It's technically feasible already and it's close to economic in very sunny areas.
The problem right now is cost, but that is going to fall. Then we hit another problem; and that is that the current cheapest technologies rely on rare materials. Thin film cells rely on indium, gallium or tellurium, and dye sensitized cells rely on ruthenium (although only in tiny amounts, I calculate the USGS' estimated producible Te to be equivalent to about 17GW/year of solar power. Add in the Ru, In, Ga and you're looking at about 50GW/yr. Not enough!).
Fortunately, we're working on new cells; dyes that use porphyrin or other materials which only use common metals like copper (though they still need iodine for the electrolyte). Organic cells, and thin-film tech that uses copper-tin-zinc-sulphur.
As soon as these techs work, we have enough resources and they should continue to fall in price. At a guess, I'm going to say cells will hit grid parity in California before 2020. In northern Europe it'll probably be around 2030.
By then solar power will start exploding, and I think 25% by the end of the century is perfectly reasonable once you include smart grid tech and energy storage.