The use of kilowatt-hours per gigabyte of data transfer in digital carbon emissions models is something that's been the source of much conversation.
A LinkedIn comment from Romain Jacob, a researcher from ETH Zurich looking into energy of networks, highlighted that use of this metric is also controversial in academic circles. His comment also shares a handy way to think about the metric next time you see it used in an emissions calculation.
Just quickly about the kWh/GB metric (which is the center of a heated debate in the research community atm): this is a used and useful metric BUT it must be understood as an "attributive" metric (think of it as "how efficiently am I using the network"). It has no "predictive" value (as in "If I send xGB less, I'll consume ykWh less").
That's completely wrong due to the lack of proportionality I mentioned above.
The problem is, that (wrong) conclusion is the one that is most intuitive :-/