What counts as a quantum computing milestone, from 9 qubits to 53

The word "milestone" gets used loosely in quantum computing. To understand what any given announcement actually demonstrates, it helps to separate three things: how many qubits a device has, how good those qubits are, and what problem the machine was asked to solve.

Google's own researchers drew that distinction early. In April 2017, John Martinis, who leads Google's quantum hardware group, told MIT Technology Review his team had a few months to reach a milestone he called quantum supremacy: a calculation beyond the reach of any conventional computer. At the time, Google had built a six-qubit chip arranged in a two-by-three grid. It was a test of whether qubits still worked when placed side by side, and of a manufacturing method that bump bonds the qubits to separate control wiring. Martinis said the team needed a grid of 49 qubits for the actual supremacy experiment. Designs for devices with 30 to 50 qubits, he added, were already in progress.

IBM answered that November. Dario Gil, who led IBM's quantum computing and artificial intelligence research division, said the company had built and measured a 50-qubit processor prototype. According to the Associated Press, he described it as the first time any company had built a quantum computer at that scale. Seth Lloyd, an MIT mechanical engineering professor not involved in the work, told the AP that glitches likely remained, but he called the announcement a sign of significant progress.

Qubit count is not the whole story

By December 2017, Martinis was pushing back on how the race was being reported. Speaking at the Q2B conference at NASA Ames, covered by Gizmodo, he said press releases always talk about the quantum space race in number of qubits, and that qubit quality mattered as much as quantity. Google was then fabricating its 49 or 50 qubit supremacy device, he said, with testing to begin within two weeks. The conference drew representatives from Volkswagen, Airbus, Citibank, Emerson and Atos, alongside venture capital firms, all listening to talks about optimization and machine learning. Caltech theorist John Preskill used the event to describe the current period as the NISQ era, for Noisy, Intermediate-Scale quantum computers.

IBM later argued the field needed a better scoreboard than raw qubit counts. At the American Physical Society March Meeting in 2019, ZDNET reported, IBM said its Q System One, a 20-qubit processor, had reached a Quantum Volume of 16. That was double the Quantum Volume of 8 recorded by its previous IBM Q system. Quantum Volume is a composite metric covering qubit count, connectivity and coherence time, while also accounting for gate and measurement errors, device cross talk and compiler efficiency. IBM said the figure would need to double every year to reach Quantum Advantage within the next decade. That is the point where quantum applications deliver significant advantages over classical computers.

Then came the supremacy claim itself. Fortune obtained a copy of a Google paper that had been posted to NASA.gov before being taken down. On 20 September 2019, the magazine reported that the company said its 53-qubit Sycamore processor had performed a sampling task in about 200 seconds that would take a state-of-the-art supercomputer approximately 10,000 years. A Google source told Fortune that NASA had published the paper early, before peer review. Google declined to confirm the paper's authenticity at the time.

Not everyone accepted the framing. Dario Gil, by then head of IBM Research, told Fortune the experiment was a highly special case laboratory experiment with no practical applications. Quantum computers, he said, would work in concert with classical machines rather than reign supreme over them. Jim Clarke, Intel Labs' director of quantum hardware, called it a notable mile marker and said a commercially viable quantum computer would require many more research and development advances.

Nature published the work in October 2019. Google's own blog post described the 200-second result as the "hello world" moment the field had been waiting for, and noted it took 13 years to get there, starting with Hartmut Neven's 2006 exploration of quantum computing for machine learning. The same post was explicit about the limits: it would be many years before a broader set of real-world applications could be implemented. Martinis had made a similar point in 2017, saying the experiment would be an academic milestone and that quantum processors would need to be far larger than 50 qubits to do useful work.