"Prime factorization" is not worth interest, for primes are their own factorization.
Factorization into primes is not used for key generation.
I conclude the question asks:
When generating primes during generation of public/private key pairs for crytosystems based on hardness of factorization (RSA, Rabin, Pailler…), is there a limit on the size of the prime factors? If so, does that leave a limited number of prime numbers that can be used for key gen? And, if so, does that leave the encryption system vulnerable?
Mathematically, there is no upper limit on the size of the prime factors. There are infinitely many primes, and (thus) primes over any size.
Some standards put a limit. For example FIPS 186-4 has an upper limits of $1536$ bit; more precisely, for this size, each of the two primes that form the composite modulus must be in the interval $[2^{1535.5},2^{1536}]$, so that the product is $3072$-bit. By the prime number theorem, there are about about $2^{1524}$ primes in this interval. That's about $600\underbrace{\text{ million }\ldots\text{ million }}_{76\text{ times the word million}}$. That's limited, but so large that it does not make the system vulnerable for any practical purpose.
