Physical qubit counts are the number every announcement leads with and the one that predicts the least. What decides whether a machine can run a useful algorithm is how many logical qubits it can sustain, which depends on error rates, connectivity and the overhead of the correcting code - often a thousand physical qubits or more for one logical one.
That is why the field's real milestones are error-correction results rather than device sizes, and why a roadmap has to be read against published thresholds instead of press releases. Tracking it as engineering rather than as news is the entire difference between a timeline and a rumour.