Fast Net News, DSL Prime, DOCSIS Report

At Best Buy a gigabit switch costs less than $50/port, so any geek can tell you even carrier-reliable gigabits just aren't very expensive any more. Sergey Brin, who personally inspired Google gigabit plans, is right on target for any network built today. Verizon is already planning to upgrade FiOS to that speed in a few years, and Juan Vela tells me they have already shipped 100,000 ports to customers. Everyone in Singapore is getting a gig connection within the next two years, and literally millions in Asia already have shared GEPON that hits top speeds close to a gig when traffic is modest.

It requires a few limits: a cap of 500-700 gigabytes, twice the highest cap today in the U.S.; efficient and experienced management; and volume that could be quickly achieved with cooperation from a city like Seattle. The cap would go up over time as Moore's Law brings down bandwidth costs. There is amazing little difference in running costs when higher speeds are installed; you don't get much more email or even watch much more TV over the net if your connection goes from 10 megabits to 100 (J:COM, Japan) or 200 (Novus Net, Canada.) The actual increase in bandwidth demand averages about 30% in those two upgrades according to their technical leads. The increased bandwidth is likely to add a few bucks at most to monthly costs in a large deployment.
Seattle's Mayor McGinn told the local papers he is going for it and he has strong backing. Qwest in the city is so second rate that they were looking to spend their own money on a fiber build, and would welcome Google's involvement. Seattle, like the also committed Somerville, Mass, has the natural medium density that Verizon FiOS proved is practical to fiber. (FiOS has problems to deal with, but Verizon's continued high profits after deploying 15M lines make clear it's not impossible.)
With local support helping bring in customers and government cutting the red tape there's a good chance Google  or anyone else competent  can make a build like this profitable, especially where the telco doesn't put in their own fiber and the cableco charges double the world price for high speeds. The 16M fiber lines deployed in the U.S. have shown how to reduce construction costs and the equipment gets better and cheaper every few months.  It's still a tough battle to overbuild. Both UTOPIA in Utah and Burlington in Vermont came in so much over budget they are struggling, but others can profit from the lessons learned.
The $3,000/home figure some press have been picking up is based on small, extremely rural buildouts. Running fiber along poles costs about $20/K mile. If you have only five homes on each side of the road, that's $2,000 apiece for the fiber run. Bell Canada recently told investors their fiber home costs are about $650 and I have a similar figure from France. Most of the European builds come in between that figure and about $1,300. So unless Google puts most of its money in very small towns, the cost will be close to Verizon's plus a factor for the smaller size of the projects. If they keep to their modest plans, the total spending will be a few hundred million and many of the projects self-sustaining and even profitable.