Fwiw, this is pretty much exactly the situation we had from the very start, since the early days before "big data" was even a thing people talked about, and why I went "bare metal" from the beginning.
At the time (about 2005 iirc) I was read/writing between 500GB and 2TB a day, with heavy calculations across two high performance desktops. Then as I came to scale up AWS was announced, but when I priced it, just what was running already on those desktops came to more for the first month than buying 6 heavy duty dual CPU Dell servers with tons of storage.
So that's exactly how we started.
The only "shock" I got was how much of a b!tch keeping all that cool is, and that took a year or two to get that "just right".
At the time (about 2005 iirc) I was read/writing between 500GB and 2TB a day, with heavy calculations across two high performance desktops. Then as I came to scale up AWS was announced, but when I priced it, just what was running already on those desktops came to more for the first month than buying 6 heavy duty dual CPU Dell servers with tons of storage.
So that's exactly how we started.
The only "shock" I got was how much of a b!tch keeping all that cool is, and that took a year or two to get that "just right".