What is a quantum computer? Counted in quantum bits and errors

That far, everyone agrees on what a quantum computer is. Then the texts part ways, and most of them are about the future, about what the machines will solve and when.

This article does the opposite. It explains what a quantum bit is and why it differs from an ordinary bit, and then counts only what has been measured and published. That goes further than you might think.

Bit and quantum bit

A bit is a switch with two positions, zero or one. An ordinary computer is billions of such switches being turned off and on in a set order.⁶ Each switch is in one position at a time, and you can always check which.

A quantum bit follows other rules. Until it is measured, it can be in a superposition of the two positions, a weighted combination of zero and one.⁶ Only when it is measured does it fall to zero or one. The popular phrase that a quantum bit "is zero and one at the same time" is a simplification of this.⁶ The quantum bit can be in a superposition until it is read, and the superposition can be calculated with.

The second property is entanglement. Two quantum bits can be linked so that they behave as a single unit even when they are apart: what happens to one decides what happens to the other.⁵ This is the experiment that won the 2022 Nobel Prize in Physics, described further down, so it is not a metaphor.⁵

Together, the two properties mean that a quantum computer can keep many possibilities open at once and let them affect each other before the answer is read.⁶ For some kinds of problem, that is a completely different route to the answer than trying one possibility at a time.

Why it is not a faster ordinary computer

The most common misunderstanding is that a quantum computer is an ordinary computer that calculates faster. It is not. It calculates in a different way, and that way suits some problems and is useless for others. Writing an email, showing a film or adding up a column are tasks a quantum computer does not do better, and in practice does not do at all.

That also explains why the comparisons in the headlines look the way they do. When it says a calculation "would take a supercomputer X years", that is a result from one particular problem, chosen because the quantum machine is good at exactly that. Such figures say something about that problem and nothing about general performance. The rule for anyone reading them is simple: ask who measured, which problem was measured, and whether the problem was chosen for the purpose.⁶ If any of the three answers is missing, the figure is worth no more than a headline. This article uses no such figures.

Sweden's quantum computer, counted

Sweden is building its own quantum computer. The project has been running at Chalmers University of Technology since 2018, within the Wallenberg Centre for Quantum Technology, WACQT, a twelve-year programme funded mostly by the Knut and Alice Wallenberg Foundation, with four main areas: quantum computers, quantum simulators, quantum communication and quantum sensors.¹

According to the university's own account in January 2023, when the testbed was presented, the machine had reached 25 quantum bits.^{1, 7} The project's goal for 2029 is 100 quantum bits.^{1, 7} That is a goal the project has set for itself, and it is given here as a goal, not as a statement about what will happen. The number of quantum bits is also the figure in this article that ages fastest; it is checked against the university's own account shortly before publication.

Quantum computers that can be reached over the internet already exist in several places. What the university itself points to as the difference is that the testbed will be cheaper for Swedish users and that it shows openly what is inside the machine, so that the algorithm can be adapted to the hardware.¹ The reason a copy is being built is that the original machine is very rarely free, because the researchers keep rebuilding it.¹ The Knut and Alice Wallenberg Foundation has, in a first stage, granted the testbed 102 million kronor.¹ Besides the computer, the testbed includes a support unit and test equipment for developing components.^{1, 8} The plan in January 2023 was for the test equipment to be in place during 2024 and for the quantum computer for running algorithms to open in 2025 with 25 quantum bits, with an upgrade to 40 after a couple of years.^{1, 8}

One more thing from the university's description deserves a mention, because it says something about what a quantum computer is physically. The machine is cooled by a unit in several stages, and the university shows two of its researchers opening the unit's final stage.¹

The error is the whole problem

Why quantum bits make errors

The Swedish machine's quantum bits are counted in tens.^{1, 7} The reason is that they make errors. A quantum bit loses its superposition after a short while, and every operation on it risks disturbing it.⁶ That is not a teething problem waiting for better components; it is a property of what quantum bits are.

