Image: Flues of data centers punctuating the skyline in a rural district of Brandenburg.
Data centres are specialized facilities designed to store, process, and transfer vast amounts of information. They serve as the backbone of distributed computing, enabling a wide range of activities: from supporting scientific research and industrial operations to delivering content for social media, gaming, and video streaming; from powering artificial intelligence by training and running machine learning models to facilitating financial transactions and managing critical services for governments and businesses. Although we associate data centres with high-end computing, they serve a wide range of purposes--from cutting-edge science to the most mundane applications. For example, a computing centre in Nuremberg prints approximately 40 tonnes of paper each day, corresponding to tax data provided—and transfered via the internet--by tax consultants in Germany.
As societies increasingly rely on computer- and data-driven services, data centres have come to be regarded as critical infrastructure, ensuring the continuous flow of information around the clock. However, there is also growing concern about their use for non-critical purposes, as well as about the significant role these facilities play in speculative financial bubbles built on inflated claims and vague promises of innovation.
A typical hyperscale data centre may look somewhat like this:
Data centres are specialized facilities designed to store, process, and transfer vast amounts of information. They serve as the backbone of distributed computing, enabling a wide range of activities: from supporting scientific research and industrial operations to delivering content for social media, gaming, and video streaming; from powering artificial intelligence by training and running machine learning models to facilitating financial transactions and managing critical services for governments and businesses. Although we associate data centres with high-end computing, they serve a wide range of purposes--from cutting-edge science to the most mundane applications. For example, a computing centre in Nuremberg prints approximately 40 tonnes of paper each day, corresponding to tax data provided—and transfered via the internet--by tax consultants in Germany.
As societies increasingly rely on computer- and data-driven services, data centres have come to be regarded as critical infrastructure, ensuring the continuous flow of information around the clock. However, there is also growing concern about their use for non-critical purposes, as well as about the significant role these facilities play in speculative financial bubbles built on inflated claims and vague promises of innovation.
A typical hyperscale data centre may look somewhat like this:
The observable exponential growth in the data centre business comes with significant environmental challenges. Running thousands of computing centres simultaneously worldwide requires vast amounts of energy, as well as extensive land and water resources--often in a scale that is difficult to quantify only locally. Estimates are that a large-scale data centre can require anything from 200 to 400 terawatt-hours of electricity annually, while cooling methods may demand up to 1.1–1.9 billion litres of water per year, equivalent to a small city.
The fictional, AI-generated image of a data centre below (the flying, giant cameras were not in the prompt) was processed within a data centre, and its creation alone is estimated—by the very same model that created it—to have emitted approximately 2.2 grams of carbon dioxide. This is roughly equivalent to fully charging a smartphone. As most composite estimation nowadays, this number excludes the emissions that were produced during the training of the models.
The fictional, AI-generated image of a data centre below (the flying, giant cameras were not in the prompt) was processed within a data centre, and its creation alone is estimated—by the very same model that created it—to have emitted approximately 2.2 grams of carbon dioxide. This is roughly equivalent to fully charging a smartphone. As most composite estimation nowadays, this number excludes the emissions that were produced during the training of the models.
However, the generation of a single image cannot reasonably represent the scale of data centre operations or their growth trajectory. According to Data Centre Map (DCM), a website that tracks data centre services, Germany had 170 commercial data centres across 31 cities in 2014. Today, that number has grown to 403 operational data centres spread across 57 cities. Similarly, Switzerland had 59 data centres in 16 cities a decade ago, compared to 109 facilities in 24 locations today.
In 2023, Europe experienced the largest increase in new cloud regions globally, with eight new zones dedicated to hyperscale data centres. Despite this expansion, Europe remains far behind the world’s largest data centre market—the United States—which has 3,390 commercial data centres listed on DCM. However, it is important to note that these figures do not fully capture the scale of the global data centre industry. Most big tech companies do not have all their data centres listed on DCM, and many countries lack transparent data on the number of facilities within their borders, making the obtaining of comprehensive assessments very difficult.
In 2023, Europe experienced the largest increase in new cloud regions globally, with eight new zones dedicated to hyperscale data centres. Despite this expansion, Europe remains far behind the world’s largest data centre market—the United States—which has 3,390 commercial data centres listed on DCM. However, it is important to note that these figures do not fully capture the scale of the global data centre industry. Most big tech companies do not have all their data centres listed on DCM, and many countries lack transparent data on the number of facilities within their borders, making the obtaining of comprehensive assessments very difficult.
