Calculating the true costs of something as complicated as a launch vehicle is difficult and subject to a lot of debate. There are several cost factors to consider. Consider the case of the Shuttle. The NASA budget for Shuttle flights is (the last time I looked) in the neighborhood of $3 billion. If they fly 3 missions a year, then it is costing a billion dollars per flight. Some will argue that the true cost is what it would take to fly a single additional mission (the marginal cost), say $300 million or so. Personnally, I believe the more accurate cost is closer to former than the later.

Now, the Ares I is a new vehicle. There are many kinds of costs that have to be determined and the answers are not clear. For example, there's the amortized research & development costs. For the Ares I, it appears to be several billion dollars. To determine the cost per launch, you have to divide the R&D costs by the expected number of launches for the vehicle. Using round, made up numbers (I can't find the actual numbers right now, if known), suppose the R&D costs are $10 billion and the realistic number of Ares I flights were 100 (say 4 flights per year for 25 years). Simple math suggests the R&D costs are $100 million per flight.

Then there's the actual hardware costs. The Ares I is planned to have a 5 segment SRB. It'll naturally cost somewhat more than the 4 segment SRBs used on the Shuttle but it may be reusable. It'll also have a couple upgraded J-2 engines, avionics, and other disposable components that are lost with each launch. Again, using rounded made up numbers, suppose the hardware cost for the launch was $50 million. Added to the R&D costs, we're up to $150 million per launch.

Now, you have to consider the cost of the infrastructure and supporting personnel. NASA appears to be designing the Ares I more as a job protection system than a booster (IMO). If the Ares I requires a cast of thousands to prepare for launch, you need to add up the cost of employing all of those people (salary, benefits, etc) and divide that cost by the number of flights per year. Suppose all of those salaries totaled $200 million a year and you did 4 flights per year - that comes to $50 million per launch to cover the personnel costs. We're now up to $200 million per launch.

Actually, this is a simplistic analysis. It's a lot more complicated than that. The actual numbers are likely not known except perhaps the actual hardware costs (and those are likely suspect). The hard answer is that even NASA may not know how much a Ares I launch will truly cost. To use a military example, consider the case of the B-2 bomber or the F-22 fighter. In both cases, many billions of dollars were spent on R&D (IIRC, about $30 billion for the B-2 alone and somewhere around that number for the F-22). They originally planned on buying about 200 B-2s so the R&D cost per plane was going to be about $150,000,000. Instead, production was cut off at 21 planes so the R&D cost per plane was almost $1.5 billion. It actually cost about $400 million or so to build a B-2 bomber but the total cost ended up being about $2 billion per plane. The same kind of thing happened with the F-22. The Air Force originally planned to purchase was somewhere in the neighborhood of 600 planes but that has been cut to fewer than 200, greatly driving up the cost per plane. The same thing will happen to the Ares I (and the Ares V) if the number of missions gets cut. Likewise, if the flight rate is low, you'll have fewer flights per year to divide the personnel costs among.