Thanks, so we're actually in agreement then :) If you look at your own source (the two Figure 3 inserts from 2017 and 2018 studies), they estimate that in 2020

Thanks, so we're actually in agreement then :) If you look at your own source (the two Figure 3 inserts from 2017 and 2018 studies), they estimate that in 2020 the kWh/GB measure would be between 0.1 and 0.01. That equates to between 60 and 6 grams of CO2 per GB, smack bang in the middle of the International Energy Agency estimate I linked to (29 grams). The reason for the big discrepancy is your source article was written in 2016, and it referenced a study from 2012 for its 5.12 kWh/GB number, and that 2012 study used bandwidth data from 2011. Telecommunication systems are still getting exponentially more efficient over time (see the second Figure 3), so estimates get old very quickly.