Book Review: Platform Capitalism
The contemporary economy often presents us with more questions than answers. On one hand, Uber has gained a near monopoly on private sector ground transportation - they’ve crowded out competitors, slashed driver pay, and raised prices - yet they’ve lost billions and remain unprofitable. On the other hand, Google offers its most crucial services for free - search, email, video - and has become one of the world’s most valuable companies.
These seem like opposite strategies, but they arose under the same economic moment. Post-2008 conditions - employment stagnation, low interest rates, and surplus capital seeking higher returns - created the environment in which these companies can exist. What can their diverging paths tell us about the future of 21st century capitalism?
Companies like Uber and Google aren’t anomalies - they represent a more fundamental shift. Traditional economic categories no longer suffice, so they represent a genuinely novel development. Rather than a curiosity, this new model is becoming the dominant form of contemporary capitalism. Political theorist and economist Nick Srnicek has observed these trends and given a name to this new form: the platform.
Crisis Conditions
During times of crisis, capitalism has a tendency to restructure itself around new technologies and forms of organization. In order to continue market growth, capitalism creates demand for new technology that can increase the profitability of a given firm against its competitors. Srnicek argues that there are three moments in the recent history of capitalism that have set the stage for today’s digital economy: the global profitability crisis of the 1970s, the dot-com bubble of the 1990s, and the financial crisis of 2008.
More specifically, the governmental response to these crises created the economic conditions that were necessary for the rise of platforms as we see them today. First, low interest rates were introduced to encourage investment and growth as a means of recovery following the 2008 market crash. One major effect of this was that surplus capital from wealthy people and institutions was funneled away from government bonds and savings accounts as investors turned towards speculative investments in pursuit of higher growth. Despite any of this potential for growth, however, job growth has remained largely stagnant outside of low-paying service jobs and unstable, independently contracted “gigs.” The stagnation of wages has forced many people into self-employment, where workers can “be their own boss” while still relying on large corporations for their paychecks.
Furthermore, Congress has been unwilling to use their powers to aid economic recovery efforts. Fiscal stimulus has been politically unpopular as both Democratic and Republican strategies have been centered around neoliberal austerity and fears of inflation. This has put the job of managing the health of the economy on the Federal Reserve and Central Bank, which have largely responded with a loosening of monetary policy. We’ve seen this manifest in the forms of historically low interest rates and the implementation of quantitative easing (in which a central bank buys up government bonds and other securities to inject more money into the economy). These monetary policies created a combination of excess corporate cash holdings and low returns in savings accounts and bonds, which resulted in large investments in speculative technology start-ups as an attempt to pursue higher profit margins.
It was in this environment - hungry for growth, flush with capital, and searching for new frontiers - that a particular kind of firm began to emerge.
Platforms as a Response
Many attempts have been made in the effort to come up with a name for the newest evolution of capitalism: the gig economy, the on-demand economy, the surveillance economy, etc. Although these terms all describe distinct phenomena, they all point to the same underlying economic shifts. Srnicek opts to examine these changes by focusing on the dominant organizational forms that present-day firms take.
Platforms weren’t the result of random luck and technological breakthroughs. Rather, they emerged as a direct response to the particular internal needs of companies. As more parts of the economy started to shift into the digital realm, the amount of data that became available for analysis began to increase at a faster rate than traditional business models could handle. As the models of Fordist factories and production lines weren’t able to process data at high speeds, platforms took shape as a response to rapid influx of new data. So, what mechanics allowed platforms to keep up while other firms fell behind?
Core Mechanics
Srnicek argues that at their core, platforms are all about extracting data as a raw material. Much like the raw materials of the industrial world, such as oil and minerals, data alone isn’t of much use to anyone. In order to harness the power of oil, the infrastructure of oil rigs, pipelines, and processing plants were created. So while the behaviors of users on a digital platform might generate massive amounts of data, the real value in a platform comes in its ability to analyze and translate that data into valuable insights for their clients. For example, Google doesn’t give all of the raw data they’ve collected from users to their advertising partners. Instead, they internally process that data into insights about users’ lives. Then, advertisers pay Google knowing that the ad network will show their advertising to the most receptive potential customers possible.
In order to help them extract data at a large scale, platforms position themselves as intermediaries that facilitate interactions between users. Placing themselves in this position then gives them privileged access to every user interaction, therefore enabling them to collect, process, and analyze their data as efficiently as possible.
