Writing software is ultimately an artistic and creative activity. It requires imagination, analysis and creative insight to produce elegant code. Not much creativity can come out of a group bogged down in commercial maintenance and patching, or under tight deadlines to deliver a product. Such people engage more in just hacking code together to fill the holes in the leaky OS made at Redmond or adding glitter to the long delayed Vista, rather than producing true leaps in technology.
Most creative and ground-breaking work is done in start-ups which eventually get acquired by larger corporations which are in what I would call maintenance mode with their existing code-base. SeaDragon the maker of PhotoSynth , now part of Microsoft labs is one of favourite illustrations for this argument.
Other than a start-up exciting new work may be done by individuals working on various Open Source projects. Microsoft has been trying to foster this with Codeplex and Shared Source. It has already borne fruit, IronPython is being adopted my Microsoft as an extension scripting language. I believe Microsoft also uses Codeplex as a recruiting ground.
Though not very accurate Ohloh offers some very insightful metrics on Open Source projects. The valuations are quite revealing. Worldwind is valued at 6.9million, GDAL is valued at 8.1million and the Linux Kernel is valued at a whooping 87million. Some companies may have this kind of money off-hand but it will save them a lot of time to exploit or bolster an existing solid codebase than make their own from scratch.