Sony and Microsoft wanted to spend about the same amount of silicone on the APU (GPU/CPU). Microsoft needed a whole bunch of RAM from the get go to achieve their vision of the XBONE with all it's OS's and multitasking. That meant from the very beginning they had to target 8GB of DDR3. DDR4 is too far away and 8Gb of GDDR5 was too expensive and complicated when they started planning the XBONE.
Because of the slow speed of DDR3, they would need to dedicate some silicone on their APU to some faster eSRAM (something consoles have had for ages). Because they needed to dedicate some silicone to eSRAM, they had to lose some silicone from either the CPU or the GPU, since the CPU is tiny anyway, it was simpler to take it from the GPU.
If the GPU was bigger, then they could probably have used more RAM specifically for it, but it isn't because they had to have some eSRAM. It kind of balances out. More RAM might have been useless without more GPU.
Sony on the other hand wanted a single dedicated pool of very fast RAM and so targeted GDDR5 from the start, they didn't want to do lots of multitasking and other stuff but knew that RAM densities would probably go up over time before launch allowing them to drop more ram in without changing anything else.
But it is expensive, so they were going with 2GB spread across 16 chips, likely 8 on each side of the motherboard in "clamshell" mode, a feature only GDDR5 has, saving room and complexity in motherboard design. Because they don't need any eSRAM to make up for slow main RAM speeds, they now have all that silicone left for more GPU. They spent it.
Since the original target spec, GDDR5 ram densities have doubled and then doubled again. Something that is very beneficial to Sony, but not something Microsoft could have taken a gamble on because they needed 8GB from the get go.
So, Microsoft have been fairly locked in to the plan from the start, because they wanted lots of RAM. Sony were happy to have a smaller chunk, knowing they would probably get to at least 4GB before launch. As it turned out Sony managed to get 8GB of very fast RAM and there was literally nothing Microsoft could do to combat it.
For Microsoft to have then added more RAM or change to GDDR5 and drop the eSRAM would have taken motherboard redesigns, controller redesigns and all sorts of stuff that would take months and months. All Sony had to do to use more RAM was buy the higher density chips.