My understanding (which is admittedly very little) though is that homomorphic encryption requires many times more computations compared to running on unencrypted data, so that could dampen it's usefulness.
Most problems worth sourcing out requires a lot of data, and piping that through a home network connection to a "meh" CPU and back isn't that useful.
Even if you spend time divvying it up, I can't imagine it would be easier/cost effecitve than just spinning up an AWS node to get it done.