Collapse Informatics, Small Files, and the Light Infrastructure of the Future

Networked media have a shockingly high and fast-growing environmental footprint. By my calculation, because most of their electricity derives from fossil fuels, these technologies contributed almost five percent of global greenhouse emissions in 2025 – double the carbon footprint of the airline industry.1 And it may be more. For I confess that doing these calculations, based on incompatible statistics released by agenda-driven organisations like the International Energy Agency (IEA) and divergent methods published by engineers, is like trying to train hedgehogs to dance dabke. Networked media, otherwise known as information and communication technologies (ICT), consist of data centres (aka servers, the ‘cloud’), networks, and devices. The most widely publicised contributor to ICT’s environmental damage are data centres, which suck up water and electricity to store files and carry out calculations or process data: for example, encoding and transcoding videos and carrying out machine-language calculations. Less well known is that devices have a large and growing carbon footprint. Much of this footprint occurs in manufacturing. Chip production is very energy-intensive, so the fancier the phone (or gaming device, computer, or smart TV), the larger its carbon footprint, even before it leaves the factory; and these factories in East Asian countries draw the great majority of their electricity from fossil fuels.2

Device use is also electricity intensive: that’s why phones and computers feel hot to your hands as they carry out the hard work of video decompression and AI inference. Under the pressure of planned obsolescence, discarded devices pollute landfills. Networks, the third component of networked media, tend to be the leanest in electricity consumption, but they too are bloating as data demand increases. Fifth-generation or 5G cellular networks are an energy vampire and spew toxic radiation.3 The next most wasteful type of network is the layers of 4G and legacy 3G and 2G networks. 

Online video makes up most of the networked media traffic, because people around the world spend hours a day streaming video – in Arabic-speaking countries YouTube tops the list of social channels, followed by social-media and gaming platforms.4 The carbon footprint of networked media is quickly rising as more global citizens stream video in high resolution, play high-resolution games, and become – seemingly overnight – dependent on AI models. These large language models are extremely energy-intensive not only in their training but also in the individual inferences of billions of users.5 Indeed, the environmental impact of AI is what first alerts people to the toxic materiality of networked media. However, it is streaming video that established the infrastructure of high-capacity data centres and servers to store and process large files, massive broadband networks to carry them, and overly sophisticated devices to receive them.

The corporate-driven fantasy of net neutrality holds that consumers worldwide should have the instantaneous access and high bandwidth that rich countries take for granted. But it’s planetary suicide for wealthy, highly infrastructured countries that rely heavily on fossil fuels to be models for the rest of the world’s networked media. Instead, I argue that the wealthy parts of the world must model sustainable infrastructures on those of poorer countries. 

Let us make no mistake: renewable energy is not going to save the planet from this catastrophe. As the rate of electricity consumption skyrockets worldwide, it becomes evident that renewable energy will not be a substitute for fossil-based energy. ‘Smart’ cities, behemoth electric SUVs, brain-destroying AI applications, cryptocurrency (strictly, not part of ICT, as its mining is not a communication technology), and ever more applications of online video – like 8K wall TVs, high-resolution video conferencing, live 8K surveillance video, and streaming virtual reality – are not clean even if they feed on renewable energy (which in any case is often spuriously supplied by carbon credits, not real reductions in fossil-fuel use).6

In the disastrous scenario of the conservative International Energy Agency, worldwide industry will need renewables and natural gas and nuclear power on top of dirtier fossil-fuel extraction to accommodate rising demand.7 And as well as extracting the toxic dregs of fossilised plants and animals (which we should revere by leaving them in their graves), the energy industry is stripping the Earth for copper and rare minerals to supply materials for wind and solar power. 

In addition, renewable energy is a new weapon of war, as Lawrence Abu Hamdan shows in his chilling installation Zifzafa (2024), based on an Israeli plan to erect 31 enormous wind turbines in the Jawlan, which Abu Hamdan and other members of the Earshot collective analysed. Through animation and a video game, Abu Hamdan dramatises how Jawlani residents would experience these towering wind machines – 256 metres high) – erected illegally close to residences on land currently used for sustainable agriculture. 

