Ask most people what a quantum computer does and you will hear some version of "it tries every answer at once." It is a lovely sentence and it is wrong in a way that matters, because it leads straight to the belief that quantum machines will simply be faster computers that make laptops obsolete. They will not. With quantum computing explained properly, the picture is stranger and more interesting: a machine that is spectacularly good at a narrow set of problems, useless for most of what you do on a computer today, and still fighting serious engineering battles before it can deliver on its biggest promises.
Bits, qubits and why "every answer at once" is misleading
A classical bit is either 0 or 1. A qubit can be in a superposition – a combination of 0 and 1 described by two numbers called amplitudes. Add more qubits and the state describing them grows exponentially: describing 50 qubits fully would take a classical computer an enormous table of numbers.
So far, that sounds like "every answer at once." The catch is measurement. When you read a qubit, you get a single 0 or 1, with probabilities set by those amplitudes, and the superposition is gone. If you just put a computation into superposition and measured, you would get one random answer – no better than guessing.
The real trick is interference. Amplitudes behave like waves: they can reinforce or cancel each other. A quantum algorithm is a carefully choreographed sequence of operations designed so that the amplitudes of wrong answers cancel out and the amplitudes of right answers add up. That only works for problems with a particular mathematical structure, which is why quantum speed-ups exist for some tasks and not others.
The third ingredient is entanglement: qubits whose states are correlated in ways no classical system can mimic. Entanglement lets the machine represent relationships across many qubits at once, and it is what makes simulating a quantum computer on a classical one so hard.
How do quantum computers work physically?
Several competing technologies build qubits:
- Superconducting circuits, cooled in dilution refrigerators to a fraction of a degree above absolute zero (the approach used by IBM and Google).
- Trapped ions, individual charged atoms held in place by electromagnetic fields and manipulated with lasers.
- Neutral atoms arranged in arrays by optical tweezers.
- Photonic systems that encode information in particles of light.
Each has trade-offs in speed, how long qubits stay coherent, how easily they connect to each other, and how well they might scale. None has clearly won.
All of them share one enemy: noise. Qubits are fragile. Heat, vibration and stray electromagnetic fields cause errors within tiny fractions of a second. Today's machines are often described as NISQ devices – "noisy intermediate-scale quantum," a term coined by physicist John Preskill – meaning they have enough qubits to be interesting but too many errors to run long algorithms reliably.
The fix is quantum error correction: spreading one reliable "logical" qubit across many physical qubits and continuously detecting and correcting errors. The overhead is large – plausibly hundreds or thousands of physical qubits per logical qubit, depending on the hardware's error rates. Getting to machines with many good logical qubits is the central engineering challenge of the field, and progress is real but gradual.
What quantum computers can do (or are expected to)
These are the areas where there are strong theoretical reasons to expect an advantage:
- Simulating molecules and materials. Nature is quantum mechanical, and classical computers struggle to simulate quantum systems exactly. Chemistry problems such as catalyst design, battery materials and certain drug-discovery questions are the most widely cited long-term applications. This was Richard Feynman's original motivation for the idea in the early 1980s.
- Factoring and discrete logarithms. Shor's algorithm, published in 1994, can factor large numbers and solve discrete logarithms exponentially faster than the best known classical methods. That matters because RSA and elliptic-curve cryptography depend on those problems being hard.
- Unstructured search. Grover's algorithm gives a quadratic speed-up – searching N possibilities in roughly the square root of N steps. Useful, but far less dramatic than Shor, and the error-correction overhead may eat much of it in practice.
- Some optimisation and sampling problems, where research is active but clear, practical advantages over the best classical algorithms are still debated.
What quantum computing can't do
This list matters just as much:
- It will not speed up everyday computing. Web browsing, spreadsheets, video, databases and most business software have no quantum advantage. Quantum computers will be specialised accelerators, reached over the cloud, not replacements for your laptop.
- It does not solve every hard problem. There is no known quantum algorithm that efficiently solves NP-complete problems in general, such as the travelling-salesman problem at scale. "Quantum will fix logistics" is a hope, not a result.
- It does not handle big data well. Loading large classical datasets into quantum states is itself slow, which undercuts many proposed quantum machine-learning speed-ups.
- It is not ready for production workloads today. Demonstrations of "quantum advantage" so far have generally involved carefully chosen tasks without commercial value, and several have later been matched by improved classical methods.
Quantum computing and encryption: the part to act on now
Here is where quantum computing affects ordinary businesses before the machines are mature. A large, fault-tolerant quantum computer running Shor's algorithm would break the public-key cryptography that protects HTTPS, VPNs, software updates and digital signatures. Symmetric encryption such as AES is in much better shape; Grover's algorithm only weakens it, which longer keys address.
Nobody can say with certainty when such a machine will exist. The risk is "harvest now, decrypt later": an adversary records encrypted traffic today and decrypts it once the capability arrives. For data that must stay secret for a decade or more – health records, state secrets, long-term contracts – that is already a concern.
The response is post-quantum cryptography: new algorithms based on maths problems believed to be hard for both classical and quantum computers. In 2024 the US National Institute of Standards and Technology published its first post-quantum standards, including ML-KEM for key exchange and ML-DSA and SLH-DSA for signatures. Major browsers and operating systems have begun deploying hybrid key exchange that combines classical and post-quantum methods.
What to do:
- Inventory your cryptography. Know where you use RSA and elliptic curves – TLS, VPNs, code signing, document signing, stored data.
- Aim for crypto-agility. Systems that can swap algorithms through configuration will migrate far more cheaply than ones with algorithms hard-coded.
- Ask vendors about their post-quantum roadmap, especially for VPNs, HSMs and identity systems.
- Prioritise long-lived secrets for early migration.
Quantum computing explained in one paragraph
A quantum computer manipulates qubits so that wrong answers interfere away and right ones remain, which gives huge speed-ups for a narrow class of structured problems – molecular simulation and breaking today's public-key encryption chief among them – and little or nothing for everyday software. The hard part now is error correction, and timelines are uncertain. For most organisations, the practical takeaway is not to buy quantum hardware but to start the slow, unglamorous work of moving to post-quantum cryptography while there is still time to do it calmly.
