The material cost may be much lower, but the machining cost of running everything through milling is dramatically higher.
Sheet Aluminium (and Steel) can be stamped, punched, spot-welded, bent etc. in a few fixed stages in a 'run'; i.e. you run a few hundred sheets through the same stamping operation in quick succession, then run those same few hundred processed sheets through another operation (e.g. bending) in quick succession, and so on. If you need multiple stamping operations, you can even use the same machine and reconfigure it once between runs.
If you needed to mill something, even if the actual milling step could be fully automated (i.e. your milling does not require someone to walk up to the machine, remove the part, flip it over, re-mouint it, and re-centre the machine) you can only process one part per machine at a time, and the actual processing take a LOT longer.
There are factories that use milling for mass-production by separating the milling stages and having a dedicated machine for each stage, but this means you can only ever build one part with that factory, which cost you a few million dollars to set up (better hope you're going to sell a several tens of millions of units!)
For cheap plastic-forming, injection moulding is king. But the startup costs of building the durable metal injection mould (which even with the best flow simulation may need one or more prototypes to get right) are so high that you need to guarantee that you are building a few tens to a few hundreds of thousands of pieces for it to be worthwhile.
3D printing might end up as a medium method, cheaper than milling but more expensive than injection-moulding, but at least for the moment it's still only marginally cheaper than milling due to the build-one-thing-at-a-time issue. There have been proposals to design SLS (Selective Laser Sintering) and FMD (Fused Material Deposition) machines with multiple parallel 'print heads' to allow a machine to build many parts at a time with one set of drive electronics and motion hardware, but so far I know of no company that has actually done so, let alone used it in production. There's also the issue that the surface finish of 3D printed parts isn't as good as moulded or milled parts yet, so you're limited to either internal parts only or adding an extra surface-finishing step e.g. tumble-polishing (which may break parts), chemical finishing (which will reduce the outer dimensions so needs to be compensated for in design), surface coating (e.g. painting with a thick paint) or finishing by hand.
