Quantum supremacy: what Google, IBM and the qubit race actually promised

In April 2017, John Martinis gave himself a deadline. The head of Google's quantum research group told MIT Technology Review that by the end of that year his team would build a device that achieved "quantum supremacy," meaning it could run a particular calculation beyond the reach of any conventional computer.

"We think we're ready to do this experiment. It's something we can do now," Martinis told MIT Technology Review. The proof would come from a drag race between Google's chip and one of the world's largest supercomputers.

That was the pitch. What followed was two and a half years of fabrication, testing, leaked papers and public skepticism about what the milestone actually measured.

What a qubit is, and why nine became 49

Quantum chips store data in qubits, devices that can shortcut through some tough calculations by exploiting quantum mechanics. A conventional bit is either 0 or 1. A qubit can be both at once, in superposition, which lets a machine process many states simultaneously for certain problems.

That is the theory. The engineering is harder. In a 2014 piece, the EE Journal explained the University of California, Santa Barbara's approach: quantum operations amount to implementing rotations on groups of entangled qubits, and measuring the result collapses the superposition into an answer that may not be the right one. Reliability is not a side issue. The article states that uncorrected reliability has to exceed 99% before error correction can handle the rest. UCSB's qubits, called transmons, operated at roughly 30 millikelvin above absolute zero.

Google's chip family grew in public. In April 2017 the company had released results from a nine-qubit chip arranged in a line. Martinis said he needed a grid of 49 qubits for the supremacy experiment. The newest device at that point had only six qubits, but arranged two-by-three so the team could test whether its technology still worked when qubits sat side by side. It also tested a manufacturing method in which qubits and their control wiring are built on separate chips and later bump bonded together, an approach meant to remove the extra control lines that interfere with qubit function.

"That process is all working," Martinis told MIT Technology Review. "Now we're ready to kind of move fast."

Google was not alone. The MIT Technology Review piece lists Intel, Microsoft, IBM and startups among those racing to build quantum processors. Simon Gustavsson, an MIT quantum computing researcher, told the publication that Google and IBM were "pretty comparable." Chris Monroe, a University of Maryland professor and cofounder of the startup IonQ, was blunter about the value of the target: "It'll be an academic milestone. Afterward you still have to figure out how to make it more scalable and programmable."

IBM answers with 50 qubits

By November 2017, IBM claimed the scale record. Dario Gil, who led IBM's quantum computing and artificial intelligence research division, said on Friday 10 November that the company's scientists had successfully built and measured a processor prototype with 50 qubits, according to the Associated Press, which reported the announcement via The Seattle Times.

Gil said it was the first time any company had built a quantum computer at that scale. Seth Lloyd, an MIT mechanical engineering professor not involved in IBM's research, told the AP that IBM likely still had glitches to work out, but that the 50-qubit announcement was a sign of significant progress.

At a quantum computing for business conference at NASA Ames in December 2017, reported by Gizmodo, Martinis pushed back on qubit counting as a scoreboard. "Press releases always talk about quantum space race in number of qubits," he said. "It's more than just quantity, it's qubit quality."

Gizmodo reported that Google was fabricating its 49 or 50 qubit supremacy device that month and would begin testing two weeks later, Christmas vacation permitting. Martinis described what the experiment would involve: entangling qubits, letting the system evolve, and measuring outcomes whose probability distribution is so complex that a classical machine would have to simulate the quantum computer to check it. Such a simulation might take weeks to do what the quantum computer could do in minutes.

Quality meant qubits that do not produce the wrong value or decay into ordinary bits by interacting with the environment. Caltech theorist John Preskill, speaking at the same conference, called the current period the NISQ era, for Noisy, Intermediate-Scale quantum computer, and said progress toward a fault-tolerant machine had to continue.

The 200-second claim

The result landed in September 2019, though not the way Google planned. Fortune reported on 20 September that it had obtained a copy of Google's paper, which was posted to NASA.gov earlier that week before being taken down. The Financial Times first reported the news. A Google spokesperson declined to confirm the paper's authenticity, and NASA did not immediately respond to a request for comment, according to Fortune.

A source at Google suggested to Fortune that NASA had published the paper early, before the claims could be vetted through peer review, a process the source said could take weeks to months.

The paper's central numbers, as reported by Fortune: Google's processor, dubbed Sycamore, contained 53 qubits, scaled back from a 72-qubit device called Bristlecone that the team had previously designed. Sampling one instance of a specialized quantum circuit 1 million times took the processor about 200 seconds, while a state-of-the-art supercomputer would need approximately 10,000 years for the equivalent task, the researchers said. On a Google Cloud server, the same experiment would take an estimated 50 trillion hours; on the quantum processor it took 30 seconds.

"While our processor takes about 200 seconds to sample one instance of the quantum circuit 1 million times, a state-of-the-art supercomputer would require approximately 10,000 years to perform the equivalent task."

The researchers wrote that quantum processors based on superconducting qubits could now perform computations beyond the reach of the fastest classical supercomputers available today, and that the experiment marked the first computation that could only be performed on a quantum processor. They also predicted quantum computing power would grow at a double exponential rate, beating the roughly two-year doubling of Moore's Law.

The pushback

IBM's Dario Gil, by then head of IBM Research, told Fortune the result should not be used as a progress metric. "The experiment and the 'supremacy' term will be misunderstood by nearly all," he said, describing it as a highly special case laboratory experiment with no practical applications. "Quantum computers will never reign 'supreme' over classical computers, but will rather work in concert with them, since each have their unique strengths."

Jim Clarke, Intel Labs' director of quantum hardware, called the update "a notable mile marker" and said a commercially viable quantum computer would require many more research and development advances. "While development is still at mile one of this marathon, we strongly believe in the potential of this technology," he said.

IBM had its own way of keeping score. In March 2019, ZDNET reported, IBM outlined a Quantum Volume of 16 for its Q System One, which has a 20-qubit processor, double the Quantum Volume of 8 for its previous IBM Q system. Quantum Volume combines qubit count, connectivity and coherence time while accounting for gate and measurement errors, device cross talk and compiler efficiency. IBM said Quantum Volume would need to double every year to reach Quantum Advantage, the point where quantum applications deliver significant advantages over classical computers, within the next decade.

That framing, quality over raw count, is the one that survived the 2019 headlines. Google's paper framed the result as quantum computing transitioning from a research topic to a technology that opens new computational capabilities, adding that the field was "only one creative algorithm away from valuable near-term applications."

Read together, the 2017 and 2019 records describe a milestone that was real, narrow and heavily contested. A 53-qubit chip sampled a circuit in 200 seconds that a supercomputer would need millennia to reproduce. It did not run a useful program, and the people building competing machines said so first.

The pattern is likely to repeat. Google set the deadline in April 2017, IBM claimed 50 qubits in November 2017, and the supremacy paper surfaced in September 2019. Each announcement reset the argument about which number means progress.