GPUs?
The question was raised in the article whether Intel would optimize for ARM as well as Intel architectures. But if the focus is on HPC, they would also have to take GPUs (especially Nvidia CUDA) into account. This strikes me as a much tougher problem.
But presumably, even in its absence, there' are a lot of of non-HPC workflows (and HPC applications that cannot or have not been CUDAfied) that would benefit.