It's not that it's hard, it's just the logistics of running the virtual machines(s) and the applications they contain. You could use a minimal linux distro like Alpine or Tiny Core, but you still need to run applications on top of that if you're testing or developing.

Spinning up a basic devstack instance (for example) take a minimum of 6GB and that's before you even deploy any test vms inside that infrastructure. Another example, if you're doing config management development you may need several VMs running which in turn may have (say) large java apps with heavy memory requirements even when fairly unladen. So, I guess the answer is, it depends on what you're doing and what the memory requirements of the thing you're running on VMs is.

Hyper-V has dynamic memory, so memory resources can be reallocated as needed, and has driver hooks so that linux vms can be resized too. There's also Intel's clear containers push which virtualizes for linux but shares a lot of kernel structures between the host and the VM.

>It seems like a really big opportunity, even if it's really hard. It's not because it's a problem that is easily solvable by spending a small amount on better hardware. 16GB RAM costs $80 which is cheap if you're only going to use it for VMs.

Unikernels are one area of research, http://unikernel.org

Yeah if someone worked out how to use less memory in a VM we could use it in normal machines too ;)