The internet is disappearing, fading into the background of everyday life. With pervasive mobile broadband, there is no longer a need to consciously connect. Our smartphones are always already online. In the absence of Ethernet cables and dial-up modems, the network itself is rarely seen or heard. For many, the internet is simply infrastructure, and infrastructure only merits attention when it breaks down. Just as we don’t identify as “plumbing users” or “toilet owners,” few of us would think to call ourselves “internet users,” no matter how much time we spend online.

As the internet recedes, something called “the cloud” rolls in. Telecom engineers have used clouds in their technical drawings since the early 1970s to represent areas of the network that are uncertain, dynamic, or unspecified. But for today’s user, “the cloud” evokes something altogether more specific: the woozy, elevated contemporary experience of ubiquitous wi-fi, infinite scroll, and limitless storage. Whereas “the internet,” “the World Wide Web,” and “the information superhighway” named some kind of entity out there that we could access, a structure of some sort with gateways and boundaries, the cloud is all around. It envelops us. We don’t “log on” to the cloud; we “sync” with it, like dreamers staring up at the afternoon sky.

But the apparent weightlessness of the cloud belies the industrial outlay on which it depends. Look beyond the vapor and you’ll find data centers that throb with thousands of spinning hard disk drives and monstrous HVAC units. It is this contradiction — between the cloud’s airy image and its material substrate, the virtual and the real, placelessness and place — that preoccupies Tung-Hui Hu in his new book, A Prehistory of the Cloud. Hu is a former network engineer, now employed as an English professor at the University of Michigan. Drawing on a diverse archive of primary sources ranging from technical journals and hobbyist magazines to “tactical media” produced by artists and activists, Hu argues that “the cloud” is only the latest in a string of metaphors conveying an ideological fantasy of the perfect communication network, always connected, and impervious to failure or attack. “The cloud,” he writes, “has exceeded its technological platform and become a potent metaphor for the way contemporary society organizes and understands itself.”

To approach the cloud, Hu begins with networks of communication, conveyance, and conservation that antedate the general-purpose computer: railroads, highways, bunkers, and community television systems. From outer space, the geographies of today’s information infrastructures look remarkably unchanged from those 19th-century manifestations. “The structure of the Internet,” argues Hu, “resembles a graft: a newer network grafted on top of an older, more established network.” The cloud, too, is neither a radical departure from nor a disruption of the preexisting order, but rather a new layer of technological growth, drawing stability and sustenance from the layers underneath. Indeed, early networking experiments brought computer scientists into a conversation about power and control that civil engineers, sociologists, military strategists, and community organizers had been having for decades.

Today’s “cloud computing” is, in many ways, a mass-scale reproduction of the time-sharing systems of the 1960s and 1970s. Time-sharing, as historians of technology like Paul E. Ceruzzi, Martin Campbell-Kelly, and Judy O’Neill have documented, emerged in the 1960s as an interactive alternative to batch-processing. Early computer programming did not require a computer. Programmers devised their data-processing tasks on paper or punched cards, dropped them off with a computer operator, and returned hours, days, or weeks later to pick up the results. Commercial data processing firms ran jobs around the clock, leaving no minute of computer time unexploited. On a time-sharing system like MIT’s Project MAC, however, programmers worked out their problems in real time while sitting at a terminal. Whereas the computer processes instructions at a regular tempo, human cognition is punctuated with moments of reflection, contemplation, and distraction: wasteful uses of costly computing time. To exploit these gaps in user activity, time-sharing systems distributed the computational resources of a single mainframe among dozens of users, giving each individual a sense of exclusive control over the machine.

The experience of time-sharing at MIT prompted researchers Robert Fano and Fernando J. Corbató to envision a future in which computing is a utility comparable to power or sewage treatment. In papers published in 1965, they predicted that large-scale computing facilities would one day provide nonspecialists with on-demand “thinking tools,” vastly expanding the role of computers in society. Commercial time-sharing services like the GE Information Services (GEIS) and Tymshare thrived during the 1970s by selling computing power over the telephone. While the arrival of cheap desktop PCs gutted the time-sharing industry in the mid-1980s, the computer-as-utility business model returned in 2006, repackaged as “cloud computing.” The mass adoption of cloud computing was swift, if largely invisible to the typical computer user. Today, a huge proportion of the web relies on “virtual machines” sold by Amazon, Microsoft, and a handful of others.

