On March 13, 2004, the Defense Advanced Research Projects Agency held the first DARPA Grand Challenge, a race for fully autonomous ground vehicles across the Mojave Desert. DARPA’s own account is blunt: “No team entry successfully completed the designated DARPA Grand Challenge route” from Barstow, California, to Primm, Nevada, and so the prize “went unclaimed.” The $1 million on offer drew 15 finalists out of a larger qualifying field at the California Speedway, but the desert beat all of them.
The best any vehicle managed was roughly 7.5 miles before getting stuck, a tiny fraction of the more than 100-mile course. By the usual measure of a race, the 2004 event was a failure: there was no winner, no finish, and no obvious sign that self-driving cars were close.
That framing misses what actually happened. The challenge had pulled university and hobbyist teams into a single hard problem with a clear, public scoreboard, and it exposed exactly where perception, planning, and control broke down on real terrain. DARPA ran it again just over a year later, and that time five vehicles finished. The 2004 washout is the canonical example of a productive failure: a prize-driven competition that looked like a flop on the day but seeded a field.
For a business reader, the lesson is about how breakthroughs are funded. A well-designed competition with a concrete goal can compress years of scattered research into a focused sprint, and a first attempt that fails completely can still be the event that makes the second attempt succeed.