I mean, imagine a future where every computer is just a chromebook, phones are no longer phones but just a “terminal” that streams the actual OS which runs in the cloud.

It will get close to happening for nearly all computing, then it will swing back the other way to local storage and compute, then after 15-20 years it will swing back toward centralized compute and storage. This has already happened 3 times.

• Original computing was mainframes. “Dumb terminals” that had zero local storage and only the most rudimentary compute power to handing the incoming data and display it, and take keystrokes, encode them, and send them on.

• Then “personal computers” became a thing with the advent of cheap CPUs. Dumb Terminals/mainframes were largely discarded and everyone had their own computer on their desk with their won compute and storage. Then the Netware/Banyan era began and those desktop computers were networked to have some remote shared storage. (there’s a slightly different branching with Sun/HPUX/DigitalUnix and Workstation grade hardware)

• Then Citrix WinFrame and Sun Ray stateless thin clients showed up once again swinging the compute and storage almost entirely remotely to centralized heavy powered servers with (mostly) dumb terminals, but these were graphic interfaces like MS Windows or Xwindow.

• Once again, powerful desktop CPUs showed up with the Pentium II etc compute was back under users desks.

• Now phones and tablets with cloud has show up, and you’re asking the question.

So what I think will swing primary compute and storage back to the user side (handheld now) is again, cheap compute and storage on the device. Right now so many services are cloud based because the massive compute and storage requirements only exist in volume in the cloud. However, bandwidth is still limited. Imagine when the next (next?) generation of mobile CPUs arrive, and with a tiny bit of power you could do today’s bitcoin mining on your phone or process AI datasets with ease in the palm of your hand. And why would you send the entire dataset to the cloud when you can process it locally and then send the result?

So the pendulum keeps swinging; centralized and distributed, back and forth.