At 60 Hudson Street in Tribeca, the web hums an ominous tune. From street level, a constant low whir is the only clue that 70 percent of the Northeast’s internet traffic is routed through the address. Otherwise, the former Western Union Building is unassuming, one of the many art-deco high-rises in the area. But data moving at that scale is hard to hide completely. The technology gets incredibly hot—requiring loud, constant cooling. So, AC drones 24/7 to keep the mostly unpopulated 290,000 square feet of dispassionate metal, plastic, and wires cool. At street level, it adds another layer to the city noise, buzzing in the background, easy enough to ignore.

Although laymen are rarely privy to their existence, 60 Hudson is one of many places where the ephemeral, digital world of the cloud has a physical footprint. Such spaces—called data centers—are considered eyesores and hidden away. For as much as our culture loves the internet, its physical manifestations are unwelcome, unsightly blemishes. The majority of data centers in America are situated in bland, suburban, and windowless warehouses. The data centers in cities are equally camouflaged, residing inside office-like buildings that match their surroundings. Concealed, they’re unwelcome reminders of the strange blinking masses of silicon that enable immediate access to our daily digital pleasures. They’re reminiscent of clumps of wires too unpleasant to untangle, instead tucked away underneath office desks.

As a result, users interact with the cloud at a remove. But the cloud, the ethereal force supporting all our modern infrastructure (when referencing the cloud, people primarily mean data stored remotely, accessed through the internet) is a real, touchable, and noisy phenomenon.

By not treating it as such, with each download or WhatsApp, we reify a relationship with the technology that is ignorant of the influences these tangible environments have on our social and physical world. These rooms ferry our happy birthday messages, weepy FaceTimes, and love letters. They store our photo catalogs, notes app diary entries, and grocery lists. They are, in effect, extensions of us—their infrastructure inevitably touching each bit that passes through them. Their engineering structurally influences both the most intimate and banal aspects of our lives.

Although relegated to abstracted exile, data centers provide one of the cloud’s few meaningful sites of analysis, concentrating a significant amount of web infrastructure into one location. When you upload a file to Dropbox, read a Wikipedia article, post on Instagram, log into Second Life, or perform any other activity online, your request either passes through or lives in a data center. Terabytes of information patiently wait for someone to reaccess them. The buildings are consolidated points of power on a distributed web, a microcosm for how the cloud operates: one of the few places where we can see the internet at a scale even marginally approaching its vastness.

“In name, the cloud pretends to exist in the air, almost like a natural phenomenon. But even as it becomes real, it operates with a complex, inhuman logic.”

Even if you’re able to locate the warehouses, however, within the racks of servers, one electron that helps, say, encode this article sits next to another that manages payroll and another that shuttles internet pornography. As a teenager, I cataloged server racks at NOAA’s Global Systems Laboratory. In my memory, although inarguably tactile—a quality the internet rarely possesses—the rooms were hard to absorb conceptually. Wires overlapped in difficult-to-discern patterns, lights blinked at seemingly random intervals, and computerized whirs floated in the air all around me. Hours spent without social interaction in windowless rooms felt extraterrestrial. With their own unique din, climate, and lighting, the rooms existed out of step with the exterior, human world. I was being paid to monitor the space, although I couldn’t tell you how it worked. In name, the cloud pretends to exist in the air, almost like a natural phenomenon. Ideally, by zooming into its hidden tangible sites, the illusion would be exposed and the cloud’s functioning exposed. But even as it becomes real, it operates with a complex, inhuman logic.

In A Prehistory of the Cloud, media scholar Tung-Hui Hu summarizes the cloud’s obscure nature, explaining, “This erasure is crucial to producing a sense of limitlessness within the digital environment: just as a pipe creates a sense of water as an infinite resource that can be summoned at the turn of a tap, the cloud creates a sense of computing power as a virtually unlimited resource.” Rather than comprehending or even witnessing the machinery, users simply guzzle the computational power from an always-running and contextless tap.

Wherever it’s encountered, the cloud resists legibility. It alienates the signs of its traceable design, masquerading as an otherworldly, alien force. With such varied responsibilities in such an opaque package, the data center is distilling and disseminating some sort of pre-existing cosmic force beyond comprehension. Yet, as much as the cloud is abstracted, its effects are not. As evidenced by neighbors’ complaints about the noise emanating from 60 Hudson, for example, the cloud places significant, felt pressures on the environment—both culturally and physically.

As Hu notes, this concept of limitlessness is often reflected in society’s laissez-faire approach to resources used for computing. The cloud demands a gigantic amount of energy—networked devices account for two percent of total global carbon emissions. While larger companies like Google, Amazon, and Microsoft have pledged to decarbonize, they primarily do so through controversial carbon buyback programs. Further, although their market share is growing, only around one-third of data globally is housed in “hyperscale” (read: massive and owned by a tech giant) data centers like these, with the rest having varying degrees of energy efficiency goals.

