Quantum supremacy and other tall tales¶

Quantum computing has accumulated more hype than a Dibbler product launch, with a similarly questionable relationship to reality. Every few months another press release announces that quantum computers are about to revolutionise everything from drug discovery to financial modelling, conveniently timed with a funding round or a grant application. The actual capabilities receive less attention, presumably because “we ran a calculation on fifty-three noisy qubits for twenty microseconds before decoherence ruined everything” lacks the inspirational quality investors prefer.

The term quantum supremacy exemplifies the problem. It originally meant a quantum computer performing any task, however useless, faster than any classical one. Google demonstrated this in 2019 by sampling from a probability distribution nobody needed samples from, a genuine milestone reported as though quantum computers had suddenly become useful, which they had not. The gap between what these machines do and what the publicity claims resembles the gap between The Patrician’s actual policies and what people imagine them to be, and separating the technology from its publicity department is most of the work.

What they can do, might do, and will not do¶

Current machines can perform specific calculations on small problems under tightly controlled conditions: sample from certain distributions, simulate small quantum systems, and run variational algorithms that occasionally produce useful results. The most genuine present application is quantum simulation, modelling quantum systems that classical computers handle poorly, which earns its keep in chemistry and materials research. Everything else is demonstration, aspiration, or marketing.

Given error correction, thousands of logical qubits, and gate accuracy far beyond today’s, they might eventually break RSA through Shor’s algorithm, simulate molecules classical machines cannot, and improve some optimisation and machine-learning tasks at the margins. These are backed by real theory, on a timeline ranging from decades to never depending on who is asking and how recently they sought funding. The whole edifice rests on error correction working at scale, which is a reasonable assumption rather than a guaranteed one.

Some claims can be dismissed outright. Quantum computers will not replace classical ones for general computing, will not solve NP-complete problems in polynomial time, will not achieve consciousness, and will not violate thermodynamics. They compute, using quantum effects, within entirely ordinary physical constraints. Confusion about measurement and observers has produced a great deal of nonsense about quantum minds; it remains nonsense.

Spotting the snake oil¶

The field has attracted the usual opportunists, and distinguishing real work from elaborate nonsense rewards a few warning signs. Vague claims of quantum advantage without problem sizes, error rates, or comparison to the best classical methods are aspirational at best. Quantum solutions to problems classical machines already handle well add expense without benefit. Talk of “trying all possibilities simultaneously” or “exploring parallel universes” signals physics learned from pop-science articles. No named hardware, qubit count, or error rate usually means a classical simulation wearing a quantum label. Aggressive timelines, months rather than years, sell either honest consulting or dishonest capability. And “quantum” applied to business strategy or creative thinking is using the word as a synonym for impressive.

Unless a business is doing quantum research itself, or has a problem classical computers genuinely cannot touch, it needs ordinary machine learning, which works, scales, and requires no refrigeration. Curiosity is worth indulging; experimentation on cloud processors is cheap and educational. Confusing experimentation with deployment is where the money goes. The revolution may well come, gradually and for specific applications rather than suddenly and for everything, and in the meantime the honest posture is to build on what works now.

The clerk’s brief¶

December 2025: The bubble, in numbers¶

Over the year to December 2025 the quantum pure-play stocks ran up by hundreds of per cent, Rigetti above 300 and D-Wave above 500, on revenues in the low hundreds of millions or less, leaving price-to-sales ratios in the hundreds where thirty is already thought unsustainable, while insiders across the sector sold far more stock than they bought. The clerks note that a market paying several hundred times sales for a company with no stated path to profit is not valuing a technology but buying a story, and stories, unlike qubits, need no error correction.

September 2025: A billion for a promise¶

In September 2025 PsiQuantum raised one billion dollars at a seven-billion valuation, led by BlackRock and joined, among others, by Nvidia’s venture arm, on the strength of a promise to deliver a commercial machine by 2027, having delivered none so far. The clerks observe that money arriving well ahead of the machine is the defining feature of the file, and that a valuation doubling on a two-year promise is arithmetic the Guild of Alchemists would recognise.

March 2025: A supremacy claim, and a rebuttal by return of post¶

In March 2025 D-Wave announced in Science that its annealer had solved a materials-simulation problem it said would take a classical supercomputer nearly a million years, claiming the first quantum supremacy on a useful real-world task. Within weeks classical researchers at the Flatiron Institute reproduced much of it with tensor-network methods on ordinary hardware, and the two sides are still arguing over what was shown. The clerks file this beside Google’s 2019 claim: the supremacy announced today is often the classical benchmark broken tomorrow, so any beyond-classical result is best read as provisional until the classical side has had its turn.

January 2025: A market moved by a sentence¶

In January 2025 Nvidia’s chief executive remarked that genuinely useful quantum computers were perhaps fifteen to thirty years away, and the pure-plays shed around sixty per cent of their value in days; by March he had walked the comment back and the stocks recovered. Nothing about the machines changed in either direction, only the mood. The clerks note that an industry whose valuation swings sixty per cent on one man’s aside is not trading on fundamentals it can name.

The credibility ledger¶

The ledger’s standing pattern is that the claims and the capabilities keep separate books. Supremacy results arrive in respectable journals and are chased down by classical algorithms, Google in 2019 and D-Wave in March 2025; valuations arrive in the billions chased by little revenue, PsiQuantum in September 2025 and the pure-plays through December 2025; and the timeline is set less by physics than by whoever last spoke to an analyst, as in January 2025. None of this proves the technology worthless, any more than the dot-com bubble proved the internet worthless. It shows only that, for now, the price of the story and the state of the machine are two different measurements. The clerks’ standing advice is the section’s oldest: judge the field by what an independent party can reproduce, and pay for the qubits, not the adjectives.