It's both. Cross section begets power, but surface area begets temperature rise. So, based on pure physics, assuming we want a given limiting temp rise -- it depends. It's not a straight proportion to either quantity, nor is it a straightforward 3/2 or whatever power law. (Well, it's probably close to a 3/2 power law, but it'll be some odd exponent near there, that comes from the convection rate as well as these factors.)
We normally assume fixed current density, i.e., cross section proportional to rating. But this does overestimate capacity at high currents, and under- at low.
A full model isn't easy, because it depends on the convection rate (or conduction for that matter, when potted). Which is nonlinear, and depends on enclosure (conduit? equipment? free air?) and orientation.
Plates have the advantage that cross section is just, whatever size you cut it to, it's very flexible that way; and, they have lots of surface area. They can also be arranged broad-facing to lower inductance, in circuits where that matters (industrial converters, pulse generators).
Or board-level SMPS for that matter, since PCBs are arguably just made with very thin plate, and benefit from the low inductance and modest dissipation, hah.
The surest answer is: check the NEC or whatever local code, and follow that. I mean -- this isn't even a matter of physics, it's simply the law!
Or if it's for equipment rather than facility wiring, this is still a fine starting point. You can shave it down from there, as long as it's not a fire hazard (passes UL/etc.), and it meets functional requirements (temp rise within some reasonably acceptable range, voltage drop / power dissipation acceptable).
Equipment still isn't a free-for-all, like, UL will test conductors connected to mains wiring, for certain overload conditions (with the fuse bypassed as part of the test!). Usually you'll have to skimp quite severely to run afoul of this. Standard wire sizes, connectors of approved ratings, should be fine. Like, maybe the most likely case is PCB traces burning off, because, well, they're just thin foil, and sometimes you can't afford much more width than is needed for nominal current rating only.
Tim