An anthropologist of cloud storage with a PhD from MIT, Steven Gonzalez Monserrate explained over the phone that these companies have “achieved carbon neutrality by offsetting, producing, investing and constructing renewable energy infrastructure, which the data centers themselves don’t run on because it’s not reliable enough.” And carbon is only one part of the environmental equation. “Water [used for cooling] is another issue and another driver of climate change. Only in 2022 did Google report its water usage, after many, many damning news reports came out. They were trying to hide water usage, FOIA requests that were denied, activists being concerned, farmers competing with data centers, and so forth.”

“Sharing its structure with the original drivers of climate change, the cloud now evades the user’s gaze while still pumping out carbon and chemically tainted water.”

It’s clear the web isn’t as immaterial as it portrays itself. The so-called virtual world instead is vulnerable to, and even contributes to, the same IRL issues we’ve dealt with for decades. As Hu observes, the internet is grafted on top of our existing world. The same power supplies and routes that were developed for highways and railways are now ferrying data. Traditional factories are being converted into data centers. Sharing its structure with the original drivers of climate change, the cloud now evades the user’s gaze while still pumping out carbon and chemically tainted water. But, as writer Everest Pipkin points out, evidence of this reappropriation is hidden: “Such information is virtually nonexistent publicly and that any attempt to access American internet infrastructure records is liable to engender a stern rebuff.”

As with most modern technologies, the internet industry’s protectiveness can be traced back to the cloud’s original military goals. As famously exposed by Edward Snowden, the web is a militarized, protected space. Many cite the internet (and as a result, the cloud) as a descendent of ARPANET, a US Department of Defense-funded project that broke cross-country communications into electronic packets of information. Created under the shadow of the Cold War, the communication format was designed to endure nuclear attacks, with the packets able to route around damaged nodes in a network.

Hu notes that ARPANET was not the sole origin of the internet, citing the much more ragtag Truckstop Network of artists laying a similar foundation: “If we only imagine the network as a product of the military, working with their contractors, to ‘invent’ ARPA and the Internet, then the network that we take away is a deeply paranoid one—a vision of nuclear strikes and distributed tanks.” Although admirable to invoke some of the more utopian underpinnings of the web, the ARPANET model appears to have won. Historians by and large identify the paranoid network as the basis for our current models.

Wartime logic still permeates the data center. Monserrate explained that “at least in the earliest days of the data centers, most people were not IT professionals, many of them were ex-military.” A surprising amount of the centers are housed in nuclear bunkers. Security measures include spy-movie-level precautions, such as retinal scanners.

Built on an anxious foundation, the internet is perpetually operating under the specter of attack. On a diffuse, networked platform, the enemy is everywhere. Viruses, hackers, scammers, and terrorists constantly pose generalized, imagined threats to be protected from. This defensive, militarized starting point leads users to see generally safe, freer peer-to-peer lines of exchange—like Soulseek, a site mostly used for exchanging audio files—as seedy, dangerous no-man’s lands full of potential threats. Yet, in reality, threats like ransomware often come from moderated, supposedly safe avenues such as phishing emails, Facebook scams, or dubious web advertisements.

Regardless, we’re directed to hegemonic and top-down mediating powers (Google, Apple, Spotify, etc) to sanitize our online experience. With the data center being one of the few large-scale sites of physical computational power, its security measures reflect the fears embedded in the digital world. The citadel, where computational activity is centralized and dispersed, is the only safe space, guarded—both digitally and physically. The wartime anxiety is so great it jumps from the digital to the material, prominently figuring into the design of these physical spaces.

“Originally billed as a utopia promoting connection, freedom, and community building, the networked world has become a dreaded arena.”

Pairing an ever-present fear of attack with a careless attitude towards resource consumption, the data center has been designed with a staggering amount of fail-safe redundancy to avoid downtime. If one section is compromised, there’s a simultaneous backup running already. Only six to 12 percent of the power some centers consume is directed towards computation. The rest is mostly insurance against disaster.

In addition to its material and political impacts, a bloated, redundancy-driven digital environment presents a subtler cultural influence. Our online ecosystem is designed to release dopamine, facilitate communication, and automate menial tasks. So, theoretically, the time spent on the web should be immensely rewarding and pleasurable. But, when dialed to 11, nothing feels good anymore. Constant stimulus inevitably plateaus. Originally billed as a utopia promoting connection, freedom, and community building, the networked world has become a dreaded arena. The draining nature of our online lives has been lamented ad nauseam—from the Surgeon General’s warning against teens using social media to studies linking loneliness to time on the internet. Although it feels reductive to blame society’s ills on our phones, perhaps the issues stem from a culture of consuming content through a redundant, seemingly infinite infrastructure.

