distribution losses vs transmission losses
> The problem seems to be power distribution capacity rather then generation. .. We should see data centres clustered round power stations.
Yes and no. Low-voltage distribution is the real problem. Long-distance HV transmission is extremely efficient (less than 2% losses in transmission, IIRC) because it happens at hundreds of thousands of volts, so you need less current for the same power, and the resistive losses of a cable (which are the bulk of the losses) go up with the square of current.
So actually siting a datacentre next to its own HV substation would be just as good as putting it next to a power station. The trouble is that 400kV transformers are incredibly expensive, so there's no point in building a small one just for a datacentre, you have to build one big enough for a city, step it down to 132kV for transmission around the city, then 33kV to the really big datacentres, 11kV for the medium sized ones, and the really crappy little ones would need to tap off 415V from the same distribution wire that supplies houses etc.
The 415V and 11kV local distribution lines in most cities (and even the 132kV and 33kV distribution mains) are hugely overstretched due to the expansion of electric heating and EVs etc, not just datacentres, and as they get hot, the resistance goes up even further. We lose about 10% in this local distribution (iirc again)
> That would include the landing points of the off-shore wind farms.
Er, probably not a good idea for reliability, wind farms have a habit of tripping offline very suddenly, even if the wind IS blowing (e.g. if it is blowing too hard, or if the grid frequency goes too high or too low)
Biting the hand that feeds IT
