The basics

8 terms
TermWhat it means
QubitThe basic unit of a quantum computer. An ordinary bit is either 0 or 1. A qubit can hold a blend of both until it is measured, which lets quantum machines work through certain problems in ways ordinary computers cannot.
SuperpositionThe blended 0-and-1 state a qubit can be in before it is measured. It is the property that gives quantum computers their potential, and also what makes them fragile.
EntanglementA link between qubits so that measuring one tells you something about the others, however far apart they are. Quantum algorithms depend on it.
MeasurementReading out a qubit's value. Measuring forces the qubit to settle on 0 or 1, so a calculation has to be designed so the right answer is the likely one.
Quantum gateA single operation on one or two qubits, the quantum version of a logic step in an ordinary chip. Programs are built from thousands or millions of them.
Quantum circuitA sequence of quantum gates that together carry out a calculation. A circuit's depth (how many steps it has) is limited by how long qubits stay stable.
Quantum algorithmA method for solving a problem on a quantum computer. Only some problems have quantum algorithms that are known to beat the best ordinary methods.
QPUQuantum processing unit: the quantum chip or system itself, often used alongside ordinary CPUs and GPUs.

Measuring quality

14 terms
TermWhat it means
Physical qubitOne actual qubit on the hardware. Machines are often advertised by this count, but on its own it says little about how useful a machine is.
Logical qubitSeveral physical qubits working together as one protected qubit, so errors can be caught and fixed. Useful large-scale machines will need hundreds or thousands of these. Always check whether a headline number is physical or logical.
FidelityHow often an operation goes right. A two-qubit gate fidelity of 99.9% means about one mistake in a thousand operations. It is the most comparable quality measure across companies.
Error rateThe flip side of fidelity: how often an operation or a stored qubit goes wrong. Long calculations need very low error rates.
NoiseDisturbances from heat, vibration, stray signals and imperfect control that knock qubits off course. Noise is the main reason today's machines make mistakes.
Coherence timeHow long a qubit keeps its quantum state before noise scrambles it. Longer coherence leaves more time to run a calculation.
DecoherenceThe loss of a qubit's quantum state to noise. Once it happens, the information the qubit held is gone.
Error correctionSpreading information across many physical qubits so mistakes can be found and fixed while a calculation runs. It is the main route to reliable machines, and it costs a lot of extra qubits.
Error detectionA lighter version of error correction that spots mistakes but does not fix them; the affected run is thrown away. Some logical qubit claims are error-detected, not error-corrected.
Surface codeThe most widely studied error correction scheme, which arranges physical qubits in a grid. Other codes need fewer qubits but place more demands on the hardware.
Error mitigationSoftware techniques that reduce the effect of errors after the fact, by running a calculation many times and adjusting the results. It helps today's machines but does not replace error correction.
Fault-tolerantA machine that can run long calculations reliably even though its parts make errors. It is the industry's main goal; no company has reached it at a useful scale yet.
BenchmarkA standard test used to compare machines. Companies often choose different benchmarks, so results are hard to compare directly.
Quantum volumeA single-number benchmark introduced by IBM that combines qubit count and error rates. Several companies report it; others prefer their own measures.

Types of machines

9 terms
TermWhat it means
Superconducting qubitsQubits made from tiny electrical circuits cooled to near absolute zero. Used by Google, IBM, Rigetti and others. Operations are fast, but the qubits are sensitive to noise.
Trapped-ion qubitsQubits made from individual charged atoms held in place by electric fields and controlled with lasers. Used by Quantinuum and IonQ. Operations are slower, but accuracy tends to be high.
Neutral-atom qubitsQubits made from uncharged atoms held in place by focused laser beams. Used by QuEra, Pasqal, Atom Computing and Infleqtion. Large arrays are relatively easy to build.
Photonic qubitsQubits carried by particles of light. Used by PsiQuantum, Xanadu and others. They work at or near room temperature for much of the system, but losing photons is a constant problem.
Spin qubitsQubits stored in the spin of single electrons in silicon chips, similar to ordinary chipmaking. Pursued by Intel, Diraq and others. Still at an earlier stage than the leading approaches.
Topological qubitsA proposed kind of qubit designed to resist errors by its physics. Microsoft is pursuing this approach and says it has built early devices; some physicists question the evidence.
Quantum annealerA specialized machine built for optimization problems, made by D-Wave. It is not a general-purpose quantum computer, so its qubit counts are not comparable with other machines.
Dilution refrigeratorThe chandelier-shaped cooler that keeps superconducting and spin qubits a fraction of a degree above absolute zero. Its size and cost are part of the challenge of scaling up.
ModalityIndustry shorthand for the type of qubit a company builds (superconducting, trapped-ion and so on). No modality has clearly won.

What they are for

6 terms
TermWhat it means
Quantum advantageA quantum computer doing a useful task better, faster or cheaper than any ordinary computer can. Claims are frequent and often disputed, because ordinary methods keep improving.
Quantum supremacyAn older term for a quantum computer beating ordinary computers at any task, even a useless one. Google's 2019 claim popularized it; many now prefer "quantum advantage".
NISQNoisy intermediate-scale quantum: today's era of machines with tens to thousands of error-prone qubits and no full error correction.
Quantum simulationUsing a quantum computer to model molecules and materials, which follow quantum rules themselves. Widely seen as the most promising early use, in chemistry, drug research and batteries.
OptimizationFinding the best option among a huge number of choices, such as delivery routes or financial portfolios. A common pitch for quantum computers, though proven speedups are rare.
Hybrid computingSplitting a job between a quantum processor and ordinary computers, each doing the part it is best at. Most practical work today is hybrid.

Security and encryption

8 terms
TermWhat it means
Shor's algorithmA 1994 method that would let a large, fault-tolerant quantum computer break the public-key encryption that protects most internet traffic. No existing machine is close to running it at that scale.
Grover's algorithmA quantum search method that speeds up brute-force attacks on some encryption. Using longer keys largely counters it.
Q-DayThe day a quantum computer can break today's widely used encryption. Estimates vary widely and no one knows the date.
Post-quantum cryptographyNew encryption methods designed to resist quantum computers while running on ordinary hardware. NIST published the first standards in 2024, and governments have set deadlines to switch.
Harvest now, decrypt laterStealing encrypted data today in the hope of unlocking it once quantum computers are powerful enough. It is why long-lived secrets need protecting now.
Quantum key distributionSharing encryption keys using quantum physics so that any eavesdropping can be detected. It needs special hardware and is a different approach from post-quantum cryptography.
Crypto-agilityBuilding systems so their encryption can be swapped out easily. It matters because post-quantum standards may change as they are tested.
"Quantum ready"A marketing label, usually meaning a product supports post-quantum encryption or can be upgraded to it. There is no standard definition, so check what the vendor actually means.

Reading the news

4 terms
TermWhat it means
PreprintA research paper posted publicly before peer review, often on arXiv. The results may change or fail to hold up.
Peer reviewChecking of a paper by independent experts before a journal publishes it. It catches many problems but does not guarantee a result is right.
RoadmapA company's published plan for future machines and dates. Roadmaps are targets, not results, and dates often slip.
Company claimA figure or result reported by the company that made the machine and not yet checked independently. The Digest labels these so readers can weigh them.