Google Claims Quantum Supremacy With Sycamore

Google published a paper in Nature on October 23, 2019 claiming that its 53-qubit Sycamore processor had achieved “quantum supremacy” — completing a specific computational task that would take the world’s fastest classical supercomputers an astronomically long time to match. The paper had leaked two days early when a draft appeared on NASA’s website (NASA collaborates with Google on quantum computing research) on September 20, 2019, generating widespread coverage before the official publication. The task chosen was Random Circuit Sampling (RCS): apply a random sequence of quantum logic gates to the 53 qubits (one of Sycamore’s 54 qubits was non-functional), then measure all qubits to obtain a 53-bit output string. Repeat this millions of times. The output distribution of a random quantum circuit is theoretically very hard to classically simulate because the quantum state evolves through a superposition of 2^53 (approximately 9 quadrillion) simultaneous values during the computation. Google estimated that the Summit supercomputer at Oak Ridge National Laboratory — the world’s fastest at 200 petaflops at the time — would require approximately 10,000 years to sample from the same distribution that Sycamore sampled in 200 seconds. Sycamore operated at approximately 15 millikelvin (colder than the cosmic microwave background at ~2.7 K) using superconducting transmon qubits, achieving a two-qubit gate error rate of approximately 0.6 percent and a single-qubit gate error rate below 0.1 percent — among the lowest error rates achieved in a superconducting quantum processor at that scale.

IBM issued a pre-publication response on October 21, 2019, two days before the Nature paper appeared, arguing that Google’s 10,000-year classical estimate was wrong. IBM’s researchers proposed that a classical simulation using a clever tensor network contraction algorithm with sufficient disk storage (approximately 2.5 petabytes) could simulate the Sycamore circuit in approximately 2.5 days on Summit — still slower than Sycamore’s 200 seconds, but vastly shorter than 10,000 years, and potentially achievable with near-future hardware improvements. Google’s original estimate had assumed classical simulation storing the quantum state vector in RAM (which would require 2^53 amplitudes × 16 bytes per complex number = ~144 petabytes, infeasible) rather than using the approximate tensor network methods IBM proposed. The debate was technical and substantive: the “supremacy” claim’s validity depended on which classical simulation algorithm counted as the appropriate comparison baseline, and that question had no clean answer because the field of classical quantum circuit simulation was itself advancing in response to quantum hardware progress. Subsequent classical simulation improvements between 2021 and 2024 — by Chinese researchers using distributed GPU simulation, and by Amazon Web Services researchers using optimized tensor network methods — further eroded the margin of classical intractability for the specific Sycamore circuit, though Google’s hardware improved simultaneously.

The term “quantum supremacy” itself was coined by John Preskill of Caltech in a 2011 lecture as shorthand for the computational milestone where a quantum device performs a specific task beyond classical tractability; critics preferred “quantum computational advantage” as less absolutist language. The Sycamore experiment demonstrated the milestone on a specifically constructed problem with no direct practical applications — RCS is not a useful computation for any real-world task, unlike factoring integers (Shor’s algorithm) or simulating molecular energy levels (quantum chemistry applications), both of which require error correction and far more qubits than Sycamore had. IBM’s quantum advantage framing centered on “Quantum Volume” — a metric combining qubit count, connectivity, gate fidelity, and circuit depth that IBM argued was more relevant to practical near-term applications than supremacy on a contrived benchmark. For the broader field, the 2019 Sycamore result represented the first experimental crossing of the boundary where quantum hardware performance exceeded what existing classical methods could match in real time, regardless of the specific problem chosen — validating that quantum processors were advancing toward a useful capability regime rather than remaining a laboratory curiosity, and accelerating investment in quantum computing research, hardware, and cloud services from IBM Quantum, Google Quantum AI, IonQ, Quantinuum, and others through the following years.