Why do I say "in ten years"? Am I high? No.
Look: in 2008 I own two dual-core computers, each of which cost significantly less than $1,000, which run at 3.0 and 2.0 GHz, respectively. They both run Windows Vista and connect to the Internet via a fast broadband connection. None of this is particularly expensive.
In 1998 you could get a single-core Pentium II PC running at about 300-500 MHz. It cost a bit, but wasn't prohibitively expensive. Most people who connected to the Internet used a 56k modem to do so. Internet service was relatively cheap thanks to AOL and other services.
In 1988 you could get a machine that booted into DOS 3.3 and could run Windows 2, and ran on an 80386 processor running at less than 50 MHz. Such a machine cost a bit less than a decent used car. The Internet was the domain of university and defense industry people; on-line services charged (and charged a lot) by the hour and the best connect speed was between 2400 and 4800 baud maximum.
In 1978 you could buy any of several computers which ran CP/M on Z80 processors with clock speeds approaching 500 kHz. Only hardcore computer and electronics freaks owned them, though, because you could buy a car for what one would cost you. Connect speed was about 300 baud, and most modems had acoustic couplers: you put the telephone handset into cups on the modem to connect it.
In 1968, if you wanted a computer, you talked to IBM and paid a lot of money--millions--to have a gigantic machine in a big, air-conditioned room.
In 1958 computers had made the switch from tubes and relays to transistors. They were programmed by flipping switches.
In 1948 the electronic computer had only just been invented a scant five years earlier.
In 1938 a "computer" was a person who sat at a desk with a mechanical calculator and "did the math".
So ten years from now, I expect massively multi-core computers to be the rule, and I expect them to be cheap. Barring any serious disruption to the world economy, in fact, my expectation may be a bit pessimistic.
My current desktop computer contains more processing power than existed in the entire world in 1948--and I mainly use it to play Diablo II, Solitaire, and to surf the Internet.
* * *
This is such a good idea. In fact, it doesn't go far enough; I want one of my own. Forget "power for 20,000 homes"; I want a compact, self-contained nuclear reactor which I can bury in my back yard, which will provide all the power I can use for 100 years.
Make such a personal-scale power supply available to people and I guarantee you will eliminate coal power in less than two decades. Not just coal--people will then buy electric cars, too, because they will have already paid for the reactor and it just makes sense to charge your electric car from it, right? Instead of also paying for gasoline to run your car?
Electric heat, electric lawn mowers, electric cars, etc--if you are worried about man-made global warming, this is your solution: people won't have to worry about their standard of living decreasing, a lot of people will make money doing it, and the use of fossil fuel will just "softly and silently vanish away" without any government intervention whatsoever.
Power companies--power companies would change from being suppliers to being brokers. They would buy energy from people with personal plants who were willing to sell it, then turn around and sell the power to people who didn't want to, or couldn't afford to, own their own personal plants. All the power companies would have to do is to maintain the wires.
The petroleum industry would change its focus from providing fuel to providing lubricants and feedstocks for making plastics. Most of the refining capacity in the US would become superfluous, as it wouldn't have to make billions of gallons of fuel week after week.
Coal would only be used for making steel.
It's physically possible. Politically, that's another story.