DIGITAL MIRROR: Water flows, data streams

Naturally, “the cloud” evokes the idea of a delicate lightweight presence, even though its actual workings rely on a gigantic physical infrastructure, with data centers at its core. Far from weightless, these huge, warehouse-like buildings, host endless rows of servers, routers and switches that process, store and share our data 24/7. Every time we call upon the services of our digital companion, we feed on a cascade of computational processes within a data center, filled with the hum and heat of thousands of calculating machines.[2]

Unsurprisingly, it takes gargantuan amounts of energy to run such a facility. U.S. data centers alone consume the electricity equivalent of 17 nuclear power plants today. The main drivers behind this enormous demand are data-intensive, cloud-based services like social media platforms, online apps, streaming services but also, and in particular, the exponential growth of artificial intelligence (AI). From chatbots like ChatGPT, image generators like Dall-E and the increasingly AI-heavy product lineup of Google, Microsoft and Meta, to new AI-applications within the fields of logistics, finance, medicine or warfare—they all rely on energy-intensive operations in data centers to train and run their AI-models. In the process, legions of high-performance computers digest trillions of data points to find patterns in texts and images that can then be used to generate more or less meaningful content. The amounts of electricity needed are baffling and, as a consequence, some data-center-dense countries, such as the U.S., India or the Netherlands, are facing strained grids already.[3]

But as it seems, the craving for energy won’t halt: Exacerbated by the current AI boom, the global electricity demand of data centers will more than double in the near future, reaching levels comparable to entire nations such as Japan.[4] Worse, a large part of the required electricity still relies on fossil fuels. And although the dirty side of nondescript server racks “might be far less visible than the billowing smokestacks of coal-fired power stations,”[5] their environmental footprint is evidently significant, growing fast, as AI researcher Kate Crawford stresses.

Welcome to the water castle
In need of cleaner energy sources and market expansion, the data center industry decided to set foot on Swiss soil. Situated amidst the Alps, the country is characterized by spectacular mountains and abundant water resources; an oasis of streams, rivers and lakes that traverse its steep slopes and rocky landscapes. These constant flows of water, fed by hundreds of glaciers, provide the perfect conditions for the mass production of hydropower, making Switzerland not only the “water castle of Europe” but turning it into a seemingly “inexhaustible alpine battery” full of clean energy. 

Attracted by these readily available hydro resources, data centers are mushrooming everywhere in Switzerland.[6] Google, Microsoft and Amazon have all plugged-into the Swiss landscape, and national and multi-national cloud-suppliers set up shop too.[7] Many of them highlight their sustainability certificates proudly on their websites telling us that their “data center is more than just a place for your data. It is a showcase project for environmental protection.”[8] In parallel, pictures of green data clouds hovering over glaciers, lakes and waterfalls on their Instagram accounts promote “the highest sustainability standards”. Switzerland, then, the new green data-hydro castle of the world? 

In fact, each data center powered by Swiss “green energy” taps into a monumental infrastructural network, which started to evolve more than 150 years ago—a time when hydropower was the country’s only way towards electrification and industrial progress. Lacking any fossil fuels, Switzerland capitalized on its topography and reengineered its watery landscapes to power the country: dams were erected, rivers channeled, and turbines installed. Driven by energy-hungry industries and growing towns, almost 90% of Switzerland’s electricity was water-based by 1970.[9] In the process, small run-of-river plants were enhanced by record-breaking alpine megadams, ready to be featured in James Bond movies.[10] In many ways, those massive hydroelectric infrastructures drove forward Switzerland’s version of hydromodernity[11]—a term coined by geographer Erik Swyngedouw that describes the operationalization of water flows to make landscapes productive and secure economic growth. Here, the “liquid power” of state-lead water-management creates the promise of an endless energy supply for profitable ventures and national progress.

