I remember, back when Instant was first ramping up staffing from prototype to production levels, saying,
"It's going to be a spectacular failure, but I am very, very glad that I work in a company where we can take risks like this."
(In my defense, remember that chart Google released today showing the resource usage of various incarnations of Instant? It was in the Prototype stage at the time, with a total cost roughly 10-20x existing websearch. We were talking about having to build new datacenters to support it. And the UI was nowhere near as polished; I found it very jarring to use at the time.)
I think there's a lesson for entrepreneurs in here somewhere: imagine where the technology could be, not where it is. Almost all innovations are pretty crappy when they first come out. That's a reason to improve them, not a reason to cancel them. Many problems that look impossible at first can be solved given enough engineering effort and creative thinking.
I think it also says a lot about the management style of Larry & Sergei vs. the MBAs at Yahoo. There's a tendency, as companies get bigger, to look at things rationally and say "We have $X to lose if this fails, and some uncertain $Y to gain if it succeeds. The chance of success is low. Let's not do it." Larry & Sergei say "Okay, we know what the risks are. $X is high but manageable, and $Y is unknown but potentially big. Let's do it."