While I've been out of the area for almost 3 years now (I sysadmin-supported GS FLX from Roche/454, MiSeq and HiSeq from Illumina, and PGM from IonTorrent), the thing to remember is that each sequencer has it's strengths and weaknesses depending on what you want. Do you want to focus on read length, run time, reads per run, reagent cost, equipment cost, accuracy, etc.

As for pacbio, if I remeber correctly, their cost was low (high equipment cost though) and sequencing was relatively fast, but at a reduction of accuracy. Never dealt with them first hand though.

Niche uses where long-read sequencing is required (denovo assembly, reference improvement, bacterial genomes, whole-isoform transcriptomes. Nanopore is starting to eat into that market, though.

pacbio has this fundamental problem that it's a single molecule process (as does MinION) and deconvoluting errors becomes a problem of unit parallelization.