Despite Amazon’s undisputed centrality in the contemporary digital economy, a close look at its core business model reveals it to be surprisingly conventional. Before the invention of e-commerce, mail-order catalog companies such as Sears, Roebuck had accustomed American consumers to purchasing goods sight unseen from vendors with whom they communicated solely via information technology.
Like Amazon, Sears, Roebuck neither manufactured goods nor owned inventory but functioned solely as information intermediary. What both companies provided was a layer of network infrastructure that links consumers to producers via a single unified interface. In the case of Sears, Roebuck, this interface was a paper catalog, for Amazon a website, but the basic services provided are identical. By organizing, consolidating, and filtering information, both the catalog and website served to simplify otherwise complicated and time-consuming informational activities and establish and maintain networks of trust across geographically dispersed networks of strangers. Concealed behind these seemingly simple user interfaces was a complex infrastructure of information-processing and communications systems, from display and advertising technologies to payment-processing systems to user support and service. And here again, it was arguably Sears, Roebuck a century earlier who was the most original and innovative; the systems that Amazon uses are perhaps more automated but are conceptually very similar.
But although both Sears, Roebuck and Amazon saw themselves essentially as information organizations, the messy reality of retail, even information-technology-mediated retail, is that eventually the goods need to be delivered. Although their sophisticated information systems could provide a competitive advantage when processing transactions, the costs associated with the management of information paled next to the costs of handling, storing, and transporting physical materials, and so both mail-order and e-commerce firms often find themselves reluctantly expanding along the distribution chain.
For Sears, this meant coordination (and occasionally partnership) with railroad companies and national postal networks and the construction of massive warehouses and distribution centers. For Amazon, this meant the coordination with (or, increasingly, ownership of) trucking companies and shipping fleets, partnership with national postal networks, and the construction of massive warehouses and distribution centers.
Within a decade of their establishment, both firms had reluctantly expanded out of informational space and into the physical environment. By 1904, Sears, Roebuck had purchased more than 40,000 square feet of office and warehouse space in Chicago alone; today, a single Amazon distribution center averages 100,000 square feet, and there are many hundreds of such centers in the United States alone. Eventually, Sears found itself constructing its own brick-and-mortar retail establishments to supplement its mail-order operations; recently Amazon, which allegedly triumphed over Sears because of its lack of such legacy brick-and-mortar, has begun doing the same.
The degree to which Amazon is fundamentally in the business of managing the movement and storage of “stuff” cannot be overstated. In 2017 alone, Amazon shipped more than five billion packages via its Prime subscription service. To accomplish this, Amazon has constructed more than 329 distribution centers in the United States, and another 380 worldwide. These include massive, million-square-foot warehouses as well as smaller, more specialized sorting and delivery stations. For delivery between its various facilities, Amazon relies on fleets of company-owned or leased vehicles. For the so-called last mile, it relies (for the moment, at least) on delivery services such as UPS or FedEx and—on extraordinarily favorable terms—the United States Postal Service. In order to further reduce its costs, Amazon has been developing an Uber-like system called Amazon Flex to further “disrupt” its dependence on third-party carriers. And famously, Amazon has announced plans to implement entirely automated drone delivery.
Amazon resembles yet another of the early-twentieth-century corporate giants, namely Standard Oil.”
In its focus on the control and consolidation of transportation and distribution networks, Amazon resembles yet another of the early-twentieth-century corporate giants, namely Standard Oil. Although Standard Oil’s dominance of the oil industry was due in part to its monopolistic consolidation of refineries, it was equally enabled by the firm’s secret manipulation of the railroad network. Like Jeff Bezos, John D. Rockefeller recognized the value of vertical integration and the necessity of access to and control over critical infrastructure. In this respect, the continuity between the industrial-era giants and the “Big Five” tech firms (Alphabet, Amazon, Apple, Facebook, and Microsoft) is all the more apparent.
When we consider the digital economy in general, and electronic commerce in particular, it seems that success is also dependent on access to infrastructure—proximity to key transportation networks such as roads, bridges, and highways; the employment of large numbers of appropriately skilled (but reasonably inexpensive) labor; the ability to construct and maintain (or at least lease) physical plants and other facilities; and, of course, access to the large amounts of capital, credit, and political influence required to secure the aforementioned resources. This perhaps explains in part why, despite the emphasis in the digital economy on light, flexible startups, many sectors of that economy are controlled by an increasingly small number of large and established incumbents. The growing belief that the United States is in the midst of a modern Gilded Age is about more than concern about wealth inequity.
Consider, for example, the one aspect of Amazon’s business model that is truly different from that of its historical counterparts in the industrial-era retail economy: namely, its integration of sophisticated computational technologies at every level of the firm, from customer-facing web interfaces to back-end databases to global positioning systems. It is because of its use of these technologies that we think of Amazon as a key player in the digital economy in the first place. And, indeed, Amazon’s implementation of these technologies was so successful that the company soon decided to package them for sale as commodity computational services and infrastructure.
