"Pros" aren't a monolithic group, and it's thus often dangerous to make such generalizations.
At one point I used a PowerMac G5 for my research. I purchased the machine with enough RAM for the calculations I anticipated I'd need to make, and that worked fine for a couple of years. But when we decided to simulate a different class of biopolymers (something that couldn't have been readily forseen when the work started), I needed more RAM. Because the RAM was expandable, I was able to upgrade it to meet our new needs.
[The machine was used for development work—I'd make sure it the code was working on a few sample biopolymers, and then send our jobs to the university's clusters.]
University labs, especially in the physical sciences (as opposed to those being supported by large NIH grants), often have tight budgets, and we could not have afforded to buy a new machine just because our RAM needs changed. Besides, since we didn't need a new machine, just more RAM, that would have been wasteful.