This is not actually where overhead really goes to. A lot of overhead simply pays for departments that don't bring in money (eg humanities, some social sciences) and I think that's fine too. But let's not say that an experimental lab with ~10 people and ~2000sqft that pays >$1mm a year in overhead spends it on "office space, internet, electricity, utilities..." because that is disingenuous (btw those numbers reflect the reality of the lab i got my phd in). Most overhead comes from expensive experimental labs (hard science + engineering + medicine).

Expected that there is some overhead to keep lights on, sure. But should lab A who brought in a 1 million dollar grant pay 500k in overhead for internet and electricity while lab B in the next bay in the same room brought in a 50k grant and only needs to pay 25k? Overhead is baked into salary budgeting too, you might budget 60k for your technician in your grant, knowing they'd receive 40k, the 20k allegedly covering the employees various benefits (which said employee will find are withheld from their 40k paychecks anyway).