I think a better analogy is the way GPU calculations are done, as the calculations are done in parallel but the process is ultimately sequential: move data from CPU RAM to GPU RAM, calculate in parallel on GPU, move results data back to CPU RAM, operate on the results data on the CPU serially.