Our photos? Stored in the cloud. Our music and TV series? In the cloud. And sharing data, collaborating on documents and running software applications? That’s all done via the cloud too. Put simply, the cloud is a digital library. We use shared spaces, but have private ones too. And whether it’s individuals, governments or businesses: almost everyone uses it.
It is, of course, very convenient to access our data, collaborate digitally, and manage processes from any location or device. And isn’t it also more sustainable to use cloud storage and services instead of hard drive storage and local servers, and to send large amounts of data over the internet? If we use shared infrastructure, we definitely need fewer data storage devices – that much is clear. Nevertheless, the answer to this question requires a more nuanced consideration.
The cloud is at home in data centres
This is primarily because clouds—unlike real clouds—don’t simply float above us. Just as books belong in a library building, our digital data ‘resides’ in data centres. And this is precisely where we need to start if we want to understand the sustainability of our cloud services. There are also other factors that determine the volume of CO2 emissions generated by our lives and work in the cloud.
“The cloud is simply someone else’s computer …
… and this computer is probably located in a huge data centre.” This is how Dr Anne Pasek from Trent University describes the cloud. Cloud computing refers to the provision of hardware and software that users can access via a network – usually the internet.
The IT infrastructure provides, for example, computing power, storage capacity, application software or database services. As a result, users require comparatively little hardware infrastructure.
Cloud services for private users include, for example, music and video streaming services on the internet or online storage. Companies often outsource their entire organisational infrastructure to cloud platforms.
CO2 emissions from data centres are casting a shadow over the cloud
For a long time, the move to the cloud was touted as the ultimate boon for the environment. It meant there was hardly any need for on-premises hardware. Yet, the more workload is shifted to data centres, the more this ‘green’ sheen begins to fade.
Let’s take a concrete example: the construction of a new Google data centre in Oregon, USA, cost more than 1.8 billion dollars. It is as big as several football pitches, and its electricity consumption accounts for more than two-thirds of the entire supply district’s total demand.
This is just one example of the thousands of ever-larger data centres being built worldwide. Europe is no exception. Germany already ranks first as the country with the most computing power in Europe – and there are massive expansion plans.
The boom in computing power is primarily driven by the increasing use of cloud services and AI, particularly generative AI. According to a recent prediction, “currently, 80 percent of data centre capacity is cloud computing compared to 20 percent in AI. However, this ratio is going to reverse by
2030.”
The social, economic and environmental impacts of this boom are manifold: in data centre hubs such as Frankfurt, electricity is already in short supply, leaving other sectors little chance of implementing their electrification plans. Moreover, a shortage of grid electricity is not deterring many Big Tech companies from building new data centres that consume as much energy as entire cities. Fossil gas is then usually the energy source of choice. These CO2 emissions add up to a power supply for our clouds that, due to the global electricity mix, still relies largely on fossil fuels anyway.
However, without a drastic shift towards 100 percent locally generated energy, the data centre industry will be unable to keep pace with the general climate protection commitments of the Paris Agreement.
Our clouds are thirsty
However, energy consumption and the associated CO₂ emissions are only one aspect of sustainability assessments. Added to this is the high water consumption of data centres. It is estimated that the new data centre in Oregon will account for around a third of the city’s total water consumption.
Little attention is paid, however, to the hardware aspect. The manufacture of hardware for data centres has, unsurprisingly, a significant carbon footprint. In particular, the AI boom requires increasingly specialised, energy-intensive chips, which are replaced on average every two to three years. As a result, emissions from the manufacture of IT infrastructure have skyrocketed in recent years.
So far, we can conclude that our cloud infrastructure as a whole is responsible for rising CO2 emissions as well as further damage to the environment. But what about the sustainability of the cloud services we use as individuals or businesses? How do we determine what role we play in these negative impacts, and how can we influence them?
To answer these questions, we need to cut through a lot of fog.
Navigating the murky fog of the carbon footprint
In reality, it is virtually impossible to accurately assess the carbon footprint of a cloud service. This is primarily because very few cloud providers offer detailed insight into environmentally relevant data such as energy and water consumption and hardware emissions.
