Quantum advantage

Few phrases in technology are as abused as quantum supremacy. It has a precise technical meaning, a far looser popular one, and a track record of claims that did not survive contact with better classical algorithms. Understanding the distinction is the single most useful defence against overstated quantum headlines.

The two terms

Quantum supremacy means a quantum computer performing some task faster than any classical computer could, where the task does not have to be useful. It is a physics demonstration: proof that the machine is doing something a conventional computer genuinely cannot follow.

Quantum advantage means doing something people actually want done, better, faster or cheaper than the classical alternative. Simulating a molecule for drug discovery, optimising a logistics network, pricing a portfolio. This is the commercial milestone, and it is the one that has not convincingly arrived.

The distinction is not pedantry. It is the difference between a laboratory result and a product.

Why the claims keep collapsing

Google announced supremacy in 2019 with its Sycamore processor, reporting that a sampling task taking its machine about 200 seconds would take the best supercomputer around 10,000 years. Within weeks IBM argued the classical estimate was wrong, and over the following years improved classical simulation methods narrowed the gap dramatically on that class of problem.

This has become the pattern. A quantum result is announced, classical algorithm researchers study the specific task, and a cleverer conventional method closes much of the distance. The lesson is not that the quantum results are fraudulent; it is that “no classical computer can do this” is a claim about every possible classical algorithm, including ones nobody has invented yet, and such claims are extraordinarily hard to make stick.

What a credible claim looks like

Three questions cut through most announcements. First, is the task useful, or was it chosen because quantum hardware happens to be good at it? Sampling from random circuits is the classic example of the latter. Second, what is the honest classical baseline, run by people motivated to beat it rather than by the same team? Third, does the advantage grow with problem size, or does it vanish as soon as the problem is scaled to a realistic case?

Where it stands

Supremacy-style demonstrations have been reported by several groups on several hardware platforms, and the physics is no longer seriously disputed. On 30 July 2026 IBM and its partners published three validated advantage claims: error-mitigated quantum dynamics with Qedma on up to 74 qubits, a simulation of heterogeneous quantum systems with Algorithmiq that has gone unmatched for eight months, and logical circuits verified with the University of Chicago. The results arrived five months ahead of IBM’s end-2026 target and are open to classical challenge on IBM’s public tracker, which, given how earlier claims have fared, is where the verdict will be decided. A week on, the claims stood: as of 6 August 2026 no rebuttal, refutation or replication had appeared, and Algorithmiq had released its classical simulation tool and invited challengers to use it. When a company claims advantage, the first thing to check is which of the two words they are using, and whether they have swapped one for the other without saying so.