Another defining feature of platforms is their reliance on the network effect. This takes place when having more users on a platform makes the platform more valuable for every other user. For example, Facebook uses a proprietary algorithm to match people with other users that they are likely to know or be related to. The biggest reason that people are drawn to Facebook instead of other websites that could do a functionally identical service is simple: everyone else is already on Facebook. Combined with the increased accumulation of data that comes with new users, the network effect creates a powerful reinforcing feedback loop towards further growth of the platform. In the context of capitalism, the monopolization that we see in giant tech companies isn’t just a product of greedy CEOs and consumer choices. Rather, the tendency towards monopoly is built in to the DNA of the platform model. As platforms grow larger and larger, their dominance over user data makes it ever more impossible for business competitors to arise.
The network effect also gives rise to another behavior that’s appeared puzzling through the lens of traditional business logic. To get more users and generate more extractable data, platforms often make services free (or at least very cheap) even at a cost to their profit margins. While profitability used to be the ultimate metric of business success, we’ve seen a rise of giant companies like Uber and AirB&B that are willing to remain unprofitable for a long period of time in hopes that they’ll gain a monopoly over their respective markets in the long run. In a process known as cross-subsidization, firms will often have multiple branches that aim to generate profits to subsidize the losses of giving some of their services out for free. For example, we’ve seen companies like Amazon and Google branch out into cloud computing as a way to subsidize their unprofitable programs like Amazon Prime and Google Drive.
The Taxonomy of the Platform
Srnicek categorizes the different types of platforms by looking at what data they use and how they profit from it. According to this framework, we can identify five unique groups: advertising, products, cloud, industrial, and lean platforms.
Advertising Platforms
By their nature, these platforms tend to be the most stable and easily recognizable out of the five categories. Companies like Facebook and Google provide “free” services such as social media, email addresses, or video streaming. This positioning gives them the ability to collect data generated by user behaviors. These behaviors can take the form of anything that the company is capable of tracking and storing as data. Some examples include everything you click on, every account you interact with, and how long you spend watching a video before turning it off. Once the companies collect massive amounts of this data, they’re able to construct detailed profiles of every user. Given the quantity of data, they’re often able to predict anything about you, from your age and gender to your passions and political beliefs.
Contrary to what we might initially assume, advertising platforms tend not to profit by selling our data to advertisers. Rather, they profit by selling a promise to the advertisers themselves: that they’ll find the most receptive customer possible for every ad. We can therefore see that they are materially motivated to collect as much data about you as possible to be of the most value to the advertisers. We see this take the form of ever increasing behavioral data collection in brand new domains: smart health devices (like fitness tracking watches and Bluetooth-enabled heart rate monitors), smart home devices (like 24/7 recordings from doorbell cameras and baby monitors), and smart TVs (which means that they’re allowed to remotely send images of every single thing you watch to their servers).
Product Platforms
Product platforms extract data in the form of transactions and sales. Companies like Amazon and eBay provide a digital hub where consumers can search for virtually any item they want from an endless catalogue of sellers. Being the intermediaries between the buyers and sellers, these companies can leverage their knowledge of exactly how much customers from all around the world are willing to pay for any given item.
Cloud Platforms
This type of platform extracts data through the providing of digital infrastructure to other businesses. Amazon Web Services or Microsoft Azure are examples of this, where they provide computation power, data storage, web hosting, and other services to anyone that’ll pay them. This is appealing to other businesses, because it means that they can pay for stable and expandable tools instead of having to build a solution to every technological need in-house. Cloud platforms profit by collecting data from all of the activity on their platforms, as well as collecting rent on the servers and processors they run.
Industrial Platforms
Industrial platforms extract data through manufacturing processes. These types of platforms aren’t as commonly recognized, since individual consumers don’t tend to interact with them as much as corporations. Examples include companies like GE and Siemens, which rent out or sell things like engines and factory parts for production and manufacturing businesses. They can then make a profit by locking those businesses into a sort of “ecosystem,” where parts from one company are incompatible with those from another company. This can also extend into providing servicing and maintenance for those industrial parts, which further incentivizes un-repairability and creates a need for platform-specialized service workers.
Lean Platforms
Lean platforms are a relatively new form that we’ve seen emerge in companies like Uber, DoorDash, and AirB&B. These types of platforms are somewhat unique among the others due to the fact that they tend to derive profits not from significant data usage, but through labor intermediation. In the case of Uber, the platform connects riders to drivers while the actual service is carried out largely without interaction from the platform. The idea of paying someone to drive you somewhere is not too different from a taxi, but the main difference is that Uber gets to decide how much each trip will cost, how much the driver will be paid for the trip, which drivers to match with each rider, and so on. These platforms are “lean” in the sense that they aim to reduce their own participation as much as possible while slashing all possible costs. Uber can label their drivers as “independent contractors” rather than employees, and can use their position of control to force drivers to pay for their own gas, repairs, cleaning costs, health insurance, car insurance, and so on.
