Does that use a lot of energy?
An interactive tool to get a sense of scale for energy consumption.
Without data, it’s incredibly hard to get a sense of scale.
One area where this is extremely apparent is in energy use and carbon emissions. It’s hard to know what matters a lot, and what very little.
A lot of my work with data is about getting a sense of whether something is big or small, effective or not. Is that a big number?
To help, I’ve built an interactive tool that lets you compare the energy use of different products and activities: from lighting and cooking, to heating and driving. This is focused on “household” energy use: what people consume at home or while commuting.
In the video below, I’ve shown a quick example of what you can do with it. Select the products or activities you want to compare, and adjust the basic inputs, which are usually the number of hours or minutes of usage, or miles driven.
To build this, I developed a dataset of energy consumption based on typical usage. As I stress in the tool, these are approximations. They will not reflect your life down to the watt-hour level. I have assumed an LED lightbulb is 10 watts; yours might be 8.5.
Getting caught up in these small differences is actually against what I’m trying to do here. I’m trying to give people (and myself) a sense of what’s small and big. What order of magnitude are we talking about? Is driving 10 miles equal to 10, 100, or 1000 hours of lighting?
If this tool looks useful and is making a difference, I might build a similar one for carbon emissions (it will allow me to include things like food and other purchases).
Do you have feedback?
This is still a work-in-progress, so if you have feedback to share, I’d really appreciate it. There are now more than 70,000 of you following this newsletter, so I can’t promise a reply to everyone, but I do promise to read your messages and use them to make improvements.
Below the tool, I’ve written a methodology document detailing my assumptions for each product. It was a long list, took quite a bit of work, and there might be a few mistakes in there.
Feedback that is useful is along the lines of:
Your assumption for the energy rating or power consumption of X is really off
This component/configuration is broken
This tool would be much more useful if it had X functionality
It would be great to add [some product/activity not listed]
Comments that are less helpful:
Those focused on small differences between my numbers and your actual usage. As I said, I’m going for typical approximations here; it won’t reflect precise reality for everyone
Things that make the tool a lot more complex. The simplicity here is important and deliberate, so people can make relatively quick comparisons. I don’t want people to have to put in the specific model of television, dishwasher, or car; or give precise details about the fabric of their home.
Hey Hannah. I have feedback on the Netflix/Youtube figures. My team's research was used by the IEA for the article you reference, and is also used by Netflix and other digital services to estimate their footprints. The figures you use are not appropriate in this context for two main reasons:
1. (The main one) Your other figures are the marginal increase in energy use for the activity, not the total energy attributed to it - so your AI doesn't include training, your laptop doesnt include manufacture, etc. The Netflix/Youtube figures you quote are attributional, and are not the *increase* in energy from an additional stream. The reason is that networking equipment doesnt change energy use significantly with data quantity transmitted. You can observe this on your home router/wifi. Our network infrastructure is always on, and provides us with access to many digital services. The marginal increase of streaming using it is minimal.
Hence it is misleading to attribute a share of energy from the network to an activity (eg streaming) in a way which implies that carrying out the activity increases energy use, and not doing it reduces energy use.
It is obvious in other activities - eg if the carbon footprint of an average swim at a public swimming pool is 3Kg, we dont believe that additional swimmers using it increase the overall footprint.
This error in reasoning was used to create scare figures that 'data networks will use large quantities of energy globally' by extrapolating increase data traffic and assuming that energy use increases linearly with it. Including in academic publications, and on R4s 'Rare Earth'
This argument also applies to the myth that you reduce emissions by making the size of websites smaller. It doesnt, and consultancies that claim otherwise in the services they offer are greenwashing.
(From memory - but can check later if you want:) I recall a figure from Netflix that the power use by the CDN equipment averages at 0.1W per stream. This is the only equipment which is additional - and so comparable to (eg) the AI figure.
2. The IEA report is old, and we send a lot more data through the internet now so networking equipment is used far more efficiently. As a result, the attributional figure is now far lower than what you have quoted.
If you want more info, best to email me: chris.preist@bristol.ac.uk
@andymasley - I messaged you before about this a few months ago, in response to a draft post you made. Here are a few more details to the argument.
Thank you for creating it.
I wonder if for context the average daily kW usage can be added? I guess it'll greatly vary by number of people but also split by US/other high/middle/low income countries? Because honestly otherwise, I struggle to know if 800kW is a lot of not, but maybe it's just my ignorance