Time-sharing brought about a fundamental change in how people relate to computers. Instead of preparing and submitting discrete tasks for the computer to process asynchronously, time-sharing required real-time interaction: hands on the keyboard, eyes on the blinking cursor. But time-sharing, according to Hu, was as much a shift in culture as in technology: “Time-sharing,” he writes, “seemed to restructure the very boundaries between work and leisure, public and private.” For the first time, the computer user was defined as an individual account identified by a unique username and protected by a secret password, an innovation which laid the groundwork for the political economy of the cloud. “Ever since time-sharing systems bestowed names upon users,” Hu writes, “those users have been interpellated as units of economic value.” The user account is both a marker of identity and a device for accounting, an odd combination that “confuses personal intimacy for economic intimacy.” In the cloud, we don’t just possess, operate, or inhabit our accounts: we are our accounts.

It matters that users engage with the cloud as atomized individuals rather than as a public with a common interest in shared infrastructure. Hu links this quarantining of individual users to the economy of cloud computing systems: “virtualization […] both literally isolates users from each other and also makes manifest an economic system of individual agency.” Echoing Marxist theorists such as Tiziana Terranova, he argues that each user is free to negotiate individually with cloud providers, customarily in the exchange of network services for personal data. Virtualization reinforces the sense of that each user is a free agent by isolating us in our own private, supposedly secure environments. The software interfaces we use further ask (and occasionally demand) that we tweak our user preferences, train our spam filters, and manage our timelines. This illusion of isolation and autonomy, an artifact of the time-sharing era, prevents us from recognizing the interdependency of all users of a given infrastructure. As Hu puts it: “We have become willing partners with the algorithms that channel our online experience.”

What separates the cloud from earlier computer networks, in Hu’s eyes, is not technology. Rather, the cloud is defined above all by its capacity to hide the material cost of its infrastructure behind a facade of individual user freedom and flexibility. In Hu’s analysis, the cloud traverses the gap between material and immaterial by performing a sort of rhetorical virtualization, “turning real things into logical objects.” Cloud-scale symbolic substitution is powerful, capable of transforming a roiling assemblage of switches, servers, software, standards, and streams of data into a “cloud drive” that appears as singular and easy to comprehend as the USB stick dangling off of your keychain.


In spite of its title, A Prehistory of the Cloud is not, primarily, a work of history. It’s an argument, or rather the prolegomena for one. Hu’s raft of historical vignettes provide the raw materials for a theoretical critique of what he calls “the politics of digital culture,” a phrase he uses throughout the book to refer to scholarship in communication and media studies that uses the language and structure of the internet to understand the relationship of new media and power. Hu traces this body of thought back to the work of thinkers like Manuel Castells, Yochai Benkler, Michael Hardt, and Antonio Negri who theorized the “network” as a radically egalitarian social form, able to resist central control by virtue of its very structure. Vanguard agents of the “network society” — activists, terrorists, interns, and freelance entrepreneurs — appeared to be capable of challenging (and defeating) the entrenched power of large institutions. The deliberate authority of the sovereign nation-state itself seemed ineffective and old-fashioned when compared with the unpredictable “emergent” behaviors of transnational networks.

Such radical theories of the network society were inspired by the internet of the 1990s, a network populated by less than a third of all American adults which skewed toward wealthier, whiter users living in metropolitan hubs and Western nations. In the mid-2000s, as more and more people came online, we began to see the limits of the notion that the internet is an inherently liberating technology. Streaming video, much of which duplicates programming found in the cinema and on cable TV, began to swallow up internet bandwidth. Proprietary platforms like Facebook and Google enclosed much of the web, enabling the mass-scale surveillance of user activity and aggregation of personal data. In 2013, classified information leaked by Edward Snowden revealed that privately owned data infrastructures were also being used as sites of state surveillance. Not only did institutional power survive the coming of the network society, it appeared to be thriving.

To understand the persistence of control and the distribution of power in networks, new theorists of technology like Wendy Hui Kyong Chun, Alexander R. Galloway, and Raiford Guins have turned their attention to the internet’s apparatus: the cookies, browsers, switches, gateways, and running code that give rise to the hybrid “network-of-networks.” Amid social media systems governed by impenetrable “terms of service” documents and arcane network protocols, these critics drew on the theoretical work of Gilles Deleuze to understand the form of power at work in the cloud. In particular, Deleuze’s model of “the control society” seemed to capture the way that power is distributed and personalized by the architecture of the cloud. Each user finds themselves ensconced in a virtual environment individually tailored by invisible systems of surveillance.

As the cloud overtakes the internet, Hu entreats us to revisit the assumptions that underpin 1990s-era theories of media power. Newer forms of power do not simply replace the old: power lingers and adapts. “[S]overeign power,” Hu cautions, “[…] has mutated and been given new life inside the cloud.” Nation-states now exploit the indeterminate, border-crossing nature of the cloud to exceed their political boundaries, exercising power across vast deterritorialized data-spaces in which traces of user behavior accumulate.

