A single iron atom can be a magnetic domain. So a surface coated with single-atom domains, spaced a few atoms apart. I would posit that's close to the 2D limit because of physics. They can't be directly next to each other or it's impossible to read or write them.
So very roughly, about 1 bit per square nanometre. Which unless I'm dropping an order of magnitude (very possible) is about 10 petabits per square centimetre, and with about 300 square centimetres for a 3.5" platter that's 3 exabits or so per side of platter.
Whether it will ever be possible to actually read or magnetize domains that small without interfering with the neighbouring domains is the question and no one knows. There have been several breakthroughs, like perpendicular recording, that have brought us much closer to the theoretical limit above, than anyone would have thought.
So very roughly, about 1 bit per square nanometre. Which unless I'm dropping an order of magnitude (very possible) is about 10 petabits per square centimetre, and with about 300 square centimetres for a 3.5" platter that's 3 exabits or so per side of platter.
Whether it will ever be possible to actually read or magnetize domains that small without interfering with the neighbouring domains is the question and no one knows. There have been several breakthroughs, like perpendicular recording, that have brought us much closer to the theoretical limit above, than anyone would have thought.