Web Site Energy Use

Leaving Our Carbon Footprints on the Information Super Highway

Jul 31, 2024

“The communications industry could use 20% of all the world’s electricity by 2025.” Source: Climate Home News.

With that in mind, I've been looking at various carbon calculators for analysing the carbon footprints of web sites, or rather, the carbon footprint we generate by using those websites. A search will turn up a list of free web site carbon calculators; they are very easy to use- you simply enter the web address you want to check out and the calculator will do its thing and give you some results. Most of the time.

They work by measuring how much data is transmitted, estimate the energy cost and may come up with a figure of grams of CO2 per page and /or rate the web site against a global average. They are inevitably an approximation as pages across a site will contain more or less data (though you can enter an individual page address to check out a specific page) and a person visiting the site may not stay on a page long enough for it to fully load, though given the speed of some folks' internet, this is not as significant as it used to be. Also, accessing a particular service on a web site may trigger more or less computation and processing power at the server end- more of that later.

Its interesting to compare results from different web sites across on different calculators. I chose three calculators and tried to find values for my own web site (www.konsk.co.uk), Substack (https://substack.com) and BBC News (https://www.bbc.co.uk/news) with the following results:

Firstly, using “The Original” Website Carbon Calculator”:

Substack is given a D and is “dirtier than 55 % of all web pages globally”
BBC news is given an F and “is dirtier than 73 % of all web pages globally”
My web site, and I quote, “Hurrah! This web page achieves a carbon rating of A+” and “is cleaner than 100 % of all web pages globally”

I'm not really sure how it can be cleaner than 100% of anything...Before I explain, here's another calculator looking at the same sites again.

This one is called The Sustainable Web Guide:

Results for Substack:
“This page transmits approximately 2400 kB and generates an estimated 0.57g CO2 during a single visit. The website's hosting provider is listed as a green host in the Green Web Directory database”. Rated at C.
Results for BBC News:
“This page transmits approximately 3665 kB and generates an estimated 1.00g CO2 during a single visit. The website's hosting provider is not listed as a green host in the Green Web Directory database.” Rated at E.
Results for my web site:
This page transmits approximately 25 kB and generates an estimated 0.01g CO2 during a single visit. Rated at A.

And finally, Web Site Emissions which takes a slightly different approach and estimates usage per year given 5,000 visits to the site per month:

Results for Substack:
“Each visit to this web page generates 0.60g of CO2e.
Based on 5000 page views a month, this webpage produces 37kg of carbon every year.”
Results for BBC News:
“Each visit to this web page generates 1.53g of CO2e.
Based on 5000 page views a month, this webpage produces 93kg of carbon every year.”
Results for my web site:
“Each visit to this web page generates 0.01g of CO2e.
Based on 5000 page views a month, this webpage produces 1kg of carbon every year.”

So what's going on here, why so much difference? My web site, which has changed very little since I first started it, other than to add more pages, is known as a static website, written in pure HTML. I put it together when internet speeds were very low1 compared to today and deliberately kept graphics simple and stuck to text based pages to cater for folk with poor connections. There is no user interaction, no comments, no fancy graphics or video, no tracking, cookies or advertising, nothing does anything, its just a location for pages of information, hence the term static.

Both the BBC and Substack, like many other sites, are dynamic sites, meaning interactive, using combinations of various flavours of HTML, XML, Java scripts, python or whatever with opportunities for comments, likes, sound, video and multiple links to other pages etc. etc. Of course, web sites like the BBC and Substack have many thousands of users so its easy to see how they can build up large energy demands.

Searching for the UK's average internet use produces a whole range of reports and varying results. An Off-Com report in May 2023 gives the average amount of daily time spent online as 3 hours and 41. This means the average online adult now spends around 56 days each year online – two more days than in 2022. Young online adults aged 18–24-spend the most time online, at 4 hours 36 minutes each day. Given these sort of figures its very easy to see how internet use consumes large amounts of energy.

What the web calculators are attempting to measure is the energy used by the servers that host the web sites we visit. When we click a link in our browser, a request is sent to the required server which then has to find the information on its storage (still largely hard disks, which require motors to spin their platters containing the data) read it and transmit it back. As the servers in data centres are upgraded from hard disks to solid state drives, the energy use will come down, but this is more than compensated for by the increasing complexities of the operations that are being called upon.