Today, hydromodernity lives on in Switzerland—with a staggering number of 188 dams, many higher than 100 meters, which stop and store billions of liters of rain- and meltwater in their reservoirs. Whenever energy is required, gates are opened and some water is dashed down into titanic turbines that generate electricity. This way, almost half of Switzerland’s white gold is produced, guaranteeing a stable supply of inexpensive energy for citizens and industries alike. Understandably, data center operators are eager to utilize this seemingly clean and renewable energy source for their electricity-hungry operations. In doing so, their AI-inflected visions give Switzerland’s hydromodern self-conception a major update: Now, the country’s perfectly managed water flows power not only the growth of chemical and metal industries of extra-human scale, but also the infinitely repeating stacks of servers that run up-and-coming AI superpowers. And while the specific consequences of this entanglement are mainly kept secret, they come at multiple costs. Dams, as infrastructural key element for electricity production, play a major role in this.

Anti-rivers and data dams
When dams block the flow of rivers at will, they become what environmental scholar Lisa Blackmore calls “antitheses of rivers,” because rivers “flow in different volumes and intensities, but always flow.” Thereby “their fluctuating courses ‘rise and fall like a heartbeat.’[12] This heartbeat is what dam structures alter and stop.”[13]

Swiss authorities have tried to mitigate some of the consequences of these broken beats, yet the socio-ecological impacts of dams are still severe: To build them, local communities are displaced and landscapes are literally cast in kilotons of concrete—a material responsible for “limestone mining, massive sand removal, release of harmful particles, rubble, and ruins.”[14] Once in place, their impenetrably thick walls destroy downstream habitats and endanger fish populations,[15] while their reservoirs lose freshwater to evaporation. And if this wasn’t enough, the reservoirs produce methane too, a potent greenhouse gas that contributes to global warming and is emitted when underwater microbes feed on organic matter that accumulates at the reservoir’s bottom.[16]

Facing these far-reaching consequences, it’s evident that the environmental impact of water-power is far from negligible.[17] Hence, thinking of hydroelectricity solely in terms of “zero-emissions”, “clean” and therefore infinitely scalable, does not only uphold the paradigm of endless resource extractivism, it will without a doubt get us into trouble in the long run. 

Nevertheless, Swiss data center operators are branding themselves as eco-friendly enterprises driven by an abundance of “sustainable” waterpower. At the same time, any environmental externalities created by their exponentially growing, energy-hungry data operations are swept under the table. A case in point are hyperscalers, the latest version of data centers whose massive size provides capacity for more servers and AI clients.[18] Labelled as highly efficient, shining sustainability clouds, a crucial detail is omitted within the euphoric press releases: a single facility consumes as much electricity as a city[19] and its red-hot servers are cooled-down with immense amounts of evaporating freshwater.[20] In order to generate the necessary “green energy”, the material geographies of alpine landscapes need to be reengineered, dammed and damaged even more.

Apparently, Switzerland is willing to do so. In 2021, authorities have announced 15 new large-scale hydropower projects to cover its surging electricity demands. Many dams, reservoirs and power plants will be enlarged or built from scratch until 2040.[21] Not even the iconic Matterhon will be spared from an 85-metre high barrage nearby its peak.[22] Once again, the Swiss state unleashes its “liquid power” to advance economic progress through brute force.[23] In return, data clouds hover over Switzerland’s damns and mountain ranges, giving birth to new landscapes of liquidity: flows of water produce data streams that generate financial liquidity. In the end, this water-turned-capital is geographically re-allocated to other cities and states,” drifting far away from its drained sites of production.

Cloud blackouts
On the horizon of those liquid landscapes, another dilemma is looming large: the water castle is running dry. Driven by a warming climate, most of Switzerland’s 1,400 glaciers are expected to vanish by 2090. With much less water and changing peak flows coming from the alpine giants in the long-run, power companies will be forced to adjust their output seasonally.[24] Experts predict that the decreasing water supplies will further heat-up conflicts over Swiss water,[25] with farmers and energy companies in the frontlines, who both depend on vast amounts of the precious resource.

