The Hidden Gaps in Our Digital Carbon Footprint: A Conversation with Jens Gröger of Öko-Institut

Back in 2020, Jens Gröger, Senior Researcher for Sustainable Digital Infrastructures at the Öko-Institut, published an infographic that visualises the carbon footprint of our digital lives. It breaks down each area of our digital lives by the amount of carbon they emit—and the ratios between these sections are especially interesting.

The infographic shows that the CO2 emissions from our devices and the consumption of digital services amount to almost one tonne. That is almost half of the emissions we should be responsible for if we want to achieve our climate targets. Two tonnes would be our total CO2 budget. “If I, as a citizen of the world, am allowed to emit around two tonnes of CO2 in total and three quarters of a tonne is already attributable to digital consumption, then there’s not much left for heating, food, mobility or clothing,” says Jens Gröger.

The other thing the graph shows is the huge impact of digital devices. The production of these alone accounts for almost a third of our digital emissions.

Öko-Institut e. V.

The infographic shows our digital carbon footprint.

We wanted to find out more about our digital carbon footprint so we met up with Jens Gröger for an interview. We asked him what is behind the term ‘digital ecological footprint’ or ‘digital carbon footprint’ and how it is calculated.

The graphic is based on figures before the big AI hype; before language models such as ChatGPT and other AI applications spread rapidly. We therefore also spoke to Jens Gröger about how AI models have changed our digital carbon footprints.

RESET: When we use the term digital carbon footprint, what does it actually mean?

Jens Gröger: My digital carbon footprint is the negative environmental impact I cause through my digital lifestyle. In other words, how the use of digital technology, the purchase of devices and the consumption of digital services affect the environment. Together, these factors refer to the digital carbon footprint, which measures CO2 as an environmental impact. But of course you can also measure other footprints such as a resource footprint or a water footprint.

How can we ensure a green digital future?

Growing e-waste, carbon emissions from AI, data centre water usage––is rampant digitalisation compatible with a healthy planet? Our latest project explores how digital tools and services can be developed with sustainability in mind.

The life cycle assessment of a product is crucial to these calculation methods. I look at the entire life cycle of a physical product: the extraction of raw materials; production and transport; and use and disposal. This allows me to identify the ecological hotspots. Where do the emissions actually occur? Is it in the extraction of raw materials, in the mines? Are there any toxic gases escaping that workers have to breathe in? Or is it in the factory where the chips are produced? Or afterwards during recycling or when devices are disposed illegally? If I know where these hotspots occur, I can also take appropriate countermeasures.

In the case of smartphones, for example, we’ve realised that energy consumption during the usage phase is not so relevant to the overall environmental impact. It’s much more relevant how long-lasting the battery is, for example, so that I don’t have to throw the device away after two years. This means that the life cycle assessment can be used to adjust the requirements for products so that they really address and reduce their emissions.

© Jens Gröger

Jens Gröger, Senior Researcher for Sustainable Digital Infrastructures at the Öko-Institut.

But calculating a digital carbon footprint is much more complex than just looking at a single product, isn’t it?

Yes, that’s right. Let’s take a look at the smartphone example again. To make this smartphone, I need a circuit board, lots of electronic components, a housing, a battery, glass for the display and so on. The production of this smartphone alone causes around 100 kilograms of CO2 emissions. Then there are the emissions from the usage phase. If I charge it every day, I cause around five kilograms of CO2 emissions per year during the usage phase. This means that it is important to differentiate between the manufacturing costs on the one hand and the utilisation phase on the other. And then at the end, in the end-of-life phase, there is of course the amount of waste: the e-waste.

Just looking at products creates a simple picture of our digital carbon footprints. However, if I look at my digital lifestyle, it doesn’t just consist of products, but also of services that I use. When I stream a video for an hour, store data in the cloud or even just make a video call, I use all kinds of digital services. All I need is my end device. But of course there’s a network attached to it: data passes to data centres, which require energy to process that data. This also requires a lot of technology and means that a lot of computing operations take place in many different locations. And that makes measuring digital carbon footprints very complicated.

To determine an individual digital carbon footprint, I would have to add up the emissions of all these devices along the entire digital supply chain. I’d have to factor in their manufacturing costs, their electricity consumption, the water consumption in data centres, to name a few. And then I would have to allocate this figure proportionately to the respective services I use and how long I spend using them.

And we don’t always have all the information to make this calculation, right? I think we’ve all heard about the high energy consumption of data centres. And many people are also now aware that our devices require a lot of resources. However, very little light is shed on the network—in other words, everything that happens between my device and the data centre. But this is a very relevant material infrastructure in terms of emissions.

