Actually, I think it’s the opposite. In the olden days every single DC had to have raised floors… IBM mainframes until rather recently would only accept intake air from underneath. These days, every piece of DC equipment is front intake, so there’s less compelling reason for the expense of a raised floor.
There’s also big, big wiring differences… In the olden days, computers would be hooked up to a LOT of cables. Serial cables, coax cables, all manner of proprietary equipment cables, etc., and having some of those under the floor, while others run over-head, made a lot of sense. Now… how many cables do you need going into your DC cage? A single internet fiber and a single cross-connect could be faster and can do everything you could possible need…
As for wiring… if you lease a cage in a proper shared data center, the floor tiles are going to be bolted down and require an appointment for staff to open them up for you. So top exit holes in racks and cable ladders are still very much a thing.
The most likely reason for top-of-rack switches was the assumption that many of the connections are not for use in the local rack, but instead will go out the top, so why waste the cable manager space for those?
If that’s not the case, and the switches are mostly used for in-rack connections, right in the middle of the rack is the best spot… minimizing cable length, reducing the size of cable bundles in the vertical cable management by up to half, etc.
I’ve previously shared my strong (educated) opinions on how rack cable management should be done: