Headlines about quantum computers usually lead with the number of qubits. It is an easy number to compare, but on its own it says little about what a machine can do. The more important question is how often those qubits make mistakes.

Why qubits are noisy

Qubits are physical systems, such as superconducting circuits, trapped ions or neutral atoms, that must stay isolated from their surroundings while still being controlled precisely. Any stray interaction disturbs the delicate quantum state, a process called decoherence. Control pulses are also imperfect. The result is that each operation on today’s hardware fails with a probability that is many orders of magnitude higher than in a classical chip.

For short programs that is manageable, and much current research works within these limits. But the algorithms with the largest expected benefits need millions or billions of operations. At current error rates, the answer would be lost in noise long before the end.

Error correction

Classical error correction copies information so that a single flipped bit can be outvoted. Quantum states cannot simply be copied, so quantum error correction spreads one qubit of information across many physical qubits in an entangled code. Extra measurements reveal which errors have occurred without revealing, and so destroying, the information itself. The errors are then corrected or tracked in software.

The protected unit is called a logical qubit. There is a catch: if the physical qubits are too noisy, adding more of them makes things worse. Below a certain error rate, the threshold, adding more physical qubits makes the logical qubit more reliable. In late 2024 Google reported a surface-code experiment in which the logical error rate fell as the code was made larger, a clear demonstration of this below-threshold behaviour on real hardware.

How to read quantum announcements

  • Ask about error rates, not just qubit counts: two-qubit gate fidelity and measurement fidelity matter most.
  • Ask whether the result uses physical or logical qubits. A logical qubit can take hundreds or thousands of physical qubits.
  • Ask how the result was checked, and whether a classical computer could reproduce it.

The path to useful, large-scale quantum computing runs through error correction. Progress there is real but gradual, and the most meaningful milestones will be counted in reliable logical operations rather than in raw qubits.