Unlike Amazon’s retail operations, the provision of these services and infrastructure is highly lucrative, bringing in more than $17 billion in revenue annually and comprising the majority of the company’s overall profits. Within the computer industry, these products are known collectively as Amazon Web Services. Colloquially, the commodity computational infrastructure that these services comprise is known simply as “the Cloud.” If trucks and warehouses are the legacy technologies that ground e-commerce companies such as Amazon to materiality and geography, the invisible and ethereal infrastructure of the Cloud seems to point the way toward a truly postindustrial and entirely digital economy.
What exactly is the Cloud? At its most basic, the Cloud is simply a set of computational resources that can be accessed remotely via the internet. The value of these resources is that they are available as discrete and idealized abstractions. A user does not need to purchase a computer, install an operating system, purchase and install applications, or worry about software maintenance, hardware failures, or data backup. All of this equipment and labor is located and performed elsewhere, rendered effectively invisible to the end user.
It is this quality of seamless invisibility that most defines the Cloud as a form of infrastructure; as Susan Leigh Starr reminds us, the whole point of an infrastructure is that you never really have to worry about what makes it all possible. No one gives much thought as to how their electricity is generated, or where, or by whom; we simply expect that when we plug in our appliances or devices, the required electrons will be available. We only notice the massive size and complexity of the underlying electrical grid when it is broken or otherwise unavailable. The same is true of all infrastructure, from sewer systems to roads and bridges to our freshwater supply—and, increasingly, the internet and the Cloud.
Despite its relative invisibility, the Cloud is nevertheless profoundly physical.”
But despite its relative invisibility, the Cloud is nevertheless profoundly physical. As with all infrastructure, somewhere someone has to build, operate, and maintain its component systems. This requires resources, energy, and labor. This is no less true simply because we think of the services that the Cloud provides as being virtual. They are nevertheless very real, and ultimately very material.
A typical large data center of the kind that Amazon operates draws between 350 and 500 megawatts of power; in 2016 such data centers consumed 70 billion kilowatt-hours of electricity in the United States alone. This represents close to 2 percent of the nation’s entire electricity consumption—roughly the equivalent to the output of eight nuclear power plants. Considered globally, the amount of power used by data centers approaches 1.4 trillion kilowatt-hours. And while some of this electricity is no doubt provided by renewable resources, much of it derives from sources that are so old-fashioned as to be prehistorical, such as coal, oil, natural gas, and uranium.
According to a 2014 Greenpeace report, if the Cloud were a country, it would be the sixth largest consumer of electricity on the planet. As these resources are consumed, they return carbon back into the atmosphere—something on the order of 159 million metric tons annually—and so the Cloud is also one of the world’s largest polluters.
Given its insatiable demand for electricity, there is at least one sense in which the Cloud is more than a metaphor. Cooling a typical data center requires roughly 400,000 gallons of freshwater daily. A very large center might require as much as 1.7 million gallons. This is independent of the massive amount of clean freshwater that is required to manufacture the data center’s computer equipment in the first place. The Cloud is a heat machine designed to circulate cool air and moisture, creating its own carefully controlled microclimate and contributing to climate change in the larger environment.
Heat, air, and water are only a few of the material resources that the Cloud hungrily devours.”
Heat, air, and water are only a few of the material resources that the Cloud hungrily devours. Also present in these computers and their associated display screens are dozens of elements, some of them rare, some of them dangerous, all of which must be painstakingly mined, purified, transported, and manufactured into finished products—processes that also involve material resources, human labor, and multiple layers of additional infrastructure, many of which are controlled by some of the least stable and most exploitive political regimes on the planet.
All of which is to say that just as Amazon’s e-commerce operations are revealed to rely to a remarkable degree on traditional, decidedly nondigital technologies such. as trucks and warehouses, so too are even its most high-tech and allegedly virtual services ultimately constructed around industrial-era systems, processes, and practices.
Today, the metaphor of the Cloud erases all connection between computing services and traditional material infrastructure (as well as the long history of public governance of infrastructural resources). As a result, the computer industry has largely succeeded in declaring itself outside of this history, and therefore independent of the political, social, and environmental controls that have been developed to constrain and mediate industrialization.
By describing itself as an e-commerce entrepreneur and not simply an email order company, Amazon was awarded a decades-long tax subsidy that allowed it to decimate its traditional competitors. In claiming to be an internet service provider and not a telecommunications carrier, Comcast can circumvent the rules and regulations intended to prevent monopolies. In rendering invisible the material infrastructure that makes possible the digital economy, the metaphor of the Cloud allows the computer industry to conceal and externalize a whole host of problems, from energy costs to e-waste pollution. But the reality is the world is burning. The Cloud is a factory. Let us bring back to earth this deliberately ambiguous and ethereal metaphor by grounding it in a larger history of technology, labor, and the built environment—before it is too late.
Nathan Ensmenger is an associate professor in the Luddy School of Informatics, Computing and Engineering at Indiana University. This essay is excerpted from “Your Computer Is on Fire” Copyright © 2021 Edited by Thomas S. Mullaney, Benjamin Peters, Mar Hicks, and Kavita Philip. Used with permission of the publisher, MIT Press.