Looks like according to an old article linked on wikipedia they spent another $100+ million over 5-6 years to keep it going. So not wildly expensive, but not cheap either.
I remember talking to a high up on the program at a FIRST robotics event who complained about them running so long. They weren't making any major new discoveries and sucking up resources and mindshare stopping him for kicking off and focusing on new projects. He said nobody would take the PR hit to decide to just end a mission like this so he was stuck.
I believe Google has publicly said a few things that are slightly contrary to this. Their vehicles are allowed to exceed the speed limit by 5%-10% in order to keep pace with a car in front of them. They do this because currently there are few things as reliable in terms of road information as another car being driven by a human. So the code makes an exception basically saying that its better to break the speed limit in order to be able to do the lazy thing and follow another car.
Having worked on an autonomous car and competed in the Urban Challenge I can attest to the safety of these vehicles. They truly can navigate obstacles that are withing 2-3 inches of the car reliably and confidently. The other interesting thing is that highways are in many ways one of the easiest types of road to drive on. Highway driving while fast, is very structured. There are relatively few rules to follow and they are rarely broken. For this reason I believe highways will be the initial realm of autonomous vehicles and that they will be allowed to drive at higher speeds(perhaps under the guise of human control of the speed in order to shift the blame to a human party.)
https://web.archive.org/web/20120504171453/http://www.msnbc....