1) Google Fiber is dead. Long Live Google ISP.
Google cannot afford to lay fiber in the ground because it's a long game and Google doesn't really want to play the long game (they just want to put pressure on competitors so they can move more bits along the wires, generating more searches and more streams with which to shove ads in your face). Fiber never was the most efficient way to do this, but there's something sexy about "Google Fiber" as opposed to "Google Point to Point Radio Towers". The reality is that wireless delivery of bits is way cheaper than fiber because you don't have to tear the ground up (over the last mile obviously since the arteries must be fiber links).
Clarification: "long" refers to 30-year payback periods for physical asset investment. There are a lot of things with higher returns on that timeline than fiber that google can invest in. I don't actually think it's profitable for anyone to build unsubsidized networks. This is why networks should be public and operators should be private, but that's a topic for another day!
2) Fiber is cheap, construction is expensive.
When I helped Comcast build out the fiber network in SF, what struck me was the relative cheapness of the assets we were putting in the ground compared to the cost of tearing up the street. The conduit and the glass inside the conduit cost almost nothing, but tearing up the street in SF is $300/sq ft. Crossing cable car tracks was like $50,000. Then there's the actual cost of construction: people. Getting contractors to arrive on time, finish on time, and avoid overtime is fraught with peril. It's actually really hard to move physical atoms around in a manner similar to programmatic systems, and so many models that have real world elements stumble against the harshness of actuality. I suspect Google's cost modeling for building a fiber network was optimistic.
3) Wireless is fast, but does it scale to city size?
It's not hard or particularly expensive to deliver gigabit over wireless. You basically need a tall building to rain down radio waves onto the masses. What I wonder about, given that we have no cities running on majority wireless point to points, is what happens when you hit scale? That is to say, point to points have a limited wireless footprint (because using beamforming we don't need to splay the signal everywhere, we just send it in one direction), but one can easily imagine a saturated wireless environment as generating a significant amount of noise. Wireless networks are easy when there's only a few objects on the network but get significantly harder as the physical area reaches device saturation. That is to say, WebPass might be super easy to operate when only a handful of buildings are on WebPass, but it might be much harder if a whole section of the city is online.
4) Google bought WebPass a while back.
The writing has been on the wall for a while that the fiber game was killing uncle Google. I can only hope that they don't bow out completely. I think that wireless makes Google significantly less of an existential threat to their carrier partners as well.
Overall, I remain cautiously optimistic about Google's future as an ISP.
On a final note: Google Fi is not an answer to Google fiber disappearing. The two are tangential, disjointed offerings that cannot, for a bunch of reasons, compete with one another (most notably the wireless data caps).