Probably not a complete break. But there needs to be paradigms that can allow for exacale computing. Instead of 20000 independent threads (e.g., SIMD / CUDA style) we will have trillions of threads that work in interleaved harmony. The von Neumann model break downs at that level.
We've made a ton of progress. They didn't have decentralized computing at all in 1978. Now most things run on remote machines. Also, there has been a move away from the imperative style that makes parallelism so difficult.
44 years later, it hasn't happened. I'm skeptical that it will happen in the next 28.