MilestoneQuantum advantageConfirmed

Classical computers catch up with Google's 2019 quantum supremacy task

Nov 4, 2021

Feng Pan, Keyang Chen and Pan Zhang report a million samples for Google's 53-qubit Sycamore task in about 15 hours on 512 GPUs. Their November 2021 preprint, published in Physical Review Letters in 2022, separately estimates that an exascale machine could beat Sycamore.

Why it matters

A task Google said would take a supercomputer 10,000 years was done classically in hours, showing that quantum advantage claims depend on the best classical methods available at the time.

Pan Zhang · Feng Pan · Keyang Chen · CAS · Google

See it on the timeline

Sources (5)

arXiv (Pan, Chen and Zhang) Solving the sampling problem of the Sycamore quantum circuits
Primary sourceNov 4, 2021

“If our algorithm could be implemented with high efficiency on a modern supercomputer with ExaFLOPS performance, we estimate that ideally, the simulation would cost a few dozens of seconds, which is faster than Google's quantum hardware.”
Physical Review Letters Solving the Sampling Problem of the Sycamore Quantum Circuits
Primary sourceAug 22, 2022

“Solving the Sampling Problem of the Sycamore Quantum Circuits”
The Quantum Insider Researchers Use Classical Supercomputer to Challenge Google's Quantum Supremacy Claim
Independent coverageAug 4, 2022

“In a study published in the Physical Review of Letters, a research team from the Institute of Theoretical Physics at the Chinese Academy of Sciences, led by statistical physicist Pan Zhang, reported they beat Google's computer — Sycamore — at the same computational task that was examined in the 2019 study.”
TechCrunch Google's quantum supremacy usurped by researchers using ordinary supercomputer
Independent coverageAug 5, 2022

“Zhang's group claims to have done it in 15 hours. And if they had access to a proper supercomputer like Summit, it might be accomplished in a handful of seconds — faster than Sycamore.”
TechCrunch Google's quantum supremacy usurped by researchers using ordinary supercomputer — GPU cluster
Independent coverageAug 5, 2022

“The mathematical relationships between these tensors (each its own set of interrelated vectors) was then calculated using a cluster of 512 GPUs.”

Added Oct 15, 2026 · Updated Oct 24, 2026. Spotted an error? Tell us and we will review it.