I'm suspicious of the author's actual experience.

The fact that scientific computing has a different pattern than the typical web app is actually a good thing. If you can architect large batch jobs to use spot instances, it's 50-80% cheaper.

Also this bit: "you can keep your servers at 100% utilization by maintaining a queue of requested jobs" isn't true in practice. The pattern of research is the work normally comes in waves. You'll want to train a new model or run a number of large simulations. And then there will be periods of tweaking and work on other parts. And then more need for a lot of training. Yes, you can always find work to put on a cluster to keep it >90% utilization, but if it can be elastic (and has compute has budget attached to it), it will rise and fall.