In Hu’s analysis, the militarization of information infrastructure has produced a novel form of state violence: “war as big data.” Targeting, a term with disquietingly similar meanings in marketing and military contexts, provides a conceptual bridge between the economic and political consequences of the cloud. The twin meanings of targeting, argues Hu, “blur the distinction between the regulatory protocols of data networks and the sovereign’s right to kill,” resulting in a hybrid form of power that he terms, “the sovereignty of data.” While the impact of targeting is typically thought in terms of curated news feeds and personalized advertising, it also drives the creation of higher-stakes technical systems: credit risk scoring, health insurance adjustment, and the anti-terrorism “disposition matrix” (better known as the “kill list”). In Hu’s analysis, the language and technology of targeting makes otherwise “bloodless” social media systems structurally proximate to the sovereign state’s mechanisms of death: “The cloud places users uncomfortably close to the mechanism of state violence.” For Hu, this proximity between the mundane and mortal is fundamental to the moral economy of the cloud. To participate is to be complicit. Hu is unequivocal: the cloud kills.


“Cloud computing” is a tricky metaphor to build an argument around for the same reason that it’s a good one to base an ideology on: it doesn’t have clear boundaries. Once we back away from the material, political, and economic conditions of individual industrial-scale computing systems like Amazon, Facebook, or Palantir, any notion of “the cloud” diffuses into generality, joining the undifferentiated ranks of “new media” and “digital culture.” Hu’s critique excels where it is most specific. For example, a detailed exploration of the former military bunkers that play host to today’s data centers reveals a return of Cold War paranoia. The architectures of both the bunkers and the data facilities they house reflect a sense of constant threat and impending disaster. The fear of losing one’s data (and, thus, one’s personhood) gives rise, in his words, to a “bunker mentality,” a conservative, defensive impulse to retreat underground, awaiting uncertain calamity.

The challenge of drawing a clear boundary around the idea of “the cloud” is clearest in Hu’s discussion of user participation. The political economy of the cloud depends upon convincing users to continuously produce new streams of data. During the 2011 war in Libya, dedicated enthusiasts provided informal (and unexpected) intelligence to NATO by monitoring military radio bands, analyzing satellite photography, and scrutinizing social media. From the perspective of the cloud, argues Hu, the practices of this amateur intelligence community and everyday social media users are indistinguishable. “Judged in terms of the cost savings they can provide to NATO,” he notes, “the RF hackers working in Libya are difficult to separate from the neoliberal consensus that turns to the marketplace for solutions, and continually seeks lower-cost or even ‘free’ alternatives.” The Facebook user and the amateur intelligence agent are both enmeshed in a “neoliberal system of free labor,” voluntarily producing value for others — NATO, Facebook, whomever — while expecting no formal commitment or compensation in return.

But there is an asymmetry to Hu’s analysis. He renders those amateur military buffs in rich detail: we learn their Twitter handles and see a screenshot of Flightradar24, one of their favorite research tools. How can we extend the same care and attention to the billion or so people who post to Facebook each day? To reduce participation in this vast social scene to a set of superficially similar activities — posting, commenting, sharing links — risks transforming the messy stuff of human life into a smooth set of data, a form of intellectual virtualization we scholars of technology should endeavor to avoid. The cloud may enclose, exploit, and re-present traces of human communication as territories of data, but public life and popular culture are not so easily displaced.

A Prehistory of the Cloud provides an opportunity to reflect on the politics of information as it is practiced today. Although its existence depends on the continuous production and aggregation of user data, Hu discourages users from thinking of the cloud as a common space. As he puts it in the conclusion, “the cloud produces users rather than publics,” shoring up a stubborn individualism that undermines the potential for collective action. Just like time-sharing before it, cloud computing offers the illusion of exclusivity, obscuring our infrastructural interdependence. So long as users experience the cloud as a personalized service rather than a common utility, matters of public concern will themselves be virtualized, recast as systemic anomalies, software bugs, or customer service complaints.

By crafting an analogy between our everyday use of computer networks and the targeted killing characteristic of the War on Terror, Hu infuses the study of media and society with an uncommon urgency. As the cloud continues to creep, its billows will exceed the bounds of our current digital economy, carrying the practice of virtualization into the domains of housing, labor, food, and fuel. To intervene in this transfer of power, we must first recognize the role that our own streams of personal data play in sustaining the current instantiation of the status quo. Only then will we cease being “users” and become citizens of the cloud, burdened with all the confusion and responsibility that citizenship entails.


Kevin Driscoll is an assistant professor of media studies at the University of Virginia. He is currently writing a book with telecom researcher Julien Mailland about the 30-year history of the French Minitel system.