Yes, exactly. We can divide the digital ecological footprint into end devices, networks and data centres. And little is actually known about what happens in these telecommunications networks. We know that the energy consumption of these networks—at least in the past—was just as high as the energy consumption of data centres.

How can I visualise these networks in concrete terms?

A network initially consists of a router or, in the case of mobile data, a mobile phone antenna. On the wired route, the router is connected to a copper or fibre optic cable that leads to a network access point. These are located in those grey boxes you see on the pavement. From there, it travels via a collective network to another network node; this continues until it reaches a core network. In other words, the network is structured like a snowflake. In the core network, huge amounts of data are transmitted very quickly to other locations and then passed on again along these branches to a data centre.

And where do emissions occur in the network?

The network has a very high standby consumption, regardless of how much data I send through it. This means that it ultimately doesn’t matter to the network whether I switch the screen on or off during a video conference, for example. I use the network either way. In a data centre, on the other hand, turning my screen off reduces emissions. This is because the data centre must process more data in order to show what’s happening on the video screen.

This means that the power consumption of the network is initially a fixed block. However, we still know very little about it. But we do know that it’s not increasing as fast as data centres. Within just three years, from 2023 to 2026, the power consumption of data centres will have doubled. Four years later, in 2030, it is expected to be four times as high. The energy consumption of telecommunications networks, on the other hand, will remain more or less constant.

If we look at where we lose most energy in the network, it’s clear that the home router and the network access point (which is located in the grey box on the street) dominate energy consumption. Further along the route towards the core network, the specific energy consumption per volume of data transmitted decreases rapidly because the infrastructure is shared by more and more data streams.

We’ve compared different network access points in research projects. We realised that it makes a big difference whether you access the internet via mobile communications or a wired network. The latter causes fewer emissions.

In the digital footprint calculator that I co-developed, we differentiate between mobile and wired access networks. We could also distinguish between 5G, 4G or 3G networks and copper cables vs fibre optic access networks. They all generate varying digital carbon footprints.

But many of the numbers are difficult to determine, right?

Yes, so it would be helpful if we knew more about the network. And if the network operators themselves had an interest in finding out where they’re wasting energy unnecessarily.

Data centres now have to report on their power consumption. Since the Energy Efficiency Directive came into force in the EU, data centres above a certain size have been obliged to register in an energy efficiency register. This is primarily intended to enable legislators to see where the large data centres are located, how much electricity they need and where new power lines may need to be laid. This means that the data collected at national level must be passed on to a European data centre register, which will then provide a Europe-wide overview of the energy consumption of data centres.

But the network operators don’t have to provide any information?

This depends on the country. The UK and Germany, for example, require network operators (like all larger companies) to produce general sustainability reports in which they share the company’s total energy consumption. But, they’re not obliged to provide information to individual customers about their digital consumption.

In France, however, telecoms companies actually have to report how much CO2 emissions each customer generates with their internet connection. This is a very good example of what we also need for digital services.

If I had a label or a key figure for the emissions of my digital services, then it would force providers to compete with one another. Online storage providers, for example, could then boast about offering low-emission gigabytes.

Of course, this assumes that there really are standardised and comparable key figures and calculation methods, and that brands can’t just greenwash their offerings.

But we could definitely approach the whole issue more proactively. For example, the EU Commission is planning to launch an efficiency label for data centres. Perhaps the result will be something similar to what we know from fridges, with a scale from A to G, from green to red.

However, networks also have a material basis, i.e. the cables, routers and so on. Are there any figures on the environmental footprint of the infrastructure?

I’m not aware of a comprehensive life cycle assessment of the telecommunications network. What’s clear is that the network has lot of copper and glass fibre. There’s a lot of steel and concrete in the mobile phone masts. And the network nodes and mobile phone stations contain a lot of technology. The reassuring thing is that we will use this infrastructure for a very long time, which will also spread the costs over several years.

In the graph you drew up for the Öko-Institut in 2020, devices, networks and data centres each account for roughly a third of emissions. Is that still up to date? Or are large language models, such as ChatGPT and other AI applications, shifting the ratio now you’ve added them to your calculations?

So, what has really changed is that online services are becoming increasingly energy-hungry. A lot is happening in the cloud, particularly as a result of artificial intelligence. However, I can’t accurately attribute these emissions to individual users. I don’t have any consumption figures for individual users from data centres and can therefore only make estimates.

