Late last year, Google announced that its “Willow” quantum chip hammered through a calculation under minutes that would take one of the world’s fastest supercomputers 10 septillion years to handle. Now, D-Wave has shown that its superconducting quantum annealing processors can simulate the complex dynamics of quantum materials in moments what would take the Frontier supercomputer millions of years to compute. This new research, published in Science, demonstrates that D-Wave’s approach outperforms classical methods by millions of years of compute time, tackling simulations previously considered impossible.
To validate these claims, the D-Wave team compared their quantum processor’s results against the most powerful classical simulation techniques available. Specifically, they employed Matrix Product State (MPS) calculations, a state-of-the-art method for simulating quantum systems that is especially effective in lower dimensions. Running these MPS simulations on the Frontier supercomputer at Oak Ridge National Laboratory still demanded enormous computational resources. While Frontier’s exascale capabilities can tackle smaller-scale versions of these quantum systems, the exponential scaling meant that matching the D-Wave processor’s performance on larger instances quickly pushed classical simulation times into astronomical territory.
Thus quantum annealers can answer questions of practical importance that may remain out of reach for classical computation.
For modestly sized problems, MPS served as a high-precision “ground truth,” with Frontier’s hybrid CPU-GPU architecture executing the necessary calculations on the Schrödinger equation. Yet as the system size grew, the time required for MPS to reach D-Wave’s accuracy increased exponentially. According to the paper, reproducing D-Wave’s largest-scale results under ideal conditions “would exceed annual global electricity consumption” and take “millions of years” of classical compute time—an estimate that underscores the extent of the speedup. As the authors note, “Thus quantum annealers can answer questions of practical importance that may remain out of reach for classical computation.”
Beyond highlighting raw computational muscle, D-Wave’s research has potential impacts across a range of industries. Their focus on simulating spin-glass models speaks directly to problems in materials science, where understanding disordered magnetic systems can shed light on next-generation components for energy storage and electronics. These same “glassy” dynamics also appear in a variety of optimization and machine learning tasks—indeed, the paper points out that certain spin-glass topologies connect to “generative artificial intelligence” (Ref. 36). If quantum annealers continue to refine these simulations at scale, they could accelerate design cycles in manufacturing and fuel breakthroughs in AI model training, especially where classical methods fail to capture the full quantum complexity.
Looking ahead, the D-Wave team acknowledges that challenges remain. Even though their device exhibits “beyond-classical” behavior on carefully chosen spin-glass systems, hardware noise and finite precision are ever-present obstacles, and classical methods will doubtless keep evolving. The paper also highlights the need for more sophisticated error-mitigation techniques and continued expansion of programmable qubit counts to tackle broader problem classes. But as the authors note, the research could also pave the way to advances in a range of industries, including in materials science, condensed matter physics and beyond while the limits of classic computers grows more obvious over time.
