[roenxi]

But this is the other way around; if I wanted to draw a computing analogy it would be arguing that if I wanted to provide massive cloud storage then the physical volume density of information (bytes/m3, a funny but practical measure to talk about) can't approach the density of 1Tb consumer grade hardware.

And that would be true, but it could still be a good target within an order of magnitude, because it could be set up as a very big RAID array. There would losses due to cooling and power and space between the drives and whatnot, but to a first approximation (Facility Volume / 1Tb density * 50% fudge factor) will be pretty reasonable; especially if someone really tried to engineer clever heating solutions.

So the paper gets (checks) ~1,300t/ha lab conditions. Why, in theory, can't they just replicate the lab environment as many times as fits in a big facility? I can see practical impediments like cost, I can't see theoretical ones if it is decided that This Must Be Done. I'd expect reasonable yields of 650 t/ha. That improves on 17 t/ha farming practice [0]. Point is that sort of calculation is totally routine when looking at academic results and I don't see what the problem is using a 3'x3' lab environment then extrapolating with a fudge factor.

[0] https://en.wikipedia.org/wiki/Wheat#Yields