Quantum issues are funny. If it weren’t for quantum, things like diodes and transisters (at least Metal Oxide Silicon Field Effect Transistors) wouldn’t work in the first place. Nand Flash memory, as I understand it, relies on quantum effects to work.

But the problem is that once you get small enough, the world stops being cooperative. At the large level, when you put something somewhere, it stays there until something moves it. At the small level, that stops being true. At the quantum level, physicists run into the old programmer’s dictum that "constants aren’t and variables don’t".

In fact, the most significant discovery of quantum mechanics is that it’s impossible to predict the action of an individual particle. For whatever reason they are pretty predictable in the aggregate, but as individuals, you can’t say anything about them at all.

That’s the fundamental challenge I see toward using things like electron spin. So far as we can tell, the individual electron doesn’t follow any laws. In one sense it’s an engineering problem to use a material which doesn’t follow any laws. In another sense, it’s a fundamental bandwidth problem. The higher your noise level, the lower the theoretical maximum bandwidth of your medium.

And the problem is that since every bit is equally precious (to the storage layer), you have to protect them all. You can’t tell the difference between an unimportant low-order chroma bit in the data stream, and a very important high-order bit in the frame rate.

And you can always overcome noise with redundancy (which all ECC inherently are), but the question is how much redundancy is necessary, and once you use it, how much better is your bandwidth really?

As you say, we’ll see. I’m just expressing why I have my doubts and wouldn’t count on electron spin storage.

Nevertheless it seems to me that all you’re posing is engineering challenges. But I guess we’ll see.

You’re right that no data transfer mechanism (storage is just transfer through time) is so unreliable that you can’t reliable store data on it. The problem, though, is that an unreliable medium means that your bandwidth is substantially reduced. I believe that people might come up with electron spin storage devices which are reliable because they contain enough redundancy and error correction; I sincerely doubt that they’ll have higher storage capacities than magnetic media (especially since there are still a few generations of improvement in magnetic media left) simply because there will be too much redundancy and error correction information required.

It’s nice that in a lab you can change the spin of an electron, but in a hard drive you have the added problem that you have to actually be able to find the thing again. You have to actually know where it is in a giant mess of electrons and atoms. And the things move around. And they all look the same. One iron atom isn’t distinguishable from another.

The great thing about tiny magnets on a disk platter is that once you put them somewhere, they stay there. But in anything less trivial than hydrogen, every atom has many electrons, so even finding each one is going to be essentially impossible (electrons aren’t really distinct from each other either; they don’t have names or faces).

So if you start doing things like give every electron on the atom the same spin, and make several atoms which are next to each other the same to, in order to get enough redundancy that you can reasonably find what you’re looking for, then you’re not very far removed from using tiny magnetic particles. It might work, but you’re not going to get enormous bandwidth improvements.

Once you get small enough, the world isn’t a reliable place. That’s why continual shrinking isn’t going to produce the same results. The universe doesn’t look the same when you get small enough; in our universe apparently only big things have to follow the rules.

I don’t know if modern RAM still has these problems, but conceptually the same rules would generally apply.