IIRC, the rumor didn't say 3G for the OS exclusively, but for "other features". I don't think either that a console OS would require as much, unless they would want to emulate a full PC. And I have no idea what other features could require them too, but that's the exciting part of the rumor.
MS has shown with the 360 that they can fit a small and efficient OS in a small amount of memory, and it's been one of the obvious strengths of the console. I can't imagine they would make a bloated OS for the next one, unless they have very good reasons to do so.
I'm sure they have a great reason for it, but is that reason tied to video games, or to something else?
First you have to assume they want Kinect available to every game without a performance hit, plus Kinect as an alternative to a traditional remote. That means Kinect needs it's own chunk of memory assigned specifically for it so that you don't see a graphical divide between Kinect and non-Kinect titles.
Second, you have the need to run DVR functionality that has been strongly rumored for a long time. Sony has a very efficient implementation of this on the PS3 already, but there is no guarantee that MS' implementation is as efficient, or that it's just basic DVR service, it could be something with expanded functionality.
Then you get into questions about the OS kernel. At this point I think it's highly likely that it'll be Windows 8 based. Consider MS' current overall corporate direction. Why would they fund Xbox 720 apps that can't also run as Windows 8 apps? Therefore the Xbox 720 needs at least cross compatibility of apps with Windows 8, and emulating that is less efficient than just building the OS off Windows 8. At that point the question is just how much Windows 8 are they putting in the box.
The rumored "display planes" feature also sounds like a real memory hog to me. My interpretation of that is MS' desire to use them for seamless plane switching. For example, you're playing Halo 5 online with your pals and one of them mentions a great new youtube video he saw yesterday. Instead of having to wait to view it after playing you can simply flip to a different, already running, web browsing plane and view the video while waiting for the next round to load.
Or lets say you're watching a blu-ray while your spouse/sibling/etc. wants to watch a video on the DVR. The Xbox 720 can without a hitch stream a pre-scaled version of the DVR'ed media to a compatible Windows 8 device.
The ultimate culmination of this that makes MS need to reserve a lot of assets is the power user. Someone who starts out watching a blu-ray, gets a friends list invite to play Halo 5 online, and during rounds flips over to the live broadcast football game that is running through the Xbox. You don't have to even close the Blu-ray if you buy Halo 5 via Xbox Live Arcade, instead it's progress is paused in the background waiting for you to flip back via a simple Kinect-sensitive snap of the wrist.
Picture the quick flipping you do on a tablet in high resolution with the Xbox 720 keeping all video at the current display setting for your TV regardless of it's native resolution (removing the need for those brief display refreshes), but instead of using a finger you use your hand and instead of flipping to angry birds or a web video that needs to buffer before being ready it's blu-ray movies, digital download movies, your whole music collection, your DVR, and all of your games. Just one flip and everything is there - instant gratification from all of your media.
All of that would eat up a lot of resources, and MS will obviously want some buffer to future proof themselves.