Start with the estimate g(n) = n log n + n log log n
*, which estimates the size of the nth prime for n > 5.
Then run a sieve on that estimate.
g(n)
gives an overestimate, which is okay because we can simply discard the extra primes generated which are larger than the desired n.
Then consider the answers in "Fastest way to list all primes below N in python".
If you are concerned about the actual runtime of the code (instead of the order of magnitude of the time complexity of the algorithm), consider using one of the solutions that use numpy (instead of one of the "pure python" solutions).
*When I write log
I mean the natural logarithm.