Not a computational linguist (or a linguist at all), but this is an area I know a bit about. The question you ask is at the centre of the study of ‘glottochronology’ [0]:
> The original method presumed that the core vocabulary of a language is replaced at a constant (or constant average) rate across all languages and cultures …
Unfortunately, the only major discovery glottochronology has revealed is that rates of change vary too much to be of any use:
> in Bergsland & Vogt (1962), the authors make an impressive demonstration, on the basis of actual language data verifiable by extralinguistic sources, that the "rate of change" for Icelandic constituted around 4% per millennium, but for closely connected Riksmal (Literary Norwegian), it would amount to as much as 20%.
And there are other factors affecting replacement rate as well. For instance, a curious trait about the non-Austronesian languages of New Guinea is that the word ‘louse’ is practically never replaced: it evolves according to normal sound change, but never gets thrown out entirely. By contrast, many languages have a taboo against mentioning the names of the deceased; this can speed up lexical replacement if those people have names homophonous with common words.
For these reasons, I suspect that a general answer to your question will be extremely difficult — if not impossible — to find.
Could a computational linguist chime in please? How many words in how many human languages will never be spoken again in 2021?