Partial hash inversion is an embarrassingly parallel problem: computing a hash from a random number is independent from the other random numbers you want to compute hashes for, also it is a computation bound problem and not memory bound. It means you can build a fairly efficient partial inverse finder by running many fragment shaders ("pixels") in parallel and only drawing the one that finds it. This is doable in WebGL today.

A key feature of WebGPU is compute shaders which sound like they can compute stuff so they would be better at hashing. But actually compute shaders give more flexibility which makes it possible to handle problems that are a bit less parallel, and they help optimize memory-bound programs (but don't help for computation bound programs). So compute shaders make it slightly less code to do a partial inverse finder, but do not make it dramatically more efficient.

So to answer your implicit point: no WebGPU won't unlock capabilities that will lead to a rise in ads mining Bitcoin. (though it could maybe help for Ethereum mining)