[ajkjk]

I find the idea that mass and energy are equivalent to information in some way plausible, personally. Just, this paper doesn't offer anything useful on the subject.

For the case of erasing a 1TB storage, I think it's fine to assume an abstract operation with no error correction or internal state. But even still, any wear-and-tear to the drive counts as data as well, and that's going to be incredibly hard to quantify. But if you got past all that, then the idea is that the quantum state of the <the drive + the erasing system>, if an isolated system, has to put that 1TB of entropy somewhere (since in principle the erasing operation must be reversible), and that it presumably becomes decoherent thermal noise. It's 10^-25 kg of energy _in the form of heat_, not anything useful like discrete photons.