At $work we run a raytracing renderfarm on AWS spot instances that are only spun up as needed. This is far cheaper than keeping the same amount of physical mach

At $work we run a raytracing renderfarm on AWS spot instances that are only spun up as needed. This is far cheaper than keeping the same amount of physical machines around to serve peak use. The caveat is that we could most likely do with a smaller pool of machines and queue up jobs over night yielding higher utilization. So the price advantage only exists when the lower time-to-completion is a requirement.