I was working in IT during the mid-70's through to 2015. The rise of the personal computer was inevitable, along with the networks and server farms to which they connect. Our terminals then were developed with more and more local processing, with integrated peripherals, multiple connections to the outside world, and increasingly complex on-board applications and processing.
As an example a requirement I had to fulfil was to enable a dumb terminal, connected via telephone line to our corporate mainframe, to run a local word processor. This was because of woeful communications speed and frequent communications failures. The specifications, simple at first, quickly grew to include interfacing a printer (golf ball; remember those?) and off-line data storage (the boss needed that report now, not when the techs in town finally sorted out their problems).
We wound up with what was effectively a personal computer, though with none of the niceties like graphic user interfaces or mice. I did incorporate a graphics tablet with wired pen for quick input later. The processor was Motorola, the firmare written by hand (I think they would call it bespoke now). The connection to the mainframe also improved over time, eventually rising to the blistering speed of 250Kb via a dedicated line.
The penultimate is the current smart device: a computer, terminal, telephone, camera, and applications in your pocket or on your wrist using voice control to connect and communicate with the rest of the world, video included; Dick Tracy indeed.
The ultimate will be neuro connectivity, bypassing the need to carry or wear something externally. Such a system is the logical extension and terminus of the distributed network, possibly sharing "spare" brain capacity the way some personal computer networks share resources now.
I don't believe that there is a place for large monolithic computers with users sharing space and resources outside of an entity such as a government department or scientific establishment "super computers"; such a workspace needs to have both its connectivity and resources rigidly controlled for safety and security but even there a distributed network of computers and servers is better in terms of speed and data backup.
Monolithic computing was never about shared access to data, it was about shared access to expensive computing cycles, very expensive memory space, massive and slow data backup, and finicky expensive peripherals like printers and plotters.
What we have now would have come about eventually anyway; the evolution of a system of massive mainframes would have led to this outcome as well. We now have machines on our laps much more powerful than the corporate silos of 10 years ago, with communications speeds undreamt of previously, and interplanetary connectivity. In terms of communications infrastructure we have so much more that we can do and need to do to enable neuro connectivity.
Humans like autonomy, feeling that they control their own information (good luck with that) and can switch off when they need to. Everybody relying on a large, centralised system with its oversight and accountability was never going to continue on a personal level and in today's world would certainly be the cause for disaster.