Google's Willow chip: a step towards quantum machines that matter

In early December 2024, Google announced a quantum processor named Willow, and the announcement was easy to misread. One part of it was genuinely remarkable: a demonstration that a superconducting qubit array could push its error rate below a threshold that scales favourably as the system grows, the first time a company had shown the crossover point in a widely reviewed result. The other part, the claim of a benchmark computation ‘faster than a modern supercomputer’, required a far more careful gloss, and the coverage that followed frequently failed to provide it.

The meaningful achievement comes down to error correction. Quantum computers gain their power from qubits, which can be entangled in states no classical bit can hold, but qubits are fragile and make errors. For years the field has chased the idea of ‘below-threshold’ performance: a regime in which adding more physical qubits reliably lowers the logical error rate, the point at which you can trust the machine to keep getting better as you build it larger. Google’s researchers reported exactly that crossover on a 105-qubit chip, and the result, published in Nature, was the headline any specialist would have given the story.

The second claim was the one that travelled further than it should have. Google measured Willow against a classical algorithm on a task involving random circuit sampling, and stated that the computation would take a classical supercomputer a near-impossible length of time. The claim was not dishonest, but it was not the breakthrough the phrase suggested. Random circuit sampling is a contrived benchmark, designed to be hard for classical machines, and the comparison depended on specific assumptions about the classical baseline. It is a demonstration of scale, not a solution to a problem anyone is waiting for.

The honest reading is more interesting than the hype. Real error-corrected quantum computing is not five years off because one chip passed a threshold; it is a sequence of such thresholds, each requiring more qubits, more control and more patience. Willow is a genuine step on that sequence, which is different from saying the machine is useful today. What it does is give the field something it badly needed: evidence that the central engineering challenge is a hard but solvable one, rather than a mystery.

There is a public-policy thread running underneath. Quantum computing is treated as a strategic race, and a company announcing a below-threshold chip tends to be read as a country gaining ground. That framing is not entirely wrong, but it flattens the most important fact: the field advances in public, through peer-reviewed results, precisely because the hardest problems are shared ones. A win in that arena is a win for everyone who cares about the technology, not a trophy for a single lab to hoard.

The measurement that matters is the one that will arrive in the next few years: a quantum machine that solves a genuinely useful problem, such as simulating a molecule that classical supercomputers cannot handle, at a scale that matters. Until that day, every chip like Willow is best read as progress on the path, not proof of arrival. The breakthrough was real, the patience is the price, and the race will be decided by the people who keep that distinction straight.