Image: Cooling infrastructure behind a data centre.
As the demand for geo-distributed data processing continues to grow, with digital technologies assuming an almost species-like role in today's anthroposphere, addressing the environmental pressures associated to this growth is crucial.
And it is not just a matter of "environmental impact": one of the trickiest problems of understanding the environmental entanglements of data centres may be in fact a matter of aesthetics. As they function as an infrastructure of digitised societies, with a growing number of both human and non-human users, these facilities remain largely opaque in everyday experience. Even if estimates are that 85% of all the data exchanged over the internet by organisations or businesses is based on cloud computing, until very recently most people were unaware of these infrastructures, and not only because the very concept of the “cloud” creates the illusion of computing as something intangible and environmentally friendly. Data centres are also most often located at exurban perimeters, nearby regional airports or remote train stations (and sometimes even within them), so they tend to go unnoticed. Plus, when discussions of anthropogenic climate change arise in public discourse, the environmental pressures associated with digital technologies are rarely highlighted.
As the demand for geo-distributed data processing continues to grow, with digital technologies assuming an almost species-like role in today's anthroposphere, addressing the environmental pressures associated to this growth is crucial.
And it is not just a matter of "environmental impact": one of the trickiest problems of understanding the environmental entanglements of data centres may be in fact a matter of aesthetics. As they function as an infrastructure of digitised societies, with a growing number of both human and non-human users, these facilities remain largely opaque in everyday experience. Even if estimates are that 85% of all the data exchanged over the internet by organisations or businesses is based on cloud computing, until very recently most people were unaware of these infrastructures, and not only because the very concept of the “cloud” creates the illusion of computing as something intangible and environmentally friendly. Data centres are also most often located at exurban perimeters, nearby regional airports or remote train stations (and sometimes even within them), so they tend to go unnoticed. Plus, when discussions of anthropogenic climate change arise in public discourse, the environmental pressures associated with digital technologies are rarely highlighted.
Image: Airplane flying over a data centre located next to an airport.
Since intensive, 24/7 computing generates a lot of heat, air conditioners and generators are necessary to keep server halls constantly refrigerated and computer processing stable. Due to this, data centres have been said to function as thermo-cultural mediators, managing thermal interactions between culture and nature, computer halls and external environments.
Considering this context, how can we make sense of the thermal mediations of planetary-scale computing infrastructures? How can we explore these relationships as an aesthetic experience? One suggestion may actually be to tune our senses to the materialities of digital technology. Researchers and artists have been doing this in several ways: some have been cartographing the life cycles of hardware parts, others have been aestheticising the images of server halls and the cooling infrastructures of data centres, or the water pipes that conduct fresh water from riverbeds to their interiors. Another idea, perhaps less explored, is to give the data centres a listen!
Since intensive, 24/7 computing generates a lot of heat, air conditioners and generators are necessary to keep server halls constantly refrigerated and computer processing stable. Due to this, data centres have been said to function as thermo-cultural mediators, managing thermal interactions between culture and nature, computer halls and external environments.
Considering this context, how can we make sense of the thermal mediations of planetary-scale computing infrastructures? How can we explore these relationships as an aesthetic experience? One suggestion may actually be to tune our senses to the materialities of digital technology. Researchers and artists have been doing this in several ways: some have been cartographing the life cycles of hardware parts, others have been aestheticising the images of server halls and the cooling infrastructures of data centres, or the water pipes that conduct fresh water from riverbeds to their interiors. Another idea, perhaps less explored, is to give the data centres a listen!
|
|
Image: Different air-conditioning systems for data centres, shown from aerial and lateral angles.
While data centres typically appear nestled in industrial districts and logistical hubs, their visual opaqueness is contradicted by the loud humming that can often be heard—sometimes near the facility, sometimes a few kilometers away. These sounds serve as acoustic performances of the operational rhythms of the cloud as a socio-technical infrastructure. The humming becomes more prominent when artificial cooling is in use—mostly during summertime (data centre ecosystems may be artificial, but they also have a seasonality). Ultimately, the noises coming from these hermetical grey boxes represent not just the workings of a single facility, but the larger metabolism of heat dissipation involved in cloud operations, when data from geographically distant areas is processed and transferred continuously.
