I often find Robert Cringley to have a lot of good ideas and such. He has, as of late, been getting into a few rants on Google. While I posted the below to the comments section of one of his articles (see here), I would also like to post it here. I think this is really likely. Enjoy...
A couple things....
1) Between AOL, Yahoo, and MSN - well, AOL needs to die. It's old, and decaying, and built on AOL-based (non-standard) protocols that AOL controls. I love AIM (which needs to somehow survive AOL dieing), but AOL needs to go. Also, MSN needs to die too. Sure, Microsoft will fight tooth and nail to keep it from it, but let's face it - it (like AOL) does not really offer anything but a lock in to a certain vender (Microsoft, as opposed to AOL), and we all know how evil Microsoft is. A nail in MSN's coffin is what is needed - well, not just one nail, but enough to seal the lid shut sow we can put it six feet under. Which leaves Yahoo - Yahoo actually offers something useful, though they may not offer it well enough to survive. Yahoo does not lock in users to any one vendor (Windows/Linux/Mac, AOL/MSN/home-town-ISP), but does provide a number of services (small business hosting, good e-mail, etc.) that do well. So, Yahoo needs to survive; but AOL and MSN need to see the mortician.
2. There are two problems with the rest of the article. (a) Interconnecting the ISPs as you suggest really only puts more Internet connections together - that is, after all, how the Internet is actually formed. ISP A agrees to allow traffic from ISP B to ISP C; ISP D - peered with ISP C - has someone trying to access ISP B, so ISP C lets it through to A which lets it through to B. It's really a simple formula. Putting more ISP-to-ISP lines in, while not riding on the central backbone providers (the really BIG ISP - L3, etc.) simply extends the Internet traffic that much more. It would primarily create a second Backbone network; sure it may not go directly to the primary backbone but it would still do the same thing with a slighly higher latency. It'll be quite hard to do otherwise.
(b) It is also just as likely that Google is looking to sell those datacenters as really big versions of their already selling search appliances for companies; thus competing more with main-frame systems than with ISPs or video, etc. Given Google (i) already has such appliances on a smaller scale, and (ii) focuses their primary business in search, I think this is a more likely scenerio.
Remember - Google is not really trying to compete with Microsoft; but if they happen to, so be it. In this case, Google would be leveraging Open Source technologies that they are already using in their own datacenters to sell a really large appliance to businesses to supplant main-frames that do heavy data crunching (analysis and searches). Think searching the FBI fingerprint databse, or facial recognition search type facilities.
They would be competing with Microsoft in terms of supplanting Microsoft Windows in the main-frame market. They would also be competing with Sun Microsystems (oh wait - isn't Eric Schmidt on their board?), IBM, SGI, HP/Compaq - and their AIX, Irix, Dynix, HP-UX, etc. products; which are already being eaten alive by Linux, which also happens to be Google's OS of choice. Ironic, isn't it? And doesn't this sound so much more like Google to start with?