How a Classical Tensor‑Network Algorithm Recreated D‑Wave’s “Quantum‑Only” Simulation

Quantum computing companies often claim breakthroughs when their hardware solves problems thought to be impossible for classical computers. But a recent study from physicists at the Flatiron Institute and Boston University shows how quickly that boundary can shift.

By developing a more efficient classical simulation method based on tensor networks and belief propagation, the researchers reproduced the same quantum‑dynamics problem used in D‑Wave’s widely publicized 2025 quantum‑supremacy claim. In some cases, their simulation was efficient enough to run on modest hardware—including a personal laptop—challenging the idea that the task required a quantum computer.

The Original Claim: A Quantum Computer vs. a Million‑Year Supercomputer

In March 2025, D‑Wave announced results from its Advantage2 superconducting quantum annealer, a system with roughly 5,000 qubits. The company reported that the machine simulated complex magnetic‑materials dynamics in under about 20 minutes, while estimating that performing the same calculation on a classical supercomputer would take close to one million years.

The task involved simulating the dynamics of disordered magnetic systems—specifically Ising spin‑glass models arranged on lattice structures. According to the company and its collaborators, existing classical techniques struggled to scale to these large quantum systems, leading them to describe the result as a demonstration of “beyond‑classical” computation.

Such claims fall under the concept of quantum supremacy (or quantum advantage): the point where a quantum device performs a task that classical computers cannot complete within any practical timeframe.

The Classical Breakthrough

The Flatiron–Boston University team revisited the exact same physical simulation problem. Instead of attempting to track the full quantum state directly—a calculation that grows exponentially with the number of qubits—they exploited mathematical structure in the system.

Their method combined several ideas:

• Two‑ and three‑dimensional tensor networks arranged to mirror the lattice geometry of the spin system
• Belief propagation, a message‑passing algorithm that helps update the network efficiently
• Tensor‑network compression techniques that keep the representation manageable as entanglement grows

By evolving lattice‑specific tensor networks and using belief‑propagation updates during the simulation, the algorithm could follow the system’s dynamics without explicitly representing the full 5,000‑qubit wavefunction. This dramatically reduced the computational cost while maintaining high accuracy.

According to the researchers, the approach can accurately and efficiently simulate the same quantum‑annealing dynamics that were previously claimed to be beyond classical reach.

Why Some Simulations Could Run on a Laptop

Tensor networks work by compressing quantum states. Instead of storing every amplitude of the exponentially large wavefunction, they capture only the correlations that actually appear in the system.

For many physical systems—especially those with structured lattice interactions—the amount of entanglement grows in a way that can still be approximated compactly. When that happens, tensor‑network representations can track the system with far fewer parameters than a brute‑force simulation would require.

In the Flatiron study, combining these tensor‑network representations with belief propagation made the simulation efficient enough that some instances could run on ordinary personal computers rather than massive supercomputers.

What This Means for “Quantum Advantage”

The result does not prove that quantum computers lack advantages. Instead, it highlights an important reality in the field: the benchmark keeps moving.

Quantum‑advantage claims typically compare quantum hardware against the best known classical algorithms at the time. But classical methods—especially tensor networks, Monte Carlo approaches, and other approximate techniques—continue to improve rapidly.

That means a task that appears impossible for classical machines today may become tractable tomorrow if someone finds a better algorithm. The Flatiron work demonstrates exactly that dynamic: the bottleneck was not fundamental computational limits, but the state of classical algorithms used for comparison.

The Ongoing Arms Race

As a result, the standard for demonstrating quantum advantage is becoming stricter. Researchers increasingly look for problems where:

• classical algorithms are unlikely to improve dramatically
• quantum hardware offers a clear scaling benefit
• results can be verified independently

The episode illustrates a broader pattern in computational science: progress often comes from both hardware and algorithms. Quantum processors are improving, but classical algorithms are evolving just as quickly.

For now, the competition between the two remains an active arms race—and each new claim of quantum advantage must survive the next breakthrough in classical computation.