ATTUNING TO THE HEAT DISSIPATION OF COMPUTING
In order to make this infrastructure more relatable, we can make use of an often-neglected sense in our aesthetic experience of digital technology: listening. This is for a very simple reason. Servers and pipes may hide behind the solid walls and brutalist concrete and iron architecture of data centres, unseen, but they still generate a great deal of noise: continuous, sometimes strange sounds, hums and whistles of various tones, etc.
|
Considering this, let us try something a bit more experimental:
1. Click on the button and play the album in the link. Close your eyes. Listen. What do you hear? 2. How can you design a navigable environment to invite listeners to walk in the dark through the geographies of different data centre technoscapes? * The recordings available at the link are free to use for prototyping and experimenting with the project, but they remain provisional, for the sake of illustration. A new round of data collection will take place in the coming months, after which longer and higher-quality soundscape recordings will be provided for the project. |
This is the sound of large-scale data centres. Well... of just a few of them. Most of these sounds were recorded in the vicinities of data centre campuses in Hesse, Brandenburg or Zurich, but there is a wider aural diversity to consider. The sounds emanating from the facilities are already distinct from one another, but they also blend with the vibrant expressions of their surroundings: cars on exurban highways, construction sites with workers, remote rural landscapes, flocks of birds near riverbeds, and logistical districts. They form a fusion of nature and the 'second nature' of built environments—they are technoscapes.
How can one experience a virtual sonic journey in this context, where the sounds (and the heat dissipation) of escalating data centre landscapes merge with the everyday acoustics of urban, rural, or industrial environments? The VR experiment can speculate with these technoscapes in several ways. Here are a few non-exhaustive suggestions:
1. Experiment creating a blind-friendly tour through a single street with different data centres, using the spatial affordances of audio: sounds may guide us geographically, through the direction of the emitting source, but can also confuse us, with echoes, ground noises, and other dynamic spatial effects. Keep in mind that ultra-high and low pitch sounds can evoke strong, bodily sensations.
2. Try imagining a virtual exploration of data centre audioscapes, but try to experiment with them in wider spatial settings. The focus would preferably be on the affective, sonic experience, so realism is not the goal here, and you may feel free to use geometric forms instead of realistic images if you would like to provide some visuals, for instance. You may also consider looping the sounds and mixing them with the soundscapes of other environments. Sound libraries and audio databases can be used creatively for this end as well.
3. Alternatively, you may instead produce a piece of speculative fiction: imagine the point at which the energy demands of growing, large-scale computing systems exceed even the extraordinary modern energy-efficiency measures of today, and cities can no longer accommodate further cloud infrastructure growth, given the already saturated computer-processing environment. What would the sonic experience of such a data centre-dense city feel like? How to tell a sonic fictional story of this environment in a VR experience?
How can one experience a virtual sonic journey in this context, where the sounds (and the heat dissipation) of escalating data centre landscapes merge with the everyday acoustics of urban, rural, or industrial environments? The VR experiment can speculate with these technoscapes in several ways. Here are a few non-exhaustive suggestions:
1. Experiment creating a blind-friendly tour through a single street with different data centres, using the spatial affordances of audio: sounds may guide us geographically, through the direction of the emitting source, but can also confuse us, with echoes, ground noises, and other dynamic spatial effects. Keep in mind that ultra-high and low pitch sounds can evoke strong, bodily sensations.
2. Try imagining a virtual exploration of data centre audioscapes, but try to experiment with them in wider spatial settings. The focus would preferably be on the affective, sonic experience, so realism is not the goal here, and you may feel free to use geometric forms instead of realistic images if you would like to provide some visuals, for instance. You may also consider looping the sounds and mixing them with the soundscapes of other environments. Sound libraries and audio databases can be used creatively for this end as well.
3. Alternatively, you may instead produce a piece of speculative fiction: imagine the point at which the energy demands of growing, large-scale computing systems exceed even the extraordinary modern energy-efficiency measures of today, and cities can no longer accommodate further cloud infrastructure growth, given the already saturated computer-processing environment. What would the sonic experience of such a data centre-dense city feel like? How to tell a sonic fictional story of this environment in a VR experience?
All images and soundscapes on this landing page by Bibiana da Silva de Paula and Eduardo Luersen.
