Quantum Supremacy Claims That Fell to Classical Simulation

In 2019, Google announced quantum supremacy: its 53-qubit Sycamore chip sampled random circuits in 200 seconds, a task estimated at 10,000 years on the best available classical supercomputer. The estimate held for roughly two years.

Between 2021 and 2022, researchers Pan and Zhang reproduced the benchmark using tensor networks on a GPU cluster, first in days, later compressed to hours. The decisive insight was not additional raw compute. It was accounting correctly for noise: Sycamore's output carried roughly 0.2 percent fidelity, meaning 99.8 percent of its samples were noise-corrupted. Matching the benchmark required only reproducing samples at that same degraded fidelity, a target exponentially easier than exact simulation. The 10,000-year estimate became a footnote.

The pattern repeats

In 2023, IBM published a 127-qubit 'utility' experiment in Nature, reporting Ising-model dynamics claimed to exceed brute-force classical methods. Within weeks, multiple groups reproduced the results classically, first with tensor networks and then with Pauli propagation, a technique that tracks how an observable spreads backward through a circuit while discarding negligible terms; some reproductions ran on a single laptop. The paper's technical claim, beyond brute-force simulation specifically, was narrower than the public reception of it, which read as quantum outperforming classical computation generally.

Annealing-based speedup claims, boson-sampling records, and several smaller advantage announcements have followed a similar arc: publication, coverage, then a classical algorithm matching the result within months.

Claim                          Year   Classical response         Time to match
Google Sycamore supremacy      2019   Tensor networks (Pan/Zhang) ~2 years
IBM 127-qubit utility          2023   Tensor networks + Pauli prop.  weeks

Why classical methods keep winning

Three structural reasons recur across these cases.

• Noise. Real hardware sets a low fidelity bar, and classical methods can trade accuracy for speed to remain beneath it.
• Structure. Any circuit computing something useful carries exploitable regularity, the same regularity that makes it useful in the first place.
• Adversarial asymmetry. A supremacy claim is a fixed target; classical algorithms continue to improve against it after publication.

The one claim that has not fallen is deep random circuit sampling on newer chips such as Google's Willow-class devices, circuits deliberately engineered to maximize entanglement and non-Clifford resources simultaneously, defeating every known compression method. These circuits compute nothing of independent interest. The current position is that quantum hardware leads on one contrived benchmark, and classical methods lead on every benchmark tied to a useful computation.

This history motivates the design of this platform directly. The techniques that answered supremacy claims, tensor networks with rigorous fidelity accounting and Pauli propagation with noise-aware truncation, are the production engines offered here, together with the accuracy statements this history has shown to be necessary. The resulting advice on hardware spending is correspondingly direct: verify that no classical method answers a given question before paying for hardware access. That check is available as a free API call.