It is not unreasonable to think that
de novo synthesis could occur by the enzyme itself providing the nucleophile for the addition of the first nucleotide. There are, in fact, examples of
protein-primed DNA synthesis where the nucleophile to which DNA polymerase adds the first nucleotide is a protein, not the 3' OH of an existing nucleic acid.
RNA polymerase, however, does not utilize a protein priming mechanism for the initiation of
de novo RNA synthesis. Rather, RNA polymerases contain pockets – distinct from the nucleotide binding pockets used during processive primer extension – capable of binding a pair of nucleotides and linking them together to form the nascent RNA chain.
Why would RNA polymerase have this ability yet DNA polymerases lack this ability? Well, according to the "RNA world" hypothesis, RNA preceded DNA as the genetic material, so the first nucleic acid synthesizing enzymes to evolve would have been the RNA polymerases, and these enzymes would have needed to carry out
de novo synthesis in order to make nucleic acids. Thus, when DNA polymerases emerged later during evolution (DNA polymerases are evolutionarily distinct from most RNA polymerases), the RNA polymerases were already capable of synthesizing primers, so the DNA polymerases may never have needed to evolve the ability to perform
de novo synthesis.