We know that by and large the biggest new proposed datacentres aren't intended to aggregate existing capacity but to add to it. Adding 26GW spread across three datacentres no doubt
is more efficient than spreading 26GW across 26 or 260 new datacentres. But anything like 100GW or 400GW of total extra datacentre capacity, in the US alone
https://www.forbes.com/sites/giacomotognini/2026/10/02/insid... , is enough of an increase that the question becomes whether the increase is economically or socially justified at all, and some (I assume) relatively modest efficiency gains from taking datacentre aggregation to the next level surely won't have a decisive impact on that. We're certainly beyond the point where such questions are largely just a problem for the investors (or even for Americans), especially when it turns out that, whoops, the DCs are going to be powered by finite, highly-dispatchable and CO_2-emitting natural gas after all, not baseload nuclear stations.
Meanwhile from a local perspective the fact that a very large datacentre next door is hypothetically aggregating what would have been several other data centres, mostly built elsewhere, is not so reassuring. (Though the other side of that coin is that from the builders' point of view it may be politically easier to get a given amount of capacity built as a few very large plants in places which are less likely or less able to object: regulatory efficiency, distinct from energy efficiency.)
And also conversely, if the mega-DCs (deca-giga-DCs?) really are markedly more efficient, and if they do aggregate lots of new capacity, most especially if they do also end up aggregating existing capacity, that will worsen problems of competition and market concentration in the cloud computing market, at least once the money spigot is turned off again and it's no longer the case that any idiot can get financing for a new datacentre.