D-Wave Quantum Inc. (NYSE: QBTS) has issued a detailed corporate and technical response defending its previously claimed benchmarks of “beyond-classical” quantum computational simulation supremacy. The statement addresses recent coverage surrounding the Flatiron Institute’s newly published Science manuscript on multi-dimensional tensor networks, which suggested that classical workstations could replicate physical quantum annealing state calculations. D-Wave directly refutes the premise that its 2025 milestones have been overturned or nullified, asserting that the classical framework fails to scale across the most complex problem classes, configurations, and physical measurements native to the original physical hardware demonstration.
The Technical Parameters of the Rebuttal
The central debate surrounds the simulation of continuous-time nonequilibrium magnetic spin dynamics within the transverse-field Ising model (TFIM). In its peer-reviewed 2025 Science publication, D-Wave utilized over 5,000 qubits on its Advantage2 superconducting processor across four distinct graph topologies: square, cubic, diamond, and biclique. The company estimated that matching its simulation quality on the largest architectures using classical matrix product states (MPS) would demand nearly one million years of runtime on the Frontier supercomputer, exceeding practical energy and memory boundaries.
According to D-Wave’s engineering team, the Flatiron team’s belief propagation tensor network (BP-TNS) algorithm represents an isolated advancement in classical approximation rather than a holistic replication of the quantum advantage dataset. D-Wave isolates four major technical dimensions left unaddressed by the classical paper:
- Observables Discrepancy: The BP-TNS algorithm failed to produce the comprehensive full-state data profiles and higher-order fourth-order physical observables captured by the quantum processor.
- Geometric Omissions: The classical simulations did not attempt or solve the highly complex, non-planar biclique topologies.
- Scale Limitations: The algorithm was restricted to localized sub-sectors and did not reproduce the maximum-size physical 3D geometries evaluated on the hardware.
- Coupling Inefficiencies: The approximation failed under strongly coupled, low-precision spin-glass ensembles where quantum correlations propagate at the fastest mathematical thresholds.
Cross-Verification via Quantum Hardware Reference
To map the exact failure boundaries of the classical approach, D-Wave researchers referenced data from a parallel arXiv investigation titled “Evaluating Classical Simulations with a Quantum Processor.” In this study, the Advantage2 QPU was deployed as a ground-truth physical baseline to track where the classical algorithm’s error scaling diverges. The benchmarking telemetry revealed that BP-TNS routines systematically fail to converge when confronting heavily frustrated, strongly coupled 3D spin glasses on cubic and diamond lattices. Furthermore, adding loop corrections to the BP-TNS pipeline proved mathematically ineffective when navigating higher-dimensional biclique problems, leaving these core simulation domains uniquely within the purview of physical quantum annealing capabilities.
You can review the official D-Wave corporate press release and technical response in its entirety here.
May 26, 2026
