The Hidden Cost of the Cloud
Nothing you store in the cloud is anywhere other than a data centre: a warehouse of computers, wired to the electricity grid and cooled continuously. The metaphor hides the physics, and the physics is where the cost lives.
Two numbers that matter
Efficiency is usually described with PUE — total facility power divided by the power that actually reaches the computers. A PUE of 1.1 means almost all the energy does useful work; 2.0 means half is spent on cooling and losses. The other number is water use for evaporative cooling, which is why data centres in dry regions attract criticism.
Why location is a design decision
Cold climates reduce cooling load, cheap hydroelectric power reduces cost, and proximity to users reduces latency. Those three pull in different directions, which is why large facilities cluster in places like Ireland, northern Sweden and the Pacific Northwest. The cloud has a geography, and it is visible on a map.
What follows for users
Efficiency is mostly a scale and design problem, so consolidation can genuinely reduce total energy compared with thousands of small servers. But demand growth is the dominant term: efficiency gains mean little if consumption doubles. That is the honest version of the argument.
Knowledge check
Two questions, no score
What does a PUE of 1.1 mean?
Now prove it