And for many digital services, such as Spotify, we don’t know the technical infrastructure behind them. Therefore, I can only estimate the CO2 emissions, as the respective providers don’t share any figures.

What the digital ecological footprint clearly shows is that we generate a lot of CO2 emissions by producing digital devices. This is probably still the case.

Yes, and unfortunately devices have an increasingly short life-span. There aren’t any major technical developments in how we build smartphones, but software development is constantly evolving. Here at the Öko-Institut, we’ve just had another Windows update which has meant we have to throw out a number of laptops because they don’t function with the new version. The same applies to other devices. If an app no longer runs on my old Android operating system, I have to throw my smartphone away, even if it still works. Software development is causing more and more electronic waste.

My digital carbon footprint depends on how long I use my devices. Say I own a smartphone which requires around 100 kilograms of CO2 emissions to produce. If I own it for just one year and then throw it away, I consume those 100 kilograms in one year. But if I use it for ten years, I will only generate 10 kilograms of CO2 per year.

However, it’s important to realise that we’re not just talking about CO2 emissions or energy consumption here. There are countless additional environmental impacts of our technology.

What do you think are the key levers if we really want to make digitalisation sustainable?

The first step is always more transparency. Because emissions that I don’t measure, I can’t reduce. Sharing emissions data would also give more efficient solutions a real chance on the market. We’d be able to clearly see their environmental benefits.

The second step is to set appropriate minimum standards. This has already been done for other products. Light bulbs, for example, must meet certain efficiency parameters, otherwise, they’ll be banned from the market. This requirement caused the boom in energy-saving LED lamps. And the same applies to digital services. Once we establish transparency, we can drive the biggest energy wasters off the market with minimum requirements.

This is still relatively easy for appliances. But, of course, it’s difficult to really compare digital services with one another. And it’s not so easy to say: “A Google enquiry is too energy-intensive, why not do a Qwant enquiry?” Because the results may vary in quality. But it would be ideal if we could mandate transparency in some way.

GOOD search engine: Why web search without advertising is worthwhile for users and the climate

The GOOD search engine has switched from a free service to a paid model. Co-founder Andreas Renner explains in an interview why this saves CO2 emissions and benefits users.

We also really need to get to grips with data centres. There should be regulations that require data centres to switch to green electricity. Then, they would be subject to the fluctuating daily supply of renewable energy. We should also enforce them to use their waste heat sensibly so they don’t contribute to local water shortages with their cooling towers. Currently, the opposite is happening. Electricity consumption is going through the roof. This means that we can’t shut down existing coal and gas-fired power plants. Data centres are currently jeopardising the phase-out of fossil fuels.

There are many starting points to improve our devices. It all comes back to the circular economy. In other words, we must design products that can use recycled raw materials, instead of new.

Why many so-called ‘sustainable’ computers are not sustainable—and how to do it right

Recycled materials, smaller packaging and CO2 certificates: PC manufacturers appear to be endeavouring to achieve sustainability. However, only a few approaches are consistently well thought out. In this interview, the French cooperative Commown reveals what truly sustainable computers look like.

A standard search engine enquiry uses much less energy than a ChatGPT conversation. We don’t talk about this enough despite the AI boom.

We could demand real prices for these digital services. If every enquiry costs one euro, I would think twice about whether I really need to use ChatGPT, AI or Google search. Currently, we share personal data or consume adverts in exchange for using these services.

Let’s assume that every service has a carbon footprint attached to it. A service that runs on fossil fuels will have a higher carbon footprint than one that runs on solar energy. I could then price such services accordingly if there were a standardised CO2 price. Services would then be cheaper during sunny or windy periods.

It would also be a start if I had a CO2 tracker in my browser. Then at the end of the day, I’d realise: “Oops, I’ve burned another 25 kilograms of CO2 on the internet today!” This would create more awareness. But we also need transparency in our services for this to work.

Ultimately, we have to start with data centres, where there’s a lot of savings potential. Perhaps at some point data centres will contribute to the energy transition or to climate protection as a whole.

To what extent could data centres contribute to the energy transition and climate protection? This is about the utilisation of waste heat, right?

We’ve identified a whole series of ways in which data centres can improve. These include low losses in the cooling systems, use of green electricity, high server utilisation, contributions to avoiding electronic waste and reuse of waste heat.

In principle, data centres could help us get rid of fossil-fuelled heating systems more quickly, balancing out the fluctuations in the electricity grid caused by solar and wind power. These digital businesses earn so much money that these same companies could also take on more social and environmental responsibility.

Thank you very much for the interview, Jens!