Bandwidth isn't an issue.
Compared to the Playstation 3 the WiiU can compress textures at a far greater ratio, compounded by the fact it will compress light maps, shadow maps, HDR textures and material properties, normal maps and bump maps just to name a few in hardware.
Thus by extension more data can be shifted if the Wii U had the same amount of bandwidth as a Playstation 3.
It will also mean that the WiiU is more thrifty with it's Ram, meaning every Megabyte will go that much farther.

It also means even more detailed texturing as higher quality assets can fit into a smaller area.

The eDRAM will assist in working as a cache if developers see fit to alleviate some of the bandwidth constraints, which to be honest is more like a bandaid measure, but does have some big bonuses on the CPU side.

Game design has everything to do with graphics, you are talking more about game mechanics, the graphic style was a part of a bigger game design they want to tailor this game with so things all fit together while considering the power of the Wii U hardware.

Nintendo's game design continues to be way beyond other first party studios about 95% of the time, and that's why Mario Kart games are always fun as shit and fun to look at no matter the iteration.

We also need to stop talking about RAM, it's first of all, fucking ridiculous to compare the PS3 or 360 memory architecture to the Wii U because the Wii U has a newer design so real world performance when done right would be faster and more avaliable than either of those platforms by far.

Also, people saying shit like this game can run on the PS3 or 360(not you,) guess what, PS4 and X1 games can also run on the PS3 or 360, but at what FPS? You can always achieve something in software rendering if hardware processing is not there, at like .5 FPS or less, what the fuck is the point of saying this game can run on older consoles is beyond me. This game does 60FPS with split screen mode and 30FPS with 4 player split screen mode, what that means is that if all they cared about was graphics without the performance at 720P 30FPS being a SP only game, they could have bumped up the graphics too, but guess what? GAME DESIGN! Durrr!

Not so much laziness for the most part, more like poorly funded, can't do the work without money or proper directions from the higher ups...... Takes time to write new code man.

It looks so good it hurts!

You'll have to excuse the graphics programmer in me that notices these things.

I watched this scene in the trailer video.  You can tell the reflection maps on metal Mario and on vehicles are static even while they drive around this track which should have many lights reflecting from the ground and the scenery.  It looks like it's using  static cube maps.

So how does that stack as Gen 8 technology?

In Gen 5 games used static environment maps.  It makes things look shiny and like they reflect the sky, but they do not update to reflect other cars, trees, or other scenery dynamically.

In Gen 6 games could do more detailed enviroment maps.  The original Xbox had native cube maps.  A basic cubemap effect looks like this:

Poor man's reflection would still use a static cube map, but many games rendered enviroment maps in real time, so you could see buildings, trees, lights reflect as you pass them.   This is much more intensive as it requires rendering up to 6 views of the world.

Since this is an expensive operation, a game might produce a single cube map shared between all cars on screen for example, so the reflections are not very accurate but make a good effect.

In Gen 7 the basic formula of using cube maps did not change, but resolutions increased, and shader complexity increased to produce more realistic metalic paint and other materials.

In Gen 8 games like Killzone Shadowfall utilize multiple cube maps simultaneously, and screen-space raytracing effects.  This allows reflections to really look connected to their source all over the screen.

And then in MK8 we are... back to doing static cube maps, somewhere around poor mans Gen 6 reflections.

Hmmm.  It's not like it couldn't be better since Gen 7 consoles can do better.  Lack of time, or just they thought it's good enough?  Could be better by launch?

useless trivia for you, in GT4 each track had several cubemaps based around the track itself that the engine would pop in and out of depending on where the car was located on the track, it was carefully timed and positioned to give minimal jump when switching maps on the back and front window but the side windows sometimes showed off the effect.

https://www.youtube.com/watch?v=wW67x0Sh00M

Video uploader mistakes this for realtime reflections even though it wasnt really, but it does show the jump between cube maps - in addition to these track based cube maps a secondary overlay cubemap was used to reflect track side lights, this was actually psuedo realtime using shaders, but not as accurate as actual realtime reflections, the secondary reflections were used over the entire car.

I thought it was real time.  The disappearing geometry could be due to distance pop-in, no?  They rendered just very near objects - also very simplified geometry compared to the full track.  Lots of cheating to do reflections and shadows on that game...  Anyway that's what I thought, but cool vid from the past .

It doesnt need extra RAM. PS3 already has games that blow MK 8 out of the water, both exclusives and multiplatforms. MK8 is a cartooney linear racer after all. There are PS3 games from 2007 that even look comparable like this (GT5P and Ratchet and Clank TOD)

Console graphics are full of compromises.
Lets say you stick to statice cube maps, by doing so you are freeing up rendering power for something like a Bloom effect, this is a common compromise that occurs all the time.

Look at the jump between Halo 3 and Reach, Bungie removed the Tessellated water, triple buffering and introduced impostering and texture streaming from disc, the result of that was they could bolster the draw distance, increase texture resolution, throw more objects on screen etc'.

The end result is, whilst they sacrificed graphics in one area, they could bolster it in another, providing an overall much larger increase in perceived image quality.

Still, could be worse, we could still be using sphere mapping. :P

common misconception because of the almost-realtime lighting effecting the rest of the car, see doing a realtime reflection of 2d surfaces (lights) is nowhere near as costly back in the day as reflecting geometry, so a series of low resolution cube maps were created per track for that purpose.

some areas used hexagonal maps with no upper and lower faces (just the 8 side faces), especially in tunnels or on night tracks where more detail is shown on the horitontal view than vertical, this basically meant it was still a cube map but with the upper and lower faces shifted to be used as side faces.

Average track would have 50 or so 128x64, 64x128, 64x64, 32x32, 64x32, or 32x64 cube maps for the record, 2d light reflection was 256x128 for night tracks (allowed since tracks generally had less geometry and detail, but more light sources), and 128x64 or 128x128 on day tracks depending on track type.

Useless trivia #2, GT4 had the ability to get punctures and blow your engine, complete with spark particles on the bare rim until you pitstopped, it was removed before prologue due to the ps2 not being powerful enough to handle several cars emitting sparks, or smoke particles for blown engines, it never got revisited even for the final version or subsequent games for the PS3.