Pixar's render farm has run on Linux for quite a few years now. I wouldn't be surprised if many of the design workstations are Macs, but the heavy lifting occurs on large Linux clusters. See:
http://blogs.computerworld.com/if_it...ects_its_linux
The ability to create large clusters using commodity hardware has been a big sell for Linux at places like Google as well. In fact, I once read that Google's search cluster is so large that they don't bother to replace machines that die for quite a while. It costs more to track down the dead server than it does to throw another one in the rack.
Let me give an alternative view to hobbes_fan's criteria list, especially as it applies to business computing.
*ease of use
I generally see two types of (non-technical) people using computers. One group, and it's a large one, consists of people for whom no computer or OS offers "ease of use." One of my clients describes some of his office workers as people who would say their computers became unworkable because you moved an icon on the desktop. For these people, any change is unwelcome, even upgrades between Windows versions. For businesses forced to upgrade Windows by decisions in Redmond, you'll still need to invest in retraining people like these.
The other group is much more comfortable with computers and are perfectly willing to muck around in some menus to figure out what they need to do. For them the choice of an operating system is really not that important if it supports the functionality they seek.
*functionality
As I said earlier, most mainstream users don't need more than perhaps a dozen or so applications. Even for people sticking with Windows, the array of FOSS applications like Firefox and OpenOffice make it possible to leave expensive closed-source applications behind. Businesses face the problem of proprietary software that appears to run only under Windows (though Wine and Mono may actually be sufficient even in these cases). In another ten years, though, many of these applications will have migrated to the web, either through third-party providers like Google Apps, or running on local servers. I always argue in favor of web-based solutions in the office environment when possible rather than having to support some proprietary application on every desktop. Commercial vendors in general would prefer the opposite solution since it provides a built-in requirement for support and upgrades. I find this especially infuriating for applications that consist largely of database access where a server running Apache and PHP could provide the entire user interface. Instead you're given some VB or .NET application that has to be installed everywhere and maintained by the provider.
I see more and more people moving back in the direction of thin clients and distributed computing to save money and to enable centralized management of resources. If everyone's running the same copy of OpenOffice, upgrading and troubleshooting takes place on the server not on hundreds of scattered desktops. Those of us with long histories of using computers are amused by this trend which harkens back to the mainframe and mini-computer days of yore.
*support
Last I heard, Microsoft provides essentially no support for Windows or its desktop applications; you're told to ask your computer's manufacturer. If you have a Windows problem, you call Dell, not Microsoft.
Also, based again more on reading and discussions than recent personal experience, professional support for things like Exchange come at a high price. A single support call to Microsoft costs a minimum of $175, I believe. Businesses usually rely on third-party system integrators to provide this type of support. Linux system integrators aren't that big a business yet (sadly for me), but you can get professional service contracts for Linux from companies like IBM, RedHat, and Novell. RH is known for writing bug fixes into the Linux or application source code in response to customer concerns then submitting them upstream to the project's maintainers.
and finally price
Price isn't always at the bottom of everyone's list. One of my clients is a community health center facing substantial budget cuts as a result of the economic crisis. We're moving to Linux-based thin clients and probably a
Linux terminal server or two to support the computing needs of the medical staff.
The true cost of Windows is, I believe, much higher than the true cost of Linux, even ignoring the cost of the software itself. Most studies show that Linux installations require fewer support staff than Windows installations. Now part of this difference might be that the average Linux administrator is probably more knowledgeable than the average Windows administrator, but a lot of it has to do with the greater reliability and security of Linux systems. If you can eliminate two or three support staff members at an average cost of $80-100K or so with salary, benefits, overheads, etc., you're looking at numbers that might attract some attention from the folks over in accounting.
None of these arguments says much about what ordinary people should be running. I fully expect Microsoft to maintain its control over computing in the home for many years to come. There are obvious benefits to home users in staying within the Windows monoculture because of its familiarity and because, as WK notes, it's really all they know. Businesses have different incentives and different resources, and it's there that I think we'll continue to see the expansion of Linux both in the server room and on the desktop. I'm not expecting these changes to happen quickly though. If Linux achieves a 20% desktop share in businesses a decade from now, that will represent a major achievement.