The giant whirring blades occupy the entire sky. Deafening winds take the very air hostage. The turbines’ noise, at 70 to 80 decibels, is as loud as a hair dryer held close to your ear for 24 hours a day. Wind power is yet another Israeli method to evict Palestinians from their land. (I wonder if one of Israel’s motives for the genocide in Gaza is to have greater access to the Mediterranean for offshore wind development.)8

It is time to recognise that streaming 8K video and using AI for everyday tasks are destructive fantasies that will turn our planet into a ball of fire. It is time to scale back and prepare for a time soon – that is, now – when the human community will have to radically reduce electricity consumption. Ideally this would happen before planetary heating reaches a tipping point, because afterwards there will be no choice. It is time to embrace collapse informatics: sociotechnical design in preparation for future scarcity. ‘We envision the possibility of a future characterized by scarcity and shrinking opportunity,’ Bill Tomlinson and colleagues from Computing within Limits write, ‘and seek to make that future a better place.’9

In terms of networked media, Arabic-speaking countries are among the world’s worst environmental criminals, but they are also decent examples of sustainability. Saudi Arabia, with the environmental catastrophe that is NEOM and its goal to be a global gaming hub, leads on the path to climate catastrophe, followed closely by Kuwait, Qatar, and the UAE: after Iceland and Norway, the top global consumers of electricity per capita in 2024.10 Wealthy Arab countries are developing renewable energy, but it remains a pitifully small contribution to their unsustainable electricity consumption.

Lightly infrastructured countries that use a lot of renewable energy, like Kenya, should be our model for sustainable networked technology. Currently Arab countries that use little electricity do so because they are ravaged by war: Syria, Yemen, Sudan, and South Sudan. Interestingly, poor countries are more on track for renewable energy, at least partly because they lack infrastructure fed by fossil fuels. Can you get your head around the idea that Sudan could be a model for sustainable IT infrastructure? 

Sudan receives most of its electricity from large-scale hydro power, though its exploitation of fossil-based energy is increasing. Unfortunately, as Maral Mahlooji and colleagues point out, large-scale hydro has very damaging impact on water and land, but Sudan could produce concentrated solar power, geothermal, or nuclear power.11 Incapacitated by years of war, Sudan’s infrastructure is light, but adequate and sustainable infrastructure could be constructed using the most efficient new technologies. 

Fat chance, I hear you saying. Who is going to give up high-resolution streaming and ChatGPT replacing our own brains once they’ve tasted those luxuries? My best hope is that government regulation will require that the environmental cost of networked media be factored into the cost of data plans, so that consumers think twice before streaming high-resolution videos and games and using energy-intensive AI apps. My darker thoughts circle around some kind of planetary disaster that would radically cut demand, but let’s not go there.

I founded the Small File Media Festival to draw attention to bloated ICT infrastructure and show that streaming video can be satisfying at a tiny fraction of standard bitrate. At the Small File Media Festival, we invite filmmakers to submit works of extremely low bitrate. We specify 1.44 megabytes (the amount of data that fits on a floppy disk) per minute. Ideally, they are filmed on older devices with relatively low resolution, or with reduced resolution on newer devices. Some are composed entirely of still images, some mostly of text. Some use animation, which has many techniques that look good in a small file. A rich soundtrack compensates for the thinness of the image. Artists employ compression skilfully, using light platforms like Handbrake or FFMPEG.

Small-file movies have no need for massive infrastructure. They travel on 2G networks, by text, and by email. They are ideal for regions where infrastructure is light, and electricity is intermittent. They are the sustainable avant-garde. Small-file movies are usually a bit soft-focus and require the viewer to lean in to experience them. Compression allows them to hide information in the image: information that was legible in the original becomes occulted in compression, only legible to viewers already in the know. Thus, the small file is useful for those who want to evade surveillance.

We at the Small File Media Festival organised small-file workshops led by local artists in Beirut, Cairo, Dhaka, Kashmir, Mexico City, the Sudanese diaspora, Tehran, and Zanzibar to share the techniques with other makers and teachers. These works followed the two main tendencies among small-file makers. Some aim to make videos whose visual quality is as high as possible – a good approach for narrative films. Others experiment with compression aesthetics, producing works that bristle with boxy pixels and flow in the juicy trails of datamoshing. They are creating small-file movies as the avant-garde of sustainable digital media.

The resulting works were thrilling. Many of the works made in the Tehran workshop, led by Mehdi Shiri at Bidar School, were animated films, which defy the limits of form and can also originate as small files that need little compression. The jarring colours of Neda Sangtarash’s Bougainvillea (2025) lost no vividness in compression, nor did their mind-bending transformations: for example, a woman in underwear crunches a yellow cucumber that turns into an irritated and hungry rat. 

Other works celebrate the way compression can occult the image. The compression of Amal Shafek’s video Snow (2024) – created at the Cairo workshop led by Mena El Shazly at Medrar – conceals the filmmaker’s nakedness in soft pinkish pixels, and because of the haptic qualities of the small file we can almost feel the snow on her skin. 

Meanwhile documentary and narrative small-file works of decent image quality led willing audiences along paths of imagination. Enjoying these movies, we collectively practise collapse informatics and find that it need not be painful. Of course, if the military-industrial complex were to disappear overnight, the planet would cool down very quickly and we could stream all the videos we want.