They generally don't, actually. Mars Rovers were about 300 million a piece. The highest end-expense ones like Cassini do come in at 3 billion or so.
And they've actually discovered a lot of amazing stuff. Name me one interesting discovery of the International Space Station.
And the total costs of all countries involved for the ISS are difficult to estimate, but I've seen figures as high as 100 billion.
In the late '90s, NASA adopted a "faster, better, cheaper" strategy. Small, straightforward probes using off-the-shelf technology and with a relatively narrow focus. I believe Pathfinder/Sojurner cost only (heh) $100 million.
The idea was to split up the tasks into many small missions rather than a single big do-everything machine. That way, if a probe got lost -- as so many Mars probes have -- it would not derail the whole mission.
The rovers Spirit and Opportunity have been the most spectacular successes of this new strategy. Conservatively expected to run for 90 or 180 (I forget) days, they're close to four years and still kicking ass. I so incredibly want a radio-controlled rover replica. With a wireless webcam. Those little buggers rock.
The Cassini/Huygens mission was the last of the do-everything missions; the planning process that began it was obsolete and discontinued by the time it launched. But Cassini, too, performed well beyond expectations.
In the first Star Trek movie, the Big Scary was V'ger -- which turned out to be Voyager VI. I give a little wistful chuckle every time I see that (which isn't often; honestly, it's a pretty crappy movie. Stick with the even-numbered ones). Voyager Six is next to Apollo Twenty in the might-have-been files.
For what it’s worth, the New Horizons mission to Pluto cost $650 million, of which $200-$250 million was the launch cost. That $650 million includes operations from Phase A studies in 2001 through Pluto encounter in 2015 and post-encounter data analysis. Either way, that’s just over half a billion spent over the course of 15 years, versus $0.5-$1 billion per shuttle launch.
IIRC, satellites on the shuttle do not incur a launch cost, as an incentive to launch from the shuttle instead of unmanned boosters. Maybe that policy has changed post-Columbia, though.