David Morgenstern of eWeek wrote a story today "Google And Its Continuing Dark Fiber Mystery" with various theories on why Google bought all that dark fiber. The latest theory, which I find a little puzzling, is that Google needs the dark (unused) fiber to support IPv6 and the trillions of resulting IP addresses. Other theories suggest the bandwidth is required for; Grid Computing, Google Video, and to tie together Google Wi-Fi networks. Mr. Morgenstern also linked to my blog "Google Data Centers and Dark Fiber Connections".
I asked Eric Schmidt about this dark fiber question a little over a year ago at the Nantucket Conference. His answer was fairly simple. Google has data centers all over the world with copies of the web index. They need the fiber bandwidth to keep the indexes in synch. Besides, it was available for pennies on the dollar from bankrupt telecom companies, so why not buy up a bunch.
I can tell you from personal experience that this is indeed true. Back in the late 90's I was director of engineering at AltaVista, the original web search engine. We had mirror sites in Germany, Singapore, Japan, and several other countries. These indexes were HUGE, and copying them to other data centers over the Internet took several days. Of course new pages are being added to the web every second, so the indexes were instantly out of date, and out of synch.
My guess is that Google's purchase of dark fiber started out as a purely practical decision, but evolved to a strategic direction once they understood all the things they could do with the enormous bandwidth. It takes a lot of bandwidth to support things like Google Video, Photo sharing, GMail, and all the SaaS applications they are cooking up. A few seconds better response time makes a big difference to impatient users.
Great move by Google, and for pennies on the dollar too. Don't worry, there is still plenty of unused fiber available for sale, but maybe for nickles and dimes on the dollar instead of pennies. Still, very reasonable.