If you launch cheaper satellites more often: - you get to iterate so you can make improvements at a much faster pace - there is less problems from a single failure - more scale effects
You can enter a virtuous cycle.
There are limits! It can be that it's hard to make a useful satellite once you go below some certain size. And reliability suffers if you try to skimp on good practices.
On many scientific satellites, the instruments are one off anyway and are complex, hard and hand built with astronomical costs, so it might not save so much even if the launch and the bus were a lot cheaper.
Microcomputers came about because, at some point, we could make small computers with just enough processing power that they could be useful.
Mainframes didn't really go anywhere - they are still doing heavy lifting in large corporations, and modern mainframes employ every trick we use to make x86's fast and then a couple that would make x86's prohibitively expensive.
The lifetime cost of the hubble is about 10 Billion and the development cost of the James webb telescope is approaching 10 Billion. James Webb will launch on an Ariane 5 rocket, which costs about 140 million per launch. This is about 1-2% of the development costs.