[nostrademons]

Depends on era, sub-industry, and to some extent location.

I think that's been largely true of startups that have succeeded since 2004, particularly the latest crop of SF-based startups. These have largely been about the deployment of existing technologies developed at big companies (the WWW, e-commerce, and mobile phones) into new industries.

But I had a mentor, back when I started my career in Boston, that said "Every startup I've seen succeed did so because some engineer did something that everybody else said was impossible." She was an early engineer and later VP at Stratus (the early-80s company that made super-redundant minicomputers for banks etc.), then later was called in to rescue Equalogic's software (multi-terabyte transactional SANs) before it was bought by Dell. There are plenty of other startups that succeeded because they did something technologically that most people thought was impossible or at least economically inadvisable - Google, VMWare, SpaceX, etc.