The location of Google's Mountain View access points can now be seen here (via Niall and GigaOM). Since counting access points on maps is one of life's simple pleasures (right after software and porn): there seem to be 364 of them (any other bean counter out there besides me?), though apparently Google (GOOG) officially reported 350 nodes (see Dennis Cheung). Tropos Networks is the royal supplier for the access points, for backhauling GoogleFi is using Alvarion (ALVR) equipment - not surprising, since Alvarion and Tropos have teamed up together to conquer the municipal Wi-Fi market. There are three Alvarion BreezeACCESS™ cells and one in seven mesh nodes acts as a gateway and connects back to those cells.
So let's do a little back-of-the-envelope Google Wi-Fi Economics. In Wi-Fi for the Masses (via Glenn), the retail price of a Tropos mesh router (that's the official marketing term Tropos uses for its Wi-Fi routers) is reported to be US$ 3,500. One might wonder why Tropos access points (some background) are so much more expensive than other carrier-grade equipment, e.g. a Cisco Outdoor Aironet access point costs around US$900-1000 on the web. Tropos charges dearly for their mesh algorithm, their secret sauce, so Tropos is actually somewhat more of a software company with a hardware shop in the basement, similar to Cisco.
Tropos access points
Alvarion Breezemax stations
Truck rolls and installation
According to the latest US Census data (via Wikipedia), there are some 70,000 people in 31,000 households (2.3 pop/household) living in Mountain View, spread across a surface of about 12 sqm. So Google is spending US$83,500 per square-mile and US$33 per household passed (for your reference, DSL and cable are usually in the US$80-90 range...), and if we assume that each person in each household is connected and can enjoy the maximally possible 1 Mbit/s, then the Capex per usable Mbit of bandwidth is US$14.31, whereas Cable and DSL would be in the US$40 range. Of couse these figures can vary depending on offering and oversubcription factors, and there would have to be some additional capex items to be considered for the Google network too really be precise and comparable, but the point is that for these KPIs cable and DSL seem to be roughly at least twice as expensive (if my calculations are correct, that is), so wireless mesh has a clear cost advantage.
Google's plan is to recover the costs through advertising, hence ad clicks by users (let's ignore the paid premium service for the moment). Assuming that 30% of Mountain View's households will use Google's service (21,000 users) and that the initial investment has to be recovered over two years (US$42,000 per month), then on a per-user-basis this turns out to be a very low US$2.0 per user and month (that ignores opex, and the estimates might be completely far from reality). I'll cover in a follow-up post if that is financable through advertising and if there will also be enough room to cover the opex of the network.