MUSCLE is a multiple sequence alignment program, now this might sound a bit confusing but this doesn't mean it is meant to align very many sequences (especially not that many), at least not to a single reference. It tries to align all sequences (maybe max a few hundred) with each other to discover a consensus, which is not what you want to do with short sequencing reads. You should instead use a short read aligner, like bwa, hisat, or bbmap, see a list here. In fact what your computer is likely telling you is that you are running out of memory because you are using the wrong tool.