In a draining water castle, one question becomes urgent: Who will be allowed to tap into the remaining water and energy flows, especially in times of scarcity? As it stands today, the Swiss data center industry doesn’t have to worry about such scenarios that could enforce a data drought. Classified as “critical infrastructure”, their hyperscale facilities wouldn’t be affected by any electricity rationings. Since they provide “essential services” for hospitals and communities, authorities are unlikely to take them off-grid.[26] However, looking at their portfolios, one can observe that services less essential for survival are ran by hyperscalers too: car dealers, shopping platforms, generative AI tools. All of them are guaranteed stable energy supply should water capacities hit all-time lows, whilst alpine farmers might be forced to quit their jobs and citizens have to sing the 2-minute shower song.[27] The reason for this prioritization is clear: Data centers have become the pillars of the digital economy, they are not only needed for essential services, but also to generate big revenues.[28]

As companies race to build the infrastructures necessary to run AI services, we need to ask who profits, pays or suffers from the green clouds that increasingly cover Swiss skies. Artistic projects like Yu Sang’s Cloud Calculator or Niels Schrader’s and Roel Backaert’s Acid Clouds can help us in this difficult endeavor, since they map the material traces of virtual data and reveal the earthbound properties of digital storage. Research like theirs pushes us to away from the rhetorical promise of the cloud as a weightless companion and make us see it as what it is: A heavy infrastructural network whose energy-hungry operations produce “deep, material marks” [29] in geological formations and hydrological cycles, which affect the livelihoods of rivers, animals, plants and citizens. Only if we understand the cloud as part of such wider socio-natural configurations, we can ask emphatically if data centers should really have access to an unlimited amount of energy; and if all of their operations are indispensable for society. Or asked differently, how much water is an AI-generated cat meme worth?
 

[1] The company Everest Link provides high-speed internet in the remote Everest Region since 2014, which enables users to interact with cloud services. https://everestlink.com.np/, accessed on May 10, 2025.

[2] Miguel Yañez-Barnuevo: “Data Center Energy Needs Could Upend Power Grids and Threaten the Climate.” Environmental and Energy Study Institute (EESI, 2025). https://www.eesi.org/articles/view/data-center-energy-needs-are-upending-power-grids-and-threatening-the-climate, accessed on May 10, 2025. 

[3] No wonder that Sam Altman—CEO of OpenAI, the company behind ChatGPT—said at last year’s World Economic Forum in Davos that an “energy breakthrough” will be needed to sustain the expansion of generative AI tools.

[4] Thomas Spencer, Siddharth Singh: Energy and AI (The International Energy Agency, Paris, 2025): 14.

[5] Kate Crawford, Atlas of AI. (Yale University Press, Yale, 2021): 41.

[6] According to Cloudscene market data, Switzerland has currently 121 data centers. https://cloudscene.com/market/data-centers-in-switzerland/all, accessed on May 10, 2025. 

[7] According to Switzerland Global Enterprise, a Swiss export and investments promotion agency, the country is ranked as the third most attractive location in the global Data Center Risk Index. Next to its secure and sustainable power supply, this is due to its politically stable environment, the central position within Europe and its reputation for technical innovation and low taxes.

[8] This phrasing is part of the claim of Stollen Lucerne Data Center, https://rz-stollen.ch/en/, accessed May 10, 2025.

[9] Monika Dommann, Max Stadler: “White Gold, Crypto Gold: Alpine Hydropolitics.” In: Data centers: Edges of a Wired Nation (Lars üller Publishers, Zurich, 2021): 76–78.

[10] That’s no metaphor, the 220-meter-high Verzasca dam, located in the South of Switzerland, was used for a stunt in the opening scene of the 1995 James Bond movie GoldenEye.

[11] Erik Swyngedouw, Liquid Power: Contested Hydro-Modernities in Twentieth-Century Spain (Cambridge, Mass.: The MIT Press, 2015).

[12] Sean Flemming, Where the River Flows: Scientific Reflections on the Earth’s Waterways (Princeton: Princeton University Press, 2017): 37, 40.

[13] Lisa Blackmore: “Turbulent River Times.” In: Liquid Ecologies in Latin America and Caribbean Art (Routledge, New York, 2020): 18.

[14] Armelle Choplin: “Heart of Concreteness.” In: Visible upon Breakdown (Spector Books, Leipzig, 2024): 71.

[15] Bernhard Truffer et al.: “Green Electricity from Alpine Hydropower Plants,” in: Mountain Research and Development, 21, 1 (2001): 19–24.

