video games and arcade games use OGG (some MP3 or proprietary) compression all the time. You can fit an entire CDs worth of streamed BGM audio tracks (60 mins+) in under 100mb. More minutes if you boost track compression. And mono sfx don't require a lot of space to sound great. You can vary your sfx between 8k to 22k depending on what the data actually is and how it sounds. There really is little reason for these games not to have stereo music and panning SFX now.
As an example- Haunted Cruise uses 100 or so panning mono sfx (some are stereo for canned animation effects) and about 30 mins of OGG stereo music. You cannot tell there is audio compression with all the overlaying music/sfx/mechanical sounds going on. Everyone who's played it comments how awesome the sound is. The entire game package is ~210MB in size- BUT that includes the unity3d runtime, lots of large 1024x1024+ textures, 3D models/animations, and the game script.
Assuming stern dmd pixels are 4bpp- all the pixel art (properly lossless compressed) could easily fit in the space of 1 4096x2048 32bpp texture thus leaving ample room for game code, core/engine code, and audio (compressed and raw).
Now that stern is entering into real coloured video (lcd), they will need to support some form of streamed video compression.
Also, I use OGG as an example here only because it's a free algorithm. Last I looked, MP3 requires licensing fees to use.