I too am very interested in this topic and was recently calculating the cost for my own methods. Without a doubt it is most cost effective to make the library yourself and submit it to your institution’s sequencing core. But again, there are some up front investments so if you don’t plan on doing sequencing or amplicon sequencing often it probably does not make sense to make those purchases.
I am aware of people paying anywhere from 50-350 USD/sample for amplicon sequencing and there are a number of choices that can drastically affect the price including sequencing depth: I am aiming for 50,000 reads/sample. For the protocol I use with everything ‘at cost’ the per-sample consumables cost to get from frozen sample to fastq is about 15$/sample. That price doesn’t include HR and amortized equipment and service contracts (mostly the sequencer) which add a significant chunk and explain a lot of the discrepancies between the consumable cost and core prices.