the biggest practical advantage to renewable energy is that we can produce it ourselves and become independent of crises that happen in another part of the world. for example the iran war.
for digital infrastructure, it’s the same. if facebook changes their policy to surveill all citizens in a new way, that creates a shock for everyone else. if we can self-host our services, then we become independent. so it’s like solar-panels in the back yard, but for digital infrastructure.
Efficiency depends. Both electrical grids and remote connections experence huge transmission losses. They also require additional work to deal with the larger concentration of power, noises, and heat compared to running the same nodes across a wider area.
An edge node can also be tapped as part of a distributed grid minimizing downtime as well.
Not to say they don’t have plenty of cases where they are more efficient. For example data processing over very large colocated datasets, if the data is going to have to travel anyways that effecicny is lost in a distributed architecture.
Following the sun or distributing across large distances to avoid localized weather disruptions to renewables requires very high voltages to do cost effectively which is more difficult (not impossible) with a dynamic and distributed production network which will need to fluctuate between voltages for safety and effectiveness on demand compared to a high pressure to lower pressure only flow that a centralized grid can make. That said, if a building can power itself 99.99% of the time, the cost needed for transmission is much lower then a 100% import model.