQuantum.
- Scaling Quantum Optimization to the Thousand-Qubit Scale with Distributed Quantum Sampling (arXiv:2610.10772) arxiv.org
- E.ON, Qoro Quantum, and NQCC Scale Constrained Optimization on IBM Heron QPUs via Distributed QAOA and Spatial Packing (Quantum Computing Report) quantumcomputingreport.com
- SparQ programme (NQCC) nqcc.ac.uk
- IBM Quantum ibm.com