Physical vs Logical Qubits: Why Error Correction Is the Real Race

Today’s best qubits still make an error roughly once in every hundred to a thousand operations. That sounds small, but useful algorithms need billions of operations. Without fixing errors, the result would be noise.

A physical qubit is a real piece of hardware: a tiny superconducting circuit, a trapped atom or a particle of light. A logical qubit is a reliable qubit built by spreading information across many physical qubits, so that errors can be spotted and fixed without destroying the information.

Quantum information can’t simply be copied, so quantum error correction uses clever tricks to check for errors indirectly, without looking at the data itself.

Error correction only works if physical qubits are already good enough, below a certain “threshold” error rate. Above it, adding more qubits just adds more errors. Below it, bigger codes make logical qubits more reliable.

In December 2024, Google reported that its 105-qubit Willow chip had crossed this line. As it grew its error-correcting code from a 3×3 to a 5×5 to a 7×7 grid of qubits, the logical error rate roughly halved at each step. It was an important sign that the approach can scale.

Estimates vary with the hardware, but one high-quality logical qubit may need hundreds to thousands of physical qubits. Breaking RSA-2048 has been estimated to need around a million noisy physical qubits. So a machine with a thousand physical qubits today is still a long way from breaking encryption or simulating complex molecules.

• ◆Ask whether the qubits are physical or logical.
• ◆Look for error rates (often called “fidelity”), not just qubit counts.
• ◆Check whether a result beat the best classical method, or only a simple one.