Yep, my bad! So commutes aren't as significant a contributor as I thought. I ran through some more calculations trying to only use the US, and it looks like the Wh / $ is somewhere around 3.51. Is there any other category of use that isn't being included? This is a pretty stark difference, even being many times more than the first calculation.
Commute: 1.028 TWh
Data centers: ~3.3TWh per trillion (for USD) = 132 TWh
ATMs: 470,135 independent in the US (1.27MWh each per year) = 0.597TWh
Banks: ~98,000 in the US (PDF wasn't clear, using 60MWh each per year) = 5.88 TWh
Card networks: ~1TWh/yr (just assuming "half" of global)
OK the main discrepancy is from the Data center calculation. The Galaxy PDF lists the top 100 banks as owning $70Tn of deposits. I calculated using the galaxy energy number but divided by total deposits wheras I should have only used the data for the top 100 banks as they did.
I think your calculation of 3.5 Wh/$ is closer to the real number. When I put $70Tn instead of $404Tn in I get 3.71Wh/$
Commute: 1.028 TWh
Data centers: ~3.3TWh per trillion (for USD) = 132 TWh
ATMs: 470,135 independent in the US (1.27MWh each per year) = 0.597TWh
Banks: ~98,000 in the US (PDF wasn't clear, using 60MWh each per year) = 5.88 TWh
Card networks: ~1TWh/yr (just assuming "half" of global)
Total: 140.5TWh/yr
https://data.worldbank.org/indicator/FB.CBK.BRCH.P5?location...
https://blog.sfgate.com/lifestyle/2020/08/31/its-electric-ho....