You're only seeing the winners. An ill-defined software product (which almost all software startups are), is similar to scientific exploration - you start with an idea, but you have no idea going in how much effort (months, years, decades) it will take to realize that idea, or even if its possible at all. After all that, what you end up producing may be not what customers want at all. In the time you took producing your new innovative software product, the market could have moved on.
The immediate example that comes to mind is "Cuil," - a search engine (large market) started by ex-googlers (great team), with $5 million investment (well-funded). It took them forever to launch (because its a hard problem), and partly because of that, they got poor PR and little traction. They've got another $25 million investment since then, and almost nothing to show for it.
Crunchbase says:
"Husband and wife team Tom Costello and Anna Patterson were joined by Russell Power. Patterson and Power are ex-Google search experts"
Anyway I don't want to cast aspersions too much on Cuil, because they may turn out to be successful eventually. My point was that software is a hugely risky business, with massive upfront costs.
You can build stuff cheaply of course, but when you do that, you have the problem of dealing with competition - namely if you can build it cheaply, others can too.
The immediate example that comes to mind is "Cuil," - a search engine (large market) started by ex-googlers (great team), with $5 million investment (well-funded). It took them forever to launch (because its a hard problem), and partly because of that, they got poor PR and little traction. They've got another $25 million investment since then, and almost nothing to show for it.