Taking this to its logical conclusion...

Google's willingness to encourage and try new ideas looks to have hit on something quite spectacular: A new data center that will fundamentally changes the "climate control" game. They are taking advantage of Belgium's cool climate, combining it with a global load-balancing system, and excising the expensive and power-hungry cooling systems usually needed to keep data center components from de-soldering themselves. (OK, the silicon bits will fail before that happens, but those computers put off a lot of heat.)

The estimate is that the center will only average seven days a year when equipment will need to be shut down due to high temperatures in Belgium. With a global system of backup and redundant servers, this is probably not much different than giving the Belgian crew an opportunity to do maintenance, anyway.

But, let's follow this out to its logical conclusion: If you could get a tube fat enough (remember, the internet is just a "series of tubes") with low-enough latency, why not stick every data center in Alaska, upper Canadia, or Siberia? (I guess I should include Antarctica, pardon the hemispherical bias, my friends down under.)

I only see a handful of troubles that need to be overcome:

1. Tube + latency
2. Energy on sight (portable, self-contained nuclear powerplants, you say?)
3. Geeks trying to overclock their data centers during wintertime
4. Geeks dying of sun exposure during summer (remember that whole sun-not-setting-in-the-summer thingy?)

(3) and (4) could be addressed by through a combination of automation and strategic employment decisions. (1) is probably the biggest technical hurdle, although I've never been overly bothered by the latency issues when reading UK rags in the US. (2) isn't so much a technical challenge as a political one; I mean, PETA is going to go absolutely ape---- when somebody draws the first cartoon of a glowing-green polar bear.

Other than that, should work.