I wrote Predicting the future: Google vs. Microsoft a few days ago and I enjoyed that exercise enough to try again with Linux vs. Windows.
I have been using Linux since downloading the core of Slackware using a 2400 baud modem and I wrote a commercial application for Windows 1.0. Unlike a lot of Linux enthusiasts, I am fairly happy with Windows dominance in the non-technical world: this makes Windows more of a target for virus attacks, trojans, etc. and makes using Linux and OS X safer. I once spent the weekend at a friend's house and Microsoft's .Net project manager was also there with his family. He told me that I was not Microsoft's target customer - and that suits me fine. I do wish that Microsoft would do the world a favor and support open file formats, but that is their bad.
Windows: Although I believe that Microsoft's desktop dominance and financial position will take a long slow slide down hill, they will be around for a long time. I think that Microsoft's future lies in the idea of charging customers a yearly fee per PC for software use in order to provide a continual revenue source without having to spoil the quality of their products with gratuitous changes in order to get people to upgrade.
This is the scenario that I would like to see: I have two PCs that I dual boot between Linux and Windows. I would prefer to pay Microsoft a (hopefully) small fee per PC per year to get security updates and hopefully rare functionality modifications. This subscription model would allow Microsoft to concentrate on quality and not on new features that a lot of people don't really care about. I would consider a $25/year fee for each of my PCs reasonable. If you buy a new PC, the PC manufacturer would pay the first year fee for you (actually you are paying via a higher purchase price). For individuals who need Microsoft office, I would think that a $50/year fee would be reasonable. I would expect Microsoft to give corporate customers a discount. I would also expect Microsoft to put more effort into making company wide deployment and maintenance of their software products less expensive in IT staff resources.
Linux: I think that the use of Linux will continue to increase in business with a much slower growth in the home desktop market. I think that professional maintenance and work on Linux will continue: large corporations will donate people out of self interest and home/small business Linux users like myself will continue making small financial donations to Debian, Ubuntu, etc. to support continued high quality distributions and improvements. I expect to see Linux Desktop use to continue to cluster more around both Gnome and KDE; alternative desktops will mostly be used for hobbiest or emotional reasons. While the competition between Gnnome and KDE is healthy, I expect to see more distributions smoothly supporting running KDE applications under Gnome, and vice versa.
I think that it will be a long time before Linux scales up to massively SMP servers like Solaris, AIX, etc. I think that for most web based applications running Linux on inexpensive yet fairly reliable hardware will continue dominate the low and middle parts of the server market, slowly but surely increasing market share.