IBM’s New Quantum Processor Did in 19 Seconds What a Supercomputer Would Need 110 Years For

September 30 (IT Home) — A research team at the US company BlueQubit has completed a random quantum circuit sampling experiment on IBM’s latest Nighthawk r2 quantum processor, generating one million samples in just 19 seconds. BlueQubit’s researchers estimate that reproducing the same task on the Frontier supercomputer would take about 110 years.

The experiment was led by Tigran Sedrakyan, a theoretical condensed-matter physicist at BlueQubit, an American quantum computing software and cloud platform company.

Nighthawk r2 is a superconducting quantum processor with 120 qubits that can be accessed through IBM’s cloud platform. The team selected 61 of those qubits and ran random quantum circuits of gradually increasing complexity, executing up to 40 cycles. In the end, they found the balance point the experiment required at 36 cycles and 918 two-qubit gates.

In that configuration, Nighthawk r2 completed one million samples in only 19 seconds. The key point here is not that a quantum processor can quickly generate large numbers of 0s and 1s, but that those results must match the probability distribution produced by a particular quantum circuit — and that is exactly the part classical computers struggle to reproduce.

Random circuit sampling (RCS) is a benchmark task used in quantum computing to test the power of a quantum processor. Researchers let qubits evolve through several rounds of randomly chosen operations, producing entanglement, and then measure them repeatedly to obtain large numbers of outcome strings made up of 0s and 1s. As the number of qubits and the number of rounds increase, the difficulty for a classical computer of calculating the corresponding probability distribution rises rapidly.

To estimate how much computation a classical computer would need for the same task, the team used a tensor network contraction method. Their calculations showed that reproducing these one million samples would require roughly 1.2 × 10²⁷ computational operations.

The researchers then used the Frontier supercomputer as a reference point. Frontier has a peak performance of more than 10¹⁸ operations per second, and based on the sustained-performance estimate the team adopted, completing an equivalent task would take about 110 years. Of course, that figure of 110 years is an estimate derived from one specific classical simulation algorithm; it is not a theoretical upper bound on how long a classical computer would need.

To verify the results produced by the quantum processor, the researchers also used two cross-checking methods. The first was to build “clipped” quantum circuits that split the 61 qubits into three or four smaller systems, so that a classical computer could simulate parts of the circuit exactly and then assess the results of the full experiment.

The second method used mirror circuits: the quantum processor carries out a set of operations and then immediately performs the exact opposite operations. If the processor worked ideally, the final state should return to the initial state; the degree of deviation reflects the noise and errors accumulated during the experiment. As circuit depth increased, both methods still produced fairly similar results.

Under the experimental conditions of 36 cycles, the fidelity of the remaining quantum signal was about 0.23%. This figure cannot be read directly as an “accuracy rate” in the ordinary sense, because during random circuit sampling most of the quantum information gradually disappears under the influence of noise; what the experiment needs to confirm is whether a measurable trace of the ideal quantum probability distribution is still present.

The hardware design of Nighthawk r2 also helped shorten the experiment. Compared with IBM’s previous-generation Heron processor, it uses a dedicated reset mechanism that can actively release the energy in the qubits, cutting the waiting time between consecutive experimental runs to as little as about 1 microsecond.

IBM says Nighthawk r2 can execute more than 100,000 quantum circuits per second, compared with roughly 4,000 per second for Heron. That allows random circuit sampling experiments, which require a great many repetitions, to be completed faster.

The team did not claim, however, that quantum computers have made classical supercomputers obsolete. The “110 years” figure rests on a particular tensor network simulation method and its associated assumptions; more efficient algorithms, the reuse of intermediate results, and other approximation methods could all substantially reduce the classical computing time required.

A similar situation arose earlier with Google’s Sycamore experiment in 2019. At the time, researchers estimated that a classical supercomputer would need about 10,000 years to complete the corresponding task, but more efficient simulation methods subsequently emerged in the classical computing world and greatly shortened that time. Quantum advantage is therefore not a fixed measure; it keeps changing as both quantum hardware and classical algorithms advance.

The significance of this experiment lies in the fact that the quantum advantage test did not rely on specialised equipment found only in a laboratory, but was carried out on a commercial quantum processor that researchers can access through a cloud platform. The team also published the quantum circuits, samples and analysis code used in the experiment so that other researchers can reproduce and verify the work.

It is worth noting that random circuit sampling is not itself a computational task aimed at practical production applications, but a benchmark experiment for testing the capabilities of quantum computers.

The research currently appears as a preprint on arXiv and has not yet been peer-reviewed.