First off, the Metro implementation on Server is not the same as it is on Desktop (aka "Client" version). Not that the theme or platform services are different, but that they support a different set of features layered on top of them. Namely, the applets or utilities that are provided for configuring, troubleshooting and maintaining a "server" as opposed to a "client". In my opinion, for the scope of features provided within the "server" paradigm on top of the Metro UX, it is a MUCH better fit. This almost certainly sounds absurd to say (and for you to read it, I'm sure). "A Metro interface on a 'Server' operating system?" Yes! It actually works. But here's the irony...
Metro is a better fit on "server" than on "Windows 8" by a factor of a Gazillion-Trillion-Billion to One.
I'm sorry to Windows 8 fans, but on the standard "desktop" configuration, I'm not a fan of Metro. If I had a tablet on which to put it through some meaningful paces, I might have a very different opinion of it. But on a traditional desktop or laptop, I still believe that the tile concept is *NOT* the most efficient UX construct for using a mouse and keyboard. It is more efficient for hand and finger gestures. The tiles are scaled on a factor that more ideally matches the scale and topography for direct hand movements. A mouse is an intermediary instrument that re-scales movement and articulation such that large tiles are actually inversely proportionate to the scale of movement. In English: From a purely engineering perspective: it is not efficient.
I relate this to comparing the handling of sugar cubes with chop sticks as opposed to using them to handle grapefruit or softballs. At some point, the relative size ratio, and weight, make it more efficient, and convenient to use bare hands.
Those are my own words.