We are not building much new transportation in the US not just because the costs are too high, but because the benefits are too low.
My 2011 post Transportation Costs Too Much made the claim that new projects were too expensive, and listed a series of hypotheses as to why that might be the case. The whole list is collected here.

When we were much younger as a nation, say 1956, and growing fast, with relatively poor connectivity, you could do almost anything and it would have a benefit/cost ratio above 1. Very little of the interstate has been reversed. But the productivity of new investments has declined over time (as shown by Nadiri, M.I. and Mamuneas, T.P. (1996) Contribution of Highway Capital to Industry and National Productivity Growth Federal Highway Administration. Office of Policy Development.)

The first 3 data points on the graph are rates of return from their work, the last is an extrapolation. This is illustrative only,but someone should update their study.

Updated Jan 10, 2012. Matt Logan points me to a study that shows the authors updated their study in 2006 (Mamuneas, T and Nadiri, M.I. (2006) Production, Consumption and the Rates of Return to Highway Infrastructure Capital)

They report net rates of return are as follows (Table 5)

Period		Net Rate of Return
1949-1950	:	0.554
1960-1969	:	0.480
1970-1979	:	0.298
1980-1989	:	0.212
1990-2000	:	0.136

My guess is that the 2000s are significantly lower.

A value of approximately zero returns in recent years is consistent with recent work by Noland and his colleagues. This is what you expect with life-cycle theory, and it applies well to existing modes.
The reason we don’t draw new lines on the map is that the net benefits are not perceived to outweigh the net costs. The costs have risen as land has gotten scarcer and for all the reasons linked above, and the benefits of additional lines drop. Here we explore the second point.

1. Diminishing marginal returns to new roads due to diminishing distance reductions as the network is increasingly complete.
   
   This is a spatial argument, illustrated in the Third Figure (Source: Figure 27.4 from The Transportation Experience). Imagine you have a network with a 1 mile grid (typical for much of the US). With development of farms, you add roads in between, say at 1/2 mile spacing, this reduces travel costs some, as people don’t need to back-track as much, and this might be a significant share of the distance for short trips. At most, you are saving someone 1 mile (1/2 mile at the beginning of the trip, and 1/2 mile at the end of the trip). Now add additional links to diminish spacing to 1/4, This requires twice as many links, but only reduces travel costs by at most 1/4 mile at each end of the trip (1/2 mile total). New links do less and less to reduce distances. Distances, along with speed, determine travel time.
2. Diminishing speed savings for new roads
   Recognition that transport links become congested, and thus slower over time due to:
   1. induced demand
   2. induced development
   3. induced driveways and interchanges, which increase friction on roadways and slow them down over time. While access management addresses this on arterials, very few interchanges are removed to speed freeways, they are almost irreversible

   Thus we have to discount the opening year forecast travel time benefits to account for the fact that the travel time savings of any expansion will in part (if not in whole) be eaten up by more travel. While this is not of itself a bad (travel is a measure of people doing something that they value), it is not perceived to be a good thing (because it creates congestion and pollution externalities which existing travelers bear).
   There is some compensation for induced demand and induced development, as more travelers may lead to more service (induced expansion of existing links and construction of new links), but this is a longer term process, and only works up to a point (and more easily with transit services than roads).

3. Diminishing demands for new roads
   1. peak travel
   2. slowed population growth, particularly in un-roaded or under-roaded exurban areas.

   Implies the benefits from new construction are falling.

4. Disillusionment with the quality of travel and transportation facilities
   The saying goes: Fast, Cheap, or Good, pick any two. It sometimes seems we have none of these. We know transport costs too much. We know transport takes too long to build. Surely if we are paying a premium, it should be of high quality. It seems not. Maybe the few facilities which are brand new, expensive, and took too long to build are, but the rest of it is not. Even then, I have doubts. Big Dig Ceiling Collapse is one of many notable failures, this one all the worse because the facility was new and expensive.An earlier post, It Just Makes You Feel Poor describes the poor quality environment around some local new LRT stations.

The arguments about benefits apply to mature systems in general, and are modally independent. If a new mode comes about that is better than existing modes for some market, it has lots of room to run while providing benefits in excess of costs. But if the new mode is inferior to existing modes, than it has little in the way of prospects.
What constitutes better or inferior is in the eyes of the customer. Certainly customers care about time and price, but they do also consider quality, and may be willing to sacrifice one for the other. But as experience with zeppelins and cruise ships (and conventional intercity rail in most of the US) show, high comfort at slow speed will not trump low comfort at high speed.
This also has implications for the Preservation vs. Expansion argument (Fix It First). If the old projects had high benefit/cost ratios, and without proper maintenance are at risk of disappearing (either failing catastrophically, or being closed to prevent such an end), it is incumbent to maintain them. In fact, given that land uses and resultant activity patterns that have evolved around transportation networks, the benefit/cost ratio of preserving those links is probably much higher than it was originally, and certainly higher than the B/C of new links.