Why qubits need correcting
Every quantum-computing headline carries a qubit count. Almost none of them tell you the number that matters: how many of those qubits you'd need to sacrifice to make the rest reliable. This is the plain-terms version.
The problem: qubits are noisy
A classical bit is a switch that stays where you put it. A qubit is a fragile physical system — a superconducting loop, a trapped ion — that interacts with its environment constantly. Operations drift, states decay, measurements misread. Current hardware performs a two-qubit operation correctly somewhere around 99–99.9% of the time. That sounds high. A useful computation needs billions of operations. At 99.9%, errors accumulate long before you finish.
The fix: trade quantity for reliability
Quantum error correction spreads one piece of quantum information across many physical qubits, so errors on any few of them can be detected and repaired. The bundle behaves as one logical qubit — the unit that actually runs algorithms.
The overhead is the whole story
The cost of that bundle — the number of physical qubits per logical qubit — is the correction overhead, and it ranges from dozens to thousands depending on hardware quality and the target error rate. A "1,000-physical-qubit" machine at an overhead of 100 is a 10-logical-qubit machine. Roadmaps that promise useful quantum computing are really promising to drive that overhead down: better gate fidelity means fewer physical qubits per logical one, which is why sub-0.1% two-qubit error rates are the threshold that changes the math.
The engineer's takeaway
Read every quantum milestone in three numbers: physical qubits, error rates, and logical-qubit yield after correction. A headline that gives you only the first is marketing. The day the third number reaches the hundreds with fault-tolerant error rates, the field stops being research and starts being infrastructure — and you'll read it here with all three numbers attached.
Independent technical analysis — not investment advice or engineering counsel.