[16] In Switzerland, this is especially true for low-lying dams, which accumulate organic matter in their reservoirs. Dams in higher altitudes collect mainly glacial sediment. See: Arushi Arora: “Dams: Economic Assets or Ecological Liabilities?” https://earth.org/dams-economic-assets-or-ecological-liabilities/, accessed on May 10, 2025.

[17] Just to make this clear: hydropower is still ecologically less harmful than powering a data center by burning good old fossil fuels. See for example: https://ourworldindata.org/safest-sources-of-energy, accessed on May 12, 2025.

[18] Hyperscale data centers are four to ten times bigger than conventional data centers and host more than 5000 servers on a surface equivalent to the size of a football field. According to the industry platform Data Center Dynamics hyperscalers mean big business: The US market alone is expected to be worth $63 billion by 2027. 

[19] For example, a new Interxion data center close to Zurich will use as much electricity as 48,000 housholds, which corresponds to the consumption of the Swiss city of Winterthur. Diego Ortiz Yepes: “Wolkenbildung: Die Schweiz als neuer Hotspot für Datenzentren” (Lucerne University of Applied Sciences and Arts, 2021), https://news.hslu.ch/datenzentren-boom/, accessed May 10, 2025.

[20] Quite a few data centers in Switzerland are located on higher altitudes, which allows for open-air cooling systems that don’t rely on freshwater. However, many centers are located in less elevated areas too and still need to be cooled by water, especially when temperatures rise during summer. According to current research, training an AI model like Microsoft’s GPT-3 can directly evaporate 700,000 liters of clean freshwater. See https://arxiv.org/pdf/2304.03271, accessed on May 10, 2025.

[21] Swissinfo: “Switzerland outlines 15 Alpine hydro projects for the future,” https://www.swissinfo.ch/eng/politics/switzerland-outlines-15-alpine-hydro-projects-for-the-future/47190566, accessed on May 10, 2025.

[22] Additionally, the legally implemented residual water flow through dams—which is mandatory since 1992 to keep river flows alive—was reduced to the absolute minimum in 2022 in order to increase the amount of water available for energy production. https://www.bafu.admin.ch/bafu/de/home/themen/wasser/aufwertung-und-schutz-der-gewaesser/restwasser.html, accessed May 10, 2025.

[23] However, Axpo, Switzerland’s biggest renewable energy producer, states that even these efforts won’t be enough, “by a long way”. To guarantee electricity supply, six times more energy is needed than what these projects would offer. 

[24] In the short run, meltwaters from glaciers might increase stream flows, leading to more dams and greater electricity production. See Henry Fountain: “Where Glaciers Melt Away, Switzerland Sees Opportunity,” https://www.nytimes.com/interactive/2019/04/17/climate/switzerland-glaciers-climate-change.html, accessed on May 10, 2025.

[25] For example, in 2023, the canton Thurgau has prohibited the withdrawal of water from its rivers and in 2022, the town of Courtételle restricted the irrigation of lawns because of water shortages.

[26] Harry Stitzel, Alessandro Massaro: „Rechenzentren verbrauchen doppelt so viel Strom wie Stadt Bern,“ https://www.srf.ch/news/wirtschaft/digitale-stromfresser-rechenzentren-verbrauchen-doppelt-so-viel-strom-wie-stadt-bern, accessed on May 10, 2025.

[27] In 2018, Cape Town, the first major city on earth to be faced with running out of water, has devised a song that lasts only two minutes and should help people save water by taking shorter showers. https://www.npr.org/2018/09/07/644918801/singing-in-the-shower-to-help-save-cape-towns-water, accessed on May 10, 2025.

[28] During the opening ceremony of a new data center from the company Green in the town of Lupfig in 2019—the biggest center at the time—, authorities announced that Switzerland aims at becoming the “data center hub of Europe” and that data centers could even become “the new banks of Switzerland.” https://www.aargauerzeitung.ch/aargau/brugg/das-birrfeld-beherbergt-nun-das-grosste-datenzentrum-der-schweiz-ld.1386492, accessed on May 10, 2025.

[29] Nick Lall, Kelly Kay, Jim Thatcher: „Computational parasites and hydropower: a political ecology of Bitcoin mining on the Columbia River,” in: Environment and Planning E: Nature and Space (2019): 2.