Quantum Supremacy Near?

Few phrases in computing are as easily misunderstood as “quantum supremacy”. It does not mean that quantum computers have replaced ordinary ones, or that they can already solve everyday problems. It means that a quantum machine has performed one specific task that a classical computer could not match in a reasonable time. As of 1 February 2026 that milestone has been claimed, challenged and partly repeated, but a large, reliable quantum computer that does useful work has not yet been built. This article explains what has been shown, what has not, and why error correction is the hinge on which everything else turns.

What “supremacy” does and does not mean

Physicist John Preskill discussed quantum supremacy in 2012, as the point where a quantum device can do something classical computers cannot do efficiently.^[1] In 2018 he introduced another term that matters here: NISQ, for “noisy intermediate-scale quantum” devices. He suggested that machines with 50 to 100 qubits might outperform classical computers on some tasks, but that such a machine “will not change the world right away”.^[2]

In 2019 a Google team reported the first claimed demonstration. Its Sycamore processor prepared states on 53 qubits and took about 200 seconds to sample a random circuit a million times, a task the authors estimated would take a state-of-the-art classical supercomputer about 10,000 years.^[3] IBM researchers argued at the time that, using disk storage as well as memory, a classical supercomputer could simulate such circuits “in a matter of days”.^[4] A supremacy claim is about one benchmark, and it can narrow when classical methods improve. The 2019 task was chosen to be hard for classical machines, not to be useful.

Different, not simply faster

The idea behind quantum computing is older than the hardware. Richard Feynman gave a lecture at MIT in May 1981, published in 1982, arguing that efficiently simulating large, highly entangled quantum systems would need a computer built from quantum mechanical elements.^[5] Qubits can be in superposition, and groups of them can be entangled, which lets some algorithms work in ways ordinary bits cannot. That does not make a quantum computer a faster laptop: speed-ups are expected only for particular kinds of problem.

The error-correction challenge

The main obstacle is noise. Qubits lose their state through interference, so errors build up quickly. Error correction spreads one reliable “logical” qubit across many physical qubits and keeps checking them. The catch is that correction only helps once physical errors are below a threshold; below it, adding more qubits makes the logical qubit better rather than worse.

In a preprint submitted in August 2024, Google Quantum AI reported surface-code memories operating below that threshold. A 101-qubit code reached an error rate of 0.143% per correction cycle, and the logical memory lasted 2.4 times longer than its best physical qubit. The authors also report rare correlated errors about once an hour, and say the results could meet the needs of large fault-tolerant algorithms only “if scaled”.^[6] IBM researchers have taken a different route. A paper by Sergey Bravyi and colleagues describes a code family that, in the authors’ analysis, could keep 12 logical qubits alive using 288 physical qubits, where the surface code would need nearly 3,000, assuming a 0.1% physical error rate.^[7] These are steps, not a finished computer.

In October 2025 Google reported a “verifiable” speed-up on its 105-qubit Willow chip: an algorithm it calls Quantum Echoes, which it says ran 13,000 times faster than the best classical algorithm on a leading supercomputer. Google itself calls the result a step towards applications, not an application.^[8]

Drug discovery: promise, not proof

Simulating molecules is the application most often linked to quantum computers, because molecules are quantum systems, which fits Feynman’s original idea. But the sources do not support claims that such problems “could be solved in seconds”. In the same Google work, a proof-of-principle experiment with UC Berkeley applied the algorithm to two molecules, of 15 and 28 atoms. Google says the results matched conventional nuclear magnetic resonance (NMR) and revealed extra information, and that such methods “could become” useful in drug discovery and materials science.^[8] That is a possibility, not a demonstrated result. We found no reliable source showing that a quantum computer has yet sped up real drug discovery, and none for earlier claims about climate modelling, so this article does not make them.

Encryption: a planning problem, not an emergency

Quantum computers matter for security because a large enough machine could break the public-key encryption used to protect websites, email and payments. In 2025 Craig Gidney estimated that a 2048-bit RSA key could be factored in under a week by a quantum computer with fewer than a million noisy qubits, down from an earlier estimate of 20 million. His assumptions include a 0.1% gate error rate and a one-microsecond error-correction cycle.^[9] Google’s Willow chip has 105 qubits, so the gap to that estimate remains very large.

Governments are not waiting. In August 2024 the US National Institute of Standards and Technology (NIST) published its first three finalised post-quantum encryption standards, designed to resist both conventional and quantum attack. NIST says some experts predict a code-breaking quantum computer “could appear within a decade”, and urges organisations to begin moving to the new standards now, because full integration takes time.^[10]

So how close are we?

Closer than a few years ago, but not there. Experiments have shown that quantum hardware can do something classical machines find hard, and that scaling up can reduce errors. What has not been shown is a machine with enough reliable logical qubits to run a useful algorithm beyond the reach of classical computers.

When you read the next headline, ask what was measured, and whether it was compared with the best classical method. Our Technology Comparisons and Reflection page offers a set of questions for judging claims like these, and Emerging Technologies explains why new technologies carry so much uncertainty.