The observation that social media is bad for one’s emotional well-being is certainly not groundbreaking. But, when understood through the perception of an infinite cloud, an insatiable appetite for content is easier to trace. Software and interface designs have taken their cues from the data center: behind each morsel of content is another primed to take its place. User journeys promote repetitive, endless motion. Now, storing mass amounts of machine code, we fear its absence. God forbid the data center goes offline or the well of information dries up. It might be best represented by the doomscroll, in which the user passively consumes massive amounts of content, occasionally “liking,” sorting some videos/tweets/status updates to their pile of preferred media. Instead of fostering open communication, the network’s paranoid military foundation has laid the groundwork for a bloated, redundant ecosystem where users inevitably burrow into password-protected bunkers of content.

In the 2009 political philosophy cult-classic Capitalist Realism, Mark Fisher observed a phenomenon similar to the modern addiction to the scroll, dubbing it “depressive hedonia.” Pleasure was pursued without joy. In one anecdote, he noted that the music dribbling out of his students’ headphones during class wasn’t even playing at a registerable volume. A pleasure-seeking connection to “the matrix” (as Fisher dubs the modern web of “communicative sensation-stimulus”) is almost universally pursued, despite producing no pleasure.

Building physical cloud architecture that accommodates ballooning amounts of data is now ushering in an expansion of Fisher’s matrix. Inside the dematerialized, faux-infinite cloud, depressive hedonia goes hyperspeed. Instagram bots spam comment sections with endless promises of sex in hopes of phishing personal data. Dating apps are designed for users to collect hundreds of matches, most of whom they are unlikely to speak with. Camera rolls, in which tens of thousands of photos lay dormant, prompt pop-ups encouraging users to buy more data to store their catalogs (it’s infinite, after all!). Rather than a network of interpersonal connection and information sharing, the stereotypical digital life is defined by wading through seas of junk living on a few corporate mega-sites. Right now, AI is poised to generate even more stuff—much of it garbage. It’s probably useful to pause, stop generating, and consider, before diving headfirst into the growing stash of data.

As we lament a militarized, polluting web, it’s important to note this current mode of operation isn’t the only one. Right now, Monserrate, the cloud anthropologist, is researching alternate futures with analog “molecular (synthetic DNA) and 5D optical ceramic (digital cuneiform) storage systems.” While sidestepping many of the issues with our current data storage infrastructures by existing without electric power, these storage systems would require a reworking of our relationship with information. Both are cold storage solutions, meaning that they’re not immediately accessible on demand.

“Our current immaterial cloud-metaphor model is not only unsustainable—it’s not useful. Encouraging mindless, unpleasant accumulation of data, it forgets the purpose of storage.”

Faced with the physical reality of our information, gone would be the days of an infinite cloud. Without immediate access to said cloud, gone, too, would be the days of unfettered access to information. But how often are you dredging up that accidental screenshot from 2019? “Data is just being stored endlessly and backed up and backed up and bloated,” says Monserrate. “If we’re not even using most of the stuff that’s on there, why does it need to be always available instantly?”

However, abandoning fingertip access to gargantuan amounts of content also forfeits the (often unfulfilled) promise of immediate pleasure. It would require thinking about the physical reality of our data—etched onto much more tangible tablets and encoded into DNA strands.

While both technologies require some more development before they’re completely viable, they present a new way of thinking about data, offering a more analog, less harmful way of engaging with the cloud. Instead of misapprehending the physical and digital as completely separate realms—leading to ignored, harmful leakages—it marries the two.

It’s a way of thinking that’s useful even with our current online infrastructure. Considering the physical reality of liking, saving, and browsing on the cloud is not only more accurate but a more responsible way to engage with the digital world. Every action is mirrored with a physical one—whether it be creating a ceramic etching or the heating of a server that consumes water and carbon.

Our current immaterial cloud-metaphor model is not only unsustainable—it’s not useful. Encouraging mindless, unpleasant accumulation of data, it forgets the purpose of storage. Rather than a legible archive, the data center is a consolidation of anxieties. The enemy is out there, and we need to collect as much information as possible as we shelter. We’re nosediving into an ocean of data knowing we can’t swim. Even if we transition to purely renewable power sources (which we should), we’ll still be saddled with a less-than-pleasant online ecosystem. If we want to emerge from our hoarder’s den, it might take some careful consideration of what we want out of a networked world—open connections or behemoth collections. In a world where we’re defined by what we consume, data provides some security, assuring us that we exist in the matrix. It’s scary to let go of the tens of thousands of photos, hundreds of Tinder matches, and thousands of Instagram followers. But it might, finally, feel good.