Regarding the question why scientists don't buy the sequencing machines themselves I think a major factor is likely that sequencing is something that everyone expects to be a service and not something you do yourself. I would not even have considered doing this in-house.
At $15 per plasmid there is also no real advantage for typical labs, they're unlikely to save much money here. It could be quicker because you don't have to send the samples, but that likely isn't a big factor here. And having another device that either someone has to run as a service locally or that you have to teach everyone is a rather big effort.
Productionizing and scaling nanopore sequencing is definitely an achievement. With that operational obstacle out of the way, you get to think about sequencing in terms of “streams” of data rather than “batches” of huge amounts of data (illumina). That confers a huge operational and commercial benefit.
I do a lot of plasmid sequencing in-house with nanopore, so I have some insights here!
For years, the problem was mainly with the rates of mutations in nanopore sequencing, in particular, non-random mutations. The stuff you couldn't just sort out by sequencing more (well, theoretically at least). This was because of the pore structure. Around 2021, R10 flow cells came out, which were a major improvement on R9 cells. In the protein pore for nanopore sequencing, they basically just stacked two proteins on top of each other, so there are more contact points to get data out of.
With the improvement in nanopore sequencing fundamentals, you now have far fewer non-random mutations. This enables plasmid sequencing. Plasmidsaurus just hit it out the park with convenience of the service.
On cost: most of the cost is in their reliability. Each customer wants their particular plasmid to be actually sequenced - so $15 price tag (and probably much less in-house for them). I currently sequence plasmids for approximately $0.25 in-house, BUT only probabilistically (900/1500 plasmids get sequenced). And to my best approximation, this is a non-random distribution, so just doing it more doesn't fix it. Also you run out of material.
There is an aspect of library prep that is cheaper. Turns out, with illumina sequencing at least, most of the cost is in the indexing step, not the actual sequencing, for plasmids. This is the step of putting barcodes onto the material to figure out which well a particular sequence came from.
Finally, you still do need software! People don't realize how much software you actually have to make to be able to tell plasmid sequence from nanopore sequencing. Plasmidsaurus makes it easy for people. Here is an example from my repo of me figuring out how the hell to interpret some weird data I've gotten:
https://github.com/Koeng101/dnadesign/pull/98
If you want a real example, take a look at this pileup file:
https://github.com/Koeng101/dnadesign/blob/sequencingPileup/...
How to read: https://en.wikipedia.org/wiki/Pileup_format
At position 165, you see a non-random nanopore error that only affects a single side of the DNA, and not the other (there would be lowercase g if the complement was also "mutated"). You need to sort through these to get good data to customers. And that's not as easy (not to mention how often plasmids are mixed!)
Sounds great. But the focus on the competition with in-house sequencing is odd. When i was doing a PhD twenty years ago, we never even considered doing it in-house - i would just post tubes to a commercial sequencing service and get results back a few days later (or, in one case, take tubes to their office myself to save some time).
So, Plasmidsaurus are actually winning against those services. It would be more useful to focus on why that is. It sounds like the two distinguishing factors are fast turnaround, and long reads.
But there are other players offering basically the same service now, and faster or cheaper:
https://eurofinsgenomics.com/en/products/nanopore-sequencing...
https://plasmidsng.com/
So we'll see how that pans out.