Every provider uses different metrics, a different definition of ‘renewable energy’ and a different level of transparency. This allows cloud service providers, for example, to claim that a data centre is ‘100 per cent renewable’, even if the servers in question are powered by a coal-fired electricity grid. This is made possible by purchasing, for example, wind power certificates from a location far away.
The big tech companies dominating the market – Amazon Web Services (AWS), Microsoft and Google – all claim they want to become climate-neutral or at least drastically reduce emissions. However, they grade their own climate commitments.
As a result, differing definitions obscure the actual impact. This is not merely a technical issue. For businesses and end consumers, the lack of data is a major obstacle to keeping track of their carbon footprint, and it also makes compliance with regulations more difficult.
Calculating CO2 emissions: Cloud service providers’ tools remain unclear
But hang on – don’t many cloud service providers offer their own tools that allow customers to calculate the carbon emissions of the services they use? “Providers cannot ignore the trend that more and more customers want information about their carbon footprint. That’s why they offer simple tools designed to meet this need,” says Jens Gröger, Senior Researcher on sustainable digital infrastructures at the Öko-Institut, in an interview with RESET. They might be useful as a very rough initial estimate. “But the figures are all very imprecise, as they are merely estimation tools based on the execution time of the programmes.”
To really find out where we can optimise and which services are particularly energy-intensive, we need more precise data. “It would be far more helpful if I received immediate feedback from the server in the data centre, telling me my actual measured energy and resource consumption,” says Gröger. After all, it is not just the CPU or GPU runtime – that is, the operating time and the load on the respective server – that determines the level of CO2 emissions. It is also important at what time of day and in what weather conditions our calculations are carried out.
If the sun is shining or the wind is blowing, we can assume that the data centre is primarily powered by renewable energy and that CO2 emissions are low. And it also matters what other resources are being used. This includes, for example, how much data is flowing through the network and how much RAM we are tying up through our cloud activities. None of this is factored into the current tools provided by cloud providers.
ECO:DIGIT demonstrates how the environmental impact of software can be measured
It is not just data that is now processed via cloud networks – our software’s computational operations are also increasingly carried out via distributed servers.
The ECO:DIGIT project, run by the German Informatics Society, highlights the challenges involved in measuring the environmental footprint of our software and has developed a methodology for doing so.
Demanding transparency and standards
Due to a lack of data and the inaccuracy of what data is available, we can only speculate about the environmental impact of our clouds. And this is unlikely to change as long as there are no standardised definitions or labels, nor effective guidelines for data centres. Transparency and reporting requirements are therefore key to obtaining more data – and we must demand them.
IT leaders, for example, can request detailed, location-specific data from their cloud providers on the carbon intensity of the electricity grid where their data is hosted. And we should all press for industry-wide guidelines and standards to be made mandatory.
We also have the option to choose alternatives.
Green data centres: Finding a sustainable home for our data
As long as we entrust our cloud services to Amazon Web Services (AWS), Microsoft or Google, the actual emissions generated by our services remain hidden due to a lack of transparency and empty promises. However, the situation is different when we opt for smaller, European providers and green data centres.
One such alternative is Infomaniak, for example. The cloud provider’s data centre is located beneath the car park of the La Bistoquette eco-housing cooperative in Geneva. It supplies 6,000 households with heating all year round, relies entirely on renewable energy and avoids electronic waste.
Although the majority of data centres are still a long way from incorporating sustainable practices into every stage of their operations, green data centres can be found, particularly in Europe. The Green Web Foundation helps with the search via its Green Hosting Directory.
Why PUE says little about the sustainability of a data centre
The PUE (Power Usage Effectiveness) is a metric often used as a sustainability indicator for data centres. However, this metric only indicates how much of the energy consumed is lost in the building services. The decimal place describes these losses.
This means that PUE is not the right factor for truly determining a data centre’s efficiency, says Jens Gröger. “The more energy that flows into the IT systems, the better the PUE becomes, because the cooling system is then utilised more effectively. The PUE does not reflect how efficiently the computers are processing data.”
