The idea of linking myriad computers across the net for distributed power is really starting to grab people. The enormous capacity of modern personal computers is wasted on mundane tasks such as web surfing and composing email, yet can be taxed when trying to do something sophisticated like system modeling or data analysis. Distributed computing -- sometimes called "grid computing" or "hive computing" -- lets machines contribute cycles or take advantage of other machines' spare power, as needed.

Robert X. Cringely at PBS has a nice, clear essay about how this all could work, and why it's important.

[...] let's think of what we could do with a hive.  For one thing, we could put a node on every desk, but instead of being limited to the power of your PC, we could have demand ebb and flow such that you could do computational fluid dynamic modeling on your desktop as easily as you could surf the web.  Hives could be cheap adjuncts to Big Iron, or they could replace mainframes completely.  A hive is a network, so why not replace all those Cisco routers with hive nodes that happened to route as needed?  Same for wireless links.  A wireless mesh hive is very interesting.  And there is no reason why we couldn't link hives together until the whole net was just an ocean of computing-on-demand.  Then every school could effectively have a supercomputer, even the high schools.