VLIW is pretty well represented in the TOP500 supercomputers and in various other performance niches. What isn't is not so much VLIW as EPIC - the explicit par

VLIW is pretty well represented in the TOP500 supercomputers and in various other performance niches. What isn't is not so much VLIW as EPIC - the explicit parallelism of Itanium, which among others didn't really support out of order or branch prediction in ways other than compiler code generator. High levels of SMT (sometimes with just a barrel execution models) are used in GPUs to smooth out the performance characteristics involved, from my understanding.