I think continuous fast-paced innovation in an industry usually creates relatively low barriers to entry as perpetual innovation cycles “self-correct” potential monopolies.
If the market creates an innovation engine, then participants in the market will compete over scraps. Unix, the Windows application environment, Apple and Android marketplaces, consumer fashion.
If a company creates an innovation engine, then it can ride a stream of development that gives it a persistent advantage: 3M, Edison, Apple, Google, Intel. Possibly IBM (though I'd count its success on other factors, despite a few periods of significant innovation).
Contrast this with companies which created one big idea and camped on it for a long time: AT&T, Xerox, Boeing, Microsoft, Comcast. These have essentially leveraged economies of scale, regulatory environments, and monopolies to capture and retain a market for an extended period of time during which little if any real innovation happened. Oftentimes these companies actively discouraged significant new innovation (AT&T and IP telephony, Boeing and high-speed rail, Microsoft and OS alternatives from DR-DOS to OS2 to Novell to Linux, Comcast with municipal broadband).
It depends on whether or not your goal is market dominance or advancing overall conditions. I'd like to see an innovation-based strategy encouraged, and demand-side benefits of scale doesn't deliver on this.
Microsoft is a business built entirely on network effects -- both their operating system business (everyone developed for Windows because it had all the users; it had all of the users because it had the most applications) and their office suite business (Office was taught in community colleges because it made you desirable to the maximum number of employers; businesses standardized on it because it was easy to hire workers that already knew it, and because they could exchange files with other businesses.)
Microsoft enriched itself enormously through network effects -- and I would argue, enriched society as well, compared to an alternate timeline where operating systems and application software were fragmented and software and skills were not portable. Though I do wish they'd done a better job with the win32 API.