Basically, the energy use jumps when the server has to do more work. So, as I noted above, a static site like mine does nothing other than serve up pages but wearing another hat I write in Cymraeg (Welsh) for Llygad Y Dydd (the Eye of the Day, or indeed, a Day’s Eye or Daisy), our local Welsh language newspaper. To check my work I use the translate facilities provided by Duckduckgo. Now the server is not simply looking up and presenting me with data but has to do some quite sophisticated operations to translate my paragraph of Cymraeg into English or vice versa. Similarly, playing games requires more computing power from the servers and hence more energy use.

Once we get into Generative AI, such as the various image creators, then there's a much bigger jump. The length of time taken to come back with an image created from your text input gives a good indication of the amount of processing being done and hence the amount of energy being used. While you sit there waiting for your picture, a high powered computer is running through zillions of calculations to generate a picture of, say, a flying pig wearing a bow tie (or whatever you fancy), for you.

This occurs whatever device you might be using. So praising our mobile phones for their low energy use and long battery life is simply ignoring the fact that pretty much all the energy use for the work it does has been offloaded to a data centre somewhere.

For example, in 2023 greenhouse gas emissions of the world famous search engine that-shall-not-be-named were 48% higher than in 2019, according to its latest environmental report. The tech giant puts it down to the increasing amounts of energy needed by its data centres, exacerbated by the explosive growth of AI. The company's target of reaching net zero emissions by 2030 is now in doubt; it admits that "as we further integrate AI into our products, reducing emissions may be challenging."

Its us at the user end that are actually externalising our energy use to the server farms which provide both the backbone of internet use and the services that we use and indeed seem to have become highly reliant and even dependant on. Externalising costs is classic capitalism; aeroplanes, for example, can only continue offering cheap flights because they do not pay for the pollution they cause, they externalise it onto the environment.

These web services, such as social networks, mapping and location services, translation and the growing number of AI enhanced applications such as search and image creation, require increasing computation and processing power and hence use more energy.

Just to put a few rough figures on it, in 2022 the world’s data centres used 460 terawatt hours of electricity, and the International Energy Agency expects this to double in just four years, roughly equivalent to the electricity consumption of Japan with a population of 125 million people.

On a different scale, nearly a fifth of Ireland’s electricity is used up by data centres, and this figure is expected to grow significantly in the next few years. This has become such a problem for Ireland that restrictions are placed on where data centres can be built as they can effectively drain a nearby town or city's power supply. There is even talk of building data centres with their own nuclear reactors to provide the power they need.

So what can we do?

Probably the most important thing is to consider how to live without any of it! Or at least, how to reduce dependency. I remember having a right laugh at some poor guy talking on the radio who, during a three day failure of the Orange mobile network, found himself in his usual coffee break, staring in horror at the blank screen on his phone with nothing to occupy his attention. He described how he was able to reduce his anxiety by watching someone else playing with their phone....I mean, do we really want to be like this??

Consider data storage- dark data is data which has been collected by organisations and individuals and is stored on hard drives or chips on the net but which has only been used once or not at all. Storage in this way still takes up large amounts of energy even when it is not being used. It is estimated that on average, 65% of the data an organisation stores is dark data. Its obvious that organisations could do a lot to greatly reduce this energy use.

We can also think about how we use storage. Anything stored in the so-called cloud needs energy to keep it live and accessible. Consider the number of photographs that now reside on Facebook, or rather, the enormous number of photos that Facebook users have uploaded to Facebook data centres. This represents a seriously HUGE amount of data stored on zillions of continually spinning disk drives or (slightly better) zillions of chips. And that's only one collection of photos on the web.

For myself, rather than taking advantage of general purpose on-line storage, such as that offered by the World Famous Search Engine or the other similarly famous sinful fruit one, I use flash drives as back up and storage. These are now very cheap, of a useful large size, fast and require no power when unplugged. I've taken a lot of pictures over the years for teaching and illustration purposes but find I can easily fit them all onto just one.

Another flash drive holds all the text I've ever written in digital format, easily. One for permaculture design, one for digital texts and so on. Its easy to copy a flash drive so I can stick copies in a different location (buried, fire proof box...) in case of disaster.

Personally, I think about the increasing number of articles I've written that are stored by Substack. Would it be better for me to transfer my past articles from Substack to my static web site, where they will remain available but can be accessed with less energy?

With my Disaster Preparation hat on, its probably a good idea for us all to dump the mobile/tablet/laptop regularly and take the opportunity to design some good (better?) alternatives for when we really need them.

Yes indeed, many thanks for reading and a warm welcome to new subscribers. Next episode of Konsk will be following shortly then I’m taking a look at art, in particular fiction and film and how they can influence our culture, or not, for better or for worse.

E.S.P. Adapt

Discussion about this post