Seems like it'd be worthwhile to trial a real computer of similar spec to what he's wanting vs. a virtual server, to see relative speed and relative cost. You can get some pretty stonking powerful VPS instances if you're willing to pay for them, and with Amazon you can optimize for different workloads such as GPU, CPU, or fast IO.
There can't be that many projects on CVS who are sticking with it only because they're waiting for someone else to manage the switch to git, and you'd need a fair number to make this worthwhile, IMO. I think he may be tilting at windmills.