While the Wall Street Journal reports that Google and DISH are in talks to build a wireless (presumably national) network, the media pickup on this is interesting. To be sure, it makes good for good reading. Google has tried to expand from its core search/advertising business before. The Android OS is probably its greatest success in the mobile space, more so than its hardware statements in Nexus smartphones and tablets. The Chromebook is great concepts to dovetail with the cloud craze the industry and business community is all really keen on these days.
People can also point to Google's capabilities in building data centers and fiber networks. The biggest testbed for a hardcore service provider play is in the suburbs of Kansas City with Google Fiber. There is no doubt that Google has the deep pockets and some technical capability to embark into the wireless service provider game. However, it lacks the expertise as it's outside Google's core competency. Though it may be true that they can acquire the talent, a national network is no small feat and tremendously capital intensive. Google's signaled its intention to play in wireless sector in 2007 with Auction 73 (700 MHz spectrum).
This time DISH has the spectrum and it needs a buildout partner. DISH is a service provider and has the correct service culture but it doesn't have any terrestrial cellular expertise. It would make better sense that DISH would approach an existing national carrier to host its spectrum. Sprint's Network Vision was built for this scenario.
One cautionary tale is Cox's attempt with its own AWS spectrum. After a year or so, the company exited the wireless retail business. Incumbent wireless/cellular carriers have the towers, tower relationships, right of way clearance expertise and backhaul networks. Google does not have any of this except perhaps the fiber/IP backhaul as a critical component. If DISH finally delivers on a wireless service promise, Google may not be the entire infrastructure builder but perhaps a funding partner with a formidable IP backbone.