I would guess that optical computers will become more common. Since you can use photorefractive crystals as transistors, and since they're also the 'magic' part of printed holograms, you should be able to 'print' a new computer with included software for fifteen cents per copy or so. (optical computers would be much faster as most people here probably know, I've read that the maximum SpeedOfElectrons in copper is something like half the speed of light, any corroboration?)

''IIRC the charge-carriers themselves bumble along very slowly, inches per minute, for "normal" current densities. The changes in the ''field'' that transmit the information move at some respectable fraction of the speed of light. The advantages of OpticalComputing are not in faster signal propagation, or even faster switching; but are in bandwidth, component density and the inherent massive parallelism (without cross-talk) of optical components.''

For advances in the current style of hardware, I'd expect it to become extremely modular, something like plugging a bunch of tetrahedrons together to make a complete system. (need more hard drive space? plug in another secondary storage tetra, need more processing power? ...)

[Legos.  I want a lego sheet back plane with lego brick components that snap on.]

----

PDA's you can accidently sit on and break without taking out a loan.

''ExternalBrainFailure?  Nooooo!''

[How about a flexible PDA so you can sit on it without breaking it?  Why do they need to be so rigid, anyway?]

Well, it can still fall in the toillet, get lost, etc. I would really like one, but am worried about loss.

[Buy an old model on eBay.  $60 gets a mint m100 these days, and that's all I need.]
----

Storage >> Processing >> Networking

Or does Networking grow faster than CPUs? I forget. In any case ...

Storage is growing far, far faster than either processing or networking capacity. Currently doubling every 9 months or so when before the first GMR revolution, it was doubling every 2 years. At the rate it's growing, within a couple of decades we'll be able to create hard drives so big that neither processors nor networks are ever able to fill them. An uncompressed movie of your entire life would never be able to fill it. Eventually it will get where a holomovie of your life will never be able to fill it. Consequences?

The industry will have to switch to ever faster storage and possibly random access storage.

----

CellularAutomata, Logic Arrays

The chips that implement ReconfigurableHardware.

And actually, there are good reasons in fundamental physics why this is a good architecture. There's a paper on ForeSight called Ultimate Limits to Computing or something which is about the physical limits of computation. Basically, the information processing in a volume outstrips the information flow between volumes.

''Solid state computers with nanomachine flywheel generators for power.''

-- DavidCary