How Quantum Computers Work

Picture a bit as a coin lying flat on a table. Heads or tails, 0 or 1. You can look at it whenever you like and the answer is always definite. Every photo, message and video game is billions of these coins, flipped very fast.

Now picture the same coin spinning on its edge. While it spins it is not heads and it is not tails — it is in a state that will become one of them when it stops. That spinning state is the qubit's superpositionA qubit's state before it is measured: a lean toward 0 or 1 that has not yet resolved into either.. Crucially, the spin is not 50/50 by default: you can bias it so that when it lands it is, say, 80% likely to be heads. The qubit stores that bias, not an answer.

The catch is reading it. The moment you look — physicists call this measurementThe act of reading a qubit, which forces it to settle into a definite 0 or 1. — the spin stops and you get a plain heads or tails, and the bias is gone. So a single qubit does not hand you more information than a bit. The power comes from what you can do to many of them before you look.

The sphere is called the Bloch sphereThe standard picture of one qubit's state: an arrow inside a sphere, with 'definitely 0' at the north pole and 'definitely 1' at the south.. The north pole means 'definitely 0'. The south pole means 'definitely 1'. Every other direction the arrow can point is a superposition — a different mix of the two. An arrow pointing at the equator is a perfectly even 50/50 lean; an arrow tilted toward the north pole leans toward 0.

This picture does two things for you. First, it makes 'superposition' concrete: it is not a mystery, it is just an arrow that is not pointing at a pole. Second, it shows what a quantum gateA basic operation on qubits — geometrically, a rotation of the state arrow. Quantum programs are built from gates the way ordinary programs are built from instructions. — the basic operation in a quantum program — actually does: it rotates the arrow. Every quantum algorithm is a choreography of rotations.

Measurement is the one operation that is not a rotation. It snaps the arrow to a pole, north or south, with odds set by how close it already was. Then the arrow stays there. That is why you cannot peek mid-calculation: peeking is measuring, and measuring destroys the superposition you were computing with.

Entangle two qubits and their arrows stop having separate directions. There is one shared state for the pair, and the only thing you can say about each qubit alone is 'random'. But the pair is perfectly correlated: if one reads 0, the other is guaranteed to read 1 (or, for a different entangled state, guaranteed to match).

The temptation is to think the answer was secretly decided in advance, like two sealed envelopes. Experiments over fifty years, recognised with the 2022 Nobel Prize in Physics, ruled that out: the correlations are stronger than any pre-agreed plan could produce. The qubits genuinely did not have individual values until one was measured.

Two things entanglement does not do: it does not send a message (the person holding the second qubit sees a random result until you tell them what you got), and it does not let you copy information. What it does do is let a quantum computer treat many qubits as one enormous shared state — which is where the exponential room to compute comes from.

A quantum program starts by setting every qubit to a known state, usually all 0. Then gates rotate and entangle them until the machine holds a superposition over every possible answer at once. So far this sounds like 'try everything' — and if you measured now, you would get one random answer, no better than guessing.

The stage that makes it a computer is interference. Each possible answer carries not just a probability but a phaseWhere in its wave cycle a quantum path sits — its crest or trough. Two paths with opposite phases cancel when they meet. — think of it as the crest or trough of a wave. The gates are arranged so that paths leading to wrong answers arrive with opposite phases and cancel out, like two ripples meeting, while paths leading to the right answer arrive in step and reinforce. Designing that arrangement is what a quantum algorithm is.

Only then do you measure. Because the wrong answers have (mostly) cancelled, the reading is very likely to be a right one. Run it a few times to be sure, and hand the result to an ordinary computer to display. Notice what this means: a quantum computer only speeds up problems where someone has found a way to make the wrong answers cancel. For most everyday tasks, no such trick exists — which is why it will never replace your laptop.

In 2019 Google's Sycamore chip had 53 qubits and performed a carefully chosen task faster than the best classical supercomputer of the day could simulate — the first widely accepted demonstration that quantum hardware could do something classical machines struggle with. By the end of 2023 IBM had built Condor with 1,121 qubits. On the chart that looks like a rocket.

The complication is noiseUnwanted disturbance — heat, vibration, stray fields — that nudges a qubit's state and introduces errors.. A qubit's superposition is fragile: heat, vibration and stray fields nudge the arrow, and after a fraction of a millisecond the state has drifted or collapsed. Every gate also has a small chance of rotating the arrow slightly wrong. Chain a few hundred gates together and those small errors swamp the answer.

The fix is quantum error correctionSpreading one qubit's state across many physical qubits so that errors can be detected and undone. Costly: it can take hundreds or thousands of physical qubits per reliable logical one.: spreading one 'logical' qubit's state across many physical qubits so that errors can be detected and undone. Depending on the error rate, that can mean hundreds or thousands of physical qubits per useful logical one. This is why the field has shifted from bragging about raw qubit counts to reporting error rates and logical qubits — and why 1,121 noisy qubits is a milestone, not a finish line.

A good test of understanding is whether you can connect the terms to each other rather than recite them one at a time. Try this chain: a qubit holds a superposition, which the Bloch sphere pictures as an arrow; gates rotate that arrow; entanglement links arrows into one shared state; interference arranges the phases so wrong answers cancel; measurement reads out a single result; and noise is why all of this is so hard to keep going long enough to be useful.

If any link in that chain feels loose, the stop that introduced it is one scroll up. That is the whole tour — six ideas, each only possible because of the one before.