You're not wrong, but people have been trying to do this for a while now without much success. So much that you can probably get a turing award by making a SAT

You're not wrong, but people have been trying to do this for a while now without much success. So much that you can probably get a turing award by making a SAT solver that can be 100-1000x faster on a GPU than on a CPU. Also that would lead to solving a bunch of NP complete problems 1000x faster, so you can see why that would be a big deal.