The solution, proposed as early as 1995, is called quantum error correction: many physical quantum bits are combined into a single logical quantum bit, and the errors of the logical one fall as more physical ones are added.^{2, 9} But it only works if the physical quantum bits already make few enough errors, below a critical threshold.² For almost thirty years nobody had shown a machine that was below the threshold.^{2, 3}

What the study in Nature showed

On 9 December 2024 a research team published a result in the journal Nature that did so.² On a superconducting processor with 105 quantum bits, they built one logical quantum bit from 101 physical ones.² Each time they increased the size of the code by two steps, the logical error rate fell by a factor of 2.14, with an uncertainty of 0.02.² The largest code gave 0.143% errors per error correction cycle, with an uncertainty of 0.003.²

The most important figure is perhaps the simplest. The logical quantum bit held its state for 291 microseconds, with an uncertainty of 6.² The best physical quantum bit on the same processor held for 119, and the median was 85.² So the combined quantum bit was better than its best component, by a factor of 2.4 with an uncertainty of 0.3.² It is the first time this has been shown for a code of this kind.^{2, 3} What people have been waiting for for almost thirty years is the second thing the same study shows: that a machine is below the threshold.^{2, 3}

How far there is still to go

The rest of the study is about how far there is still to go, and the researchers say so themselves. The decoder that works out which errors have occurred ran in real time with an average delay of 63 microseconds in the smaller of the study's two codes, against a cycle time of 1.1 microseconds.² In the study's repetition codes, run on a smaller processor, rare errors that hit many quantum bits at once occurred about once an hour, that is, once every three billion cycles.² According to the authors, their cause is not yet understood. The study states that the performance, if scaled up, could meet the requirements for large-scale error-corrected algorithms.² The condition "if scaled up" is the authors' own. The research group's own commentary on the study also says that at current error levels more than a thousand physical quantum bits per code grid may be needed to reach a logical error rate of one in a million, and that the engineering challenge ahead is enormous.³

Put the two figures side by side. Chalmers reported 25 quantum bits in January 2023, and a single error-corrected quantum bit of that kind required 101 physical ones in the study from December 2024.^{1, 2} That is the distance between what exists and what is written about.

What the 2022 Nobel Prize rewarded

Everything above rests on entanglement being real and not a mathematical convenience. That was settled in experiments that won the 2022 Nobel Prize in Physics, awarded to Alain Aspect, John F. Clauser and Anton Zeilinger "for experiments with entangled photons, establishing the violation of Bell inequalities and pioneering quantum information science".⁵ The prize was announced by the Royal Swedish Academy of Sciences on 4 October 2022 and was shared equally between the three.⁵

The background is a question the founders of quantum mechanics argued about: can the strange connections be explained by the particles carrying hidden information from the start, so-called hidden variables?¹³ In the 1960s John Stewart Bell showed that the question can be settled by measurement.^{5, 13} If there are local hidden variables, the correlation between a large number of measurement results can never exceed a certain value.^{12, 13} Clauser turned the idea into an experiment, and the measurements clearly broke Bell's limit.^{5, 12} Aspect developed the set-up and closed an important loophole, by switching the measurement settings after the pair had left its source.^{5, 12} Zeilinger's group demonstrated quantum teleportation, which moves a quantum state from one particle to another at a distance.⁵

The conclusion is that quantum mechanics cannot be replaced by a theory with local hidden variables.^{12, 13} Entanglement is real. The Academy stresses that there is now a large field of research covering quantum computers, quantum networks and secure quantum-encrypted communication, among other things.⁵ That is the foundation of the field, established in experiments, and that is as far as this article follows it.

What a quantum computer does not do today

It does not break today's encryption.⁶ What has happened is more concrete. The standards body NIST has published new encryption standards that it describes as designed to withstand attacks from quantum computers.^{4, 10} That a standard has been developed with that starting point is a fact that has happened. When or whether a quantum computer breaks today's methods is not, and it is not stated here.

Nor does it solve any practical problem for you today. What has been measured and published is a logical quantum bit that lasts longer than its parts, and a Swedish machine with 25 quantum bits that companies will be able to try.^{1, 2} The result on the logical quantum bit was published on 9 December 2024. The figure of 25 quantum bits for the Swedish machine comes from Chalmers' account of 23 January 2023.¹ The two describe different machines.

What to take away

A quantum computer calculates in a different way from an ordinary computer, not faster at everything.⁶ Chalmers' machine had 25 quantum bits in January 2023, and in the study from December 2024 it took 101 physical quantum bits to make a single logical one, which held for 291 microseconds against 119 for the best physical one.^{1, 2} What makes this work that spans decades is that error correction is the problem itself. The researchers write that more than a thousand physical quantum bits per logical one may be needed to reach an error level they themselves call modest.^{2, 3} Follow Chalmers' own account of the number of quantum bits against the goal of 100 by 2029, the opening of the testbed to companies, and the NIST standards, where a fifth algorithm was chosen in March 2025.^{1, 7, 11}

Education, not advice.