in theory. but if your input for that hash is 32 characters long no datacenter in the world would be able to store such a table. and that table would only be good for one specific hash algorithm.
The length of the plaintext is irrelevant. You can find a collision, with very high probability, with a much smaller database. All you need is a forward string within the constraints of any input validation that matches every hash output, of which there are a fixed number.
There aren't that many hash algorithms in wide use. Salting deals with the problem, but many, many applications don't use it, or don't use it properly.