- If we are talking about storing one qubit per defect: reproducibly/reliably/scalably creating such defects in large numbers is impossible for now, so the answer is either "they simply can not be built" or "just as long as a single defect if we imagine we could build them"

- If we are talking about storing one qubit in a 100 defects: same issues as above hold, but if we imagine that we can build a 100 defects... and if we imagine we have control over such ensemble of defects... and if we imagine we can perform logic gates between neighboring defects, then we can use error correcting codes, so a hundred atomic defects can store the equivalent of one logical qubit for (exponentially) longer than a single defect. But all the things I am imagining here are impossible for the moment (but people are working on it).