How about them apples?

Based on the FCC Fiber Build Workshop and USAC service provider training, it's becoming a little clearer what applicants are supposed to do to make the apples-to-apples cost comparison for dark vs. lit fiber.

First, you determine what your term will be.  USAC mentions terms of "5, 7, 10, or even 20 years."  I've seen 40-year IRUs, but 20 years does seem sufficient.  For one thing, the highest speed I could find anyone cramming down current fiber cable is 15.5 terabits/second, and you'll probably need more speed than that before we reach year 30.  Within the next 20 years, you'll want to move to a better fiber optic cable.

Once you've figured out how long the term of your dark contract, you just compare the total cost of a dark fiber (self-provisioned or leased) and compare it to the total cost of leased fiber (or a service agreement for a certain amount of bandwidth).

Simple, right?  Not really.

Let's look at what ought to be the simplest: the 20-year cost of a service agreement.  Just multiply your current monthly cost by 240, right?  Well, that hides two bad assumptions: 1) your bandwidth needs will stay flat for 20 years, and 2) the cost of bandwidth will stay flat for 20 years.  Neither is true.  You could assume that you'll keep your monthly cost flat, and just take whatever bandwidth you could get for that price, but I'll bet that hasn't been true for you in the past.

Not to worry: I am coming to your rescue with a spreadsheet to calculate future costs.  First, to project future bandwidth needs, I just used Nielsen's Law of Internet Bandwidth: bandwidth increases by 50% per year.  (OK, it's based on one guy's residential connection speed, so it's pretty bogus, but it's amazing how often I see it quoted.)  For future costs, I found data showing annual decreases in cost/Mbps between 27% and 35%.  So let's look at the results for 20 years out, assuming you're currently paying $1,000/month for a 100 Mbps circuit:

Bandwidth Increase	Cost/Mb decrease	20-year cost
50%	27%	$649,466.79
50%	35%	$190,709.91

Well, that's quite the stark difference, eh?  That's why your financial adviser is always going on about compound interest.  The bad news for the telcos is that the 27% is probably a more accurate number, since the 35%  number is a based on a small sample and covers backhaul prices, not retail.

My cost-estimation spreadsheet is cooler, though; you just fill in the cells shaded yellow with data from two years from your own network: the current service (or from the best bid) and any year from the past.  The example that's in the network shows a T-1 in the year 2000, which cost $400/month, and the current 100 Mbps Ethernet circuit, which costs $1,000/month.  The spreadsheet annualizes the cost/Mbps decrease (19.67%) and bandwidth increase (32.31%), and then extrapolates those costs over 40 years.  If you go to cell F39, you will see that, based on the past for this school district's network, your total cost for a 20-year service agreement is $455,883.34.

By the way, column B estimates how much bandwidth you'll need each year.

Also, none of those cells are locked, so you can mess with the spreadsheet any way you want.  And if you mess it up irreparably, you can just download a fresh copy.  By the way, the PriceDecrease worksheet shows how I came up with the 27% and 35% annual cost decreases, if you're interested.

I wouldn't bet much on the numbers actually turning out to be right 20 years from now (or 5 years from now), but they are based on real data....

Tomorrow, let's look at the cost of leased fiber.