qubit yield

Qubit yield is the fraction of qubits on a chip that actually come out good enough to use. Building a quantum processor is a lot like baking a tray of identical cookies: you set out to make a hundred the same, but they emerge slightly different, and some are burnt or misshapen. A qubit counts as a keeper only if it lands inside spec on several things at once — its frequency is close to where you designed it, it holds a quantum state long enough (good coherence), and its tiny junction works at all. If even one of those misses, that qubit is dead weight, and the share that pass is your yield.

Yield matters because the usable size of a processor is set by the qubits that survive, not the ones you drew. The trouble is that the failures often multiply rather than add. A larger chip needs more qubits, more junctions, and more couplers all working together, so if each element passes with some probability, the chance of the whole chip being flawless drops fast as it grows — roughly like that per-element pass rate raised to the power of the element count. Worse, qubits are not independent: two neighbors that happen to land on the same frequency can spoil each other even though both are individually fine, so on a packed chip the effective yield can fall faster than a simple part count suggests.

This single number quietly drives much of quantum chip engineering today. It is why teams push for tighter, more repeatable fabrication, why they tune frequencies after fab by aging or laser-trimming junctions, and why they split big designs into small pre-tested tiles joined into multi-chip modules so a dud costs one tile instead of the whole processor. Honest status: yields good enough for a few dozen to a few hundred qubits are achievable with screening and tuning, but reaching the millions of high-quality qubits that large-scale error correction would need — without hand-picking every one — is still an open manufacturing problem, not a solved one.