I mentioned microgeneration (of power) in my previous piece, just as a random Tory policy that I didn't give a hoot about either way. But the revival of the concept calls to mind one of the first pieces I wrote here, back in 2005, in one of my rare forays into things I actually know about.
If a 500MW power station could only be built by putting fifty thousand small 10kW generators in racks, with expensive complicated machinery to try to keep as many as possible fueled and running at once, then I don't think the concept of an electricity grid would ever have caught on. But that's what a "computing" power station looks like.

There are some slight economies of scale to computer hardware, mainly in management overhead, but compared to the cost of putting your own computer at the other end of a wide area network, they're negligible.

Now we are being told that central power stations are not, or at least will not be, a good idea even for power.

When I read about this in the Metro (yes, yes, I know), the idea was attributed to "a group of environmental and economics experts", which in fact turn out to be a bunch of lunatic lefties. So it's safe to assume that their facts are all rubbish. But the logic is sound, and just the same as my point about computers. For centralization to be efficient, the economies of scale have to outweigh the cost of having stuff a long way from where you need it.

The economies of scale are threefold: mechanical (a big turbine is more efficient than a small one), organizational (it's easier to keep two big machines fed & maintained than 200 small ones), and pooling (If everyone has their own little resource, they each need enough capacity to meet their own peak demand, which is much more capacity than overall peak demand).

For computing, as I argued, the first is non-existent or negative, the second marginal, and the third not enough to outweigh transmission costs.

Power generation that is based on making things very hot (i.e. all current major methods except hydro, but not photovoltaic solar) have large mechanical economies of scale because capacity varies with volume, and heat loss with surface area. But on the other hand, transmission costs are very high - both in building and maintaining the Grid and in transmission losses.

(I also seem to remember reading that the seasteading people looked at OTEC and found that it couldn't be made effective on a small scale, though it might on a larger scale. It still seems to be turbines and things, so perhaps it's the same fundamental scale effect.)

The management scale issue is questionable.

The pooling is a big deal for power where the cost of production is large compared to the cost of having capacity - that's fuel-burning but not most renewables (or nuclear). They talk about offsetting the pooling problem with small-scale grids, or hydrogen storage, but that brings in a whole lot more management overhead. If microgeneration becomes sensible it will be because generation methods which are all capital costs and no fuel costs take over.

And no, my reader who knows who he is, a national torque grid of spinning axles will never be sensible.