To assess the efficiency of a data centre, therefore, further metrics are required, such as water use efficiency (WUE), the proportion of waste heat utilised (ERF) and the average server utilisation (ITEUsv).
Unfortunately, not all green electricity is created equal. When choosing our data centres, we should therefore also ensure that genuinely renewable energy sources – ideally from the local region – are used. This is because companies often simply purchase green electricity certificates, whilst in reality, fossil fuels are powering their operations.
A data centre’s carbon footprint improves even further if waste heat is utilised, the IT infrastructure is operating at full capacity, and the equipment is used for long periods.
“Companies can ask their service providers for these figures,” says Jens Gröger. And the public sector could issue tenders that set specific minimum requirements for data centres. “As a private individual, however, I often have no choice but to look at the website and find that there is no information apart from perhaps a self-designed green electricity logo.”
SUSFECIT – Shifting computing workloads to where green electricity is currently available
What if we always shifted particularly energy-intensive computing processes to data centres that are currently being powered by renewable energy? This is precisely what researchers in the SUSFECIT (Sustainable Federated Compute Infrastructures) project are working on.
The idea behind this is that various research data centres, spread across Germany, always run their computing tasks at the data centres where the proportion of green electricity is currently particularly high. To ensure that the loads can be coordinated, the data centres are interconnected.
Clearing the fog with alternative cloud solutions
It is not just the choice of data centre, but also the software that determines the level of cloud emissions. The current ‘bigger is better’ mentality, fuelled by generative AI, is unfortunately heading in the wrong direction. Running a model with a trillion parameters to summarise a simple internal document is like using a sledgehammer to crack a nut. Yet this is precisely how Big Tech providers, in particular, are developing their cloud services.
There is another aspect to consider: the walled garden effect. “When customers opt for Microsoft products, it’s clear that they’re effectively locked into a Microsoft data centre – and then can’t get out,” says Jens Gröger. Cloud providers also decide which services are compatible with their clouds. As a result, we often have no option to choose independent, green or democratic services.
“In that respect, perhaps the first step is to free ourselves from proprietary software providers. Only then do I have real freedom of choice regarding where my software runs and my data is stored,” says Gröger.
Open-source cloud solutions such as Nextcloud, Open Desk or Leaf Cloud can be adapted to your own needs and further developed. And we can run them on servers that meet high environmental and transparency standards.
Digital sovereignty in the cloud
Sovereign Cloud Stack (SCS) is a European initiative that aims to create an open, transparent and vendor-neutral cloud ecosystem that guarantees sovereignty.
It defines a ‘cloud stack’ that can be installed in one’s own, medium-sized or large data centre. As a counterbalance to the major cloud providers, customers have the freedom to choose whether they wish to go with Telekom or Plusserver, for example. The focus is on cloud technology with genuine freedom of choice.
Establishing data-efficient routines
Of course, the amount of data we store in our personal digital libraries also affects CO₂ emissions. The more data we store, the more storage space we use. And this requires a constant supply of energy. In this context, Jens Gröger points out that saving data also makes financial sense, since we pay our cloud providers based on the number of terabytes we use.
So why not declutter? Our devices are usually always full of old backups, documents and emails. In individual cases, our obsolete data may amount to a manageable quantity, but taken together, it quickly adds up to vast volumes.
Cleaning up your data once a year can certainly help. Of course, it’s even better not to let data clutter build up in the first place and to establish data-efficient routines right from the start. There are plenty of ways to do this: don’t keep sending the same document back and forth via email; instead, save it straight away and just send the link; use lightweight software; and avoid bloating documents with images and videos.
Out of the fog and into the green cloud
The path to a sustainable cloud is thus clearly laid out before us: the first and second steps are choosing a green data centre and alternative cloud providers. By doing so, we have already left the walled garden of Big Tech behind. Then we can get started with the fine-tuning. This involves establishing data-efficient practices. “It’s about handling the data I produce, manage and have to store in a conscious manner. And that also applies to the entire software layer,” says Gröger.
The search for sustainable data centres is not made easy for us at present, as many operators are keeping us in the dark. That is why, as customers and businesses, we must demand transparency and effective regulations at every level.
