[Buge]

Those numbers don't make sense to me. I just watched this video[1] in 4K 60fps on Youtube. I think it's ~5GB. I watched it at double speed so it took 15 minutes to watch. Your second link says it takes 5.12 kWh to transfer 1GB, so it took 25 kWH to watch that video. That's about $5 at California power prices. Did it really cost $5 to watch that free Youtube video?

My ISP caps me at 1.2TB/month. That would be $1200/month, but I don't pay anywhere near that.

25kWh in 15 minutes is 100kW. That's 150x the max amount of power my overpowered desktop uses.

On HN people claim AWS overcharges on egress pricing. They charge between $0.05/GB and $0.09/GB[2]. Their "overcharging" price is much lower (5% - 11%) than the cost of electricity if your numbers are correct, which doesn't make sense. Their ingress price is free, so they're not making up for lost money there.

[1] https://www.youtube.com/watch?v=gidJopKtcnc

[2] https://aws.amazon.com/ec2/pricing/on-demand/

[nanidin]

The source for power per GB links to a paper published in 2012 [0] that describes how they arrived at 5.12kWh per GB. Maybe things have gotten more efficient in the last decade?

Note that the power figure includes all of the networking equipment between your desktop and the server, and the infrastructure required to run it like cooling equipment. Also note that watching a video at 2x speed may result in the server sending less data than watching a video at 1x speed.

[0] https://www.emergeinteractive.com/wp-content/uploads/2020/02...

[jonas21]

I wouldn't put too much weight into that number. Studies of energy usage for data transfer vary by over 5 orders of magnitude depending on their assumptions. This paper [1], which is an analysis of 14 different studies, came up with 0.06 kWh/GB in 2015, with energy usage halving every 2 years. Assuming the trend continued, that would put it at less than 0.01 kWh/GB today.

The claim about CO2 emissions for a Google search was also contested at the time and later retracted by the original source [2][3]. Unfortunately, once on the internet, these things just keep getting repeated.

[1] https://onlinelibrary.wiley.com/doi/pdf/10.1111/jiec.12630

[2] https://techcrunch.com/2009/01/12/revealed-the-times-made-up...

[3] https://googleblog.blogspot.com/2009/01/powering-google-sear...

[himinlomax]

> The source for power per GB links to a paper published in 2012 [0] that describes how they arrived at 5.12kWh per GB

Even for 2012, that number seems utterly implausible.

[jcims]

They are just dividing total consumption of data by total consumption of energy. While I would imagine both have risen substantially, I would imagine that consumption of data has risen by some multiple of consumption of energy.

[Buge]

>Also note that watching a video at 2x speed may result in the server sending less data than watching a video at 1x speed.

That's an interesting point, but I don't think that's the case for Youtube.