Despite their seemingly massive rise, lean platforms have so far been largely unprofitable. Their ability to capture markets has mostly been a result of surplus capital seeking start-ups for high growth potential, and easy access to low-interest corporate debt.
Combined Platforms
We’ve also seen an increase in the number of firms willing to combine aspects of multiple types of platforms under one corporate umbrella. For example, Amazon is largely thought of along the lines of the product platform, yet a bulk of their profits tend to come from Amazon Web Services, their cloud platform wing. As the capitalist demand for profit creates a corresponding demand for increasing amounts of data, it’s natural that large firms will attempt to expand their platforms into new areas.
Srnicek’s framework gives us a powerful tool to understand the convoluted worlds of technology and capitalism. Yet, there are still some questions that I think should remain up for debate - and some ways that pushing past the framework can reveal even more about how platforms work.
What Data Tells Us About Capitalism
Data as Oil
Srnicek’s framework gives us a powerful tool to understand the convoluted worlds of technology and capitalism. Yet, there are still some questions that I think should remain up for debate.
First, we should examine Srnicek’s assertion that we should think about data as a natural resource like oil. There are a few strong points in favor of this metaphor. If we think of the capturing of data through platforms, we can draw a strong parallel to the ways in which oil and other natural resources are extracted.
In the case of digital platforms, data exists in small pockets out in the world — in the form of users. Every time a user joins a platform, a new batch of data becomes available for extraction. To make use of this data, the platform needs infrastructure that can extract, transport, and refine it into useful end-products. Much like the pumps, pipes, and plants of the oil world, user behaviors are captured, transported through fiber-optic cables, stored in server rooms, and refined through analytics into business insights.
Another similarity between data and natural resources can be found in the advantages of accumulation. Much in the same way that oil companies compete to be the first to buy (or steal) oil-rich land so that they can build their own infrastructure on it, early-movers in the digital world get their own advantages. Once they establish themselves as the dominant platform in an area, they can profit from their exclusive control over new user data. This profit then lets them reinvest into bigger databases, better algorithms, and advertising to recruit more users, which creates a natural tendency towards monopolies.
Capitalists extracting natural resources for profit will enclose and protect their territory with a threat of violence, whether it’s literal or legal. Similarly, digital platforms rely on government regulations to protect their intellectual property rights and algorithms. Furthermore, the fact that digital platforms must be run on physical servers requires a corresponding level of physical (possibly armed) security on premises.
We could even think of the natural resource status of data in terms of pollution. When the security of a platform database is compromised, we tend to see massive leaks of private user details. Things like names, addresses, financial information, and social network connections can be weaponized to harm victims of corporate data breaches, much like how oil spills and other pollution incidents cause real harm to the people nearby.
The Differences
That being said, there are clearly some ways in which data differs from natural resources. One of the most obvious is that data doesn’t deplete in the same way as a limited stock of resources. When you refine and use data, it doesn’t go away or get transformed into a new form. In fact, you can create as many copies of any dataset you want as long as you have enough digital storage.
One could argue there’s a practical limit: a finite amount of user activities can be captured, and digital storage requires Earth’s resources. But data still creates novel behaviors that finite resources don’t. You can copy a hard drive full of data to another drive — giving you two identical copies. You can’t do that with oil.
Furthermore, the massive scale of data can often create new unexpected behaviors or insights. In one form, you can combine all of the different datasets you have about one person to give you a complete picture of their life. In another, you can combine endless amounts of data in the form of books, wikis, internet posts, and so-on to create the large language models (LLMs) that power things like ChatGPT. We could say that we have a similar effect with natural resources, in that you can combine larger amounts of it to create new behaviors. While one drop of oil won’t be very useful, a huge tank of oil can power an engine that moves a vehicle. Despite that, the way that data can be extracted from so many sources and turned into a wide variety of new forms intuitively seems to be on a more complex level.
Another difference is that data doesn’t seem to be out there in the same way that some like oil or wood is. There’s a physical space where that substance would exist regardless of whether or not anyone was trying to extract it and make use of it. We can’t say the same of data. While we might have natural behaviors, those behaviors don’t get transformed into data unless there’s someone actively collecting it. In that sense we might try to conceive of behaviors as the “raw material” that must exist prior to the creation of any data. However, we run into issues when we see the ways in which data can influence and reshape our behaviors. The idea of the online trend shows this perfect: when someone is able to present data showing that a lot of people are doing the same behavior, other people might react to that data by engaging in that same thing. Data is also often collected in a motivated manner - if someone wants to prove that their thing is the best out of all of the things, they might look for reviews or focus-groups or usage rates that can enforce their narrative.
On Natural Resources
While some of the differences between data and previously existing natural resources might seem to point to the inherent properties of data being new and novel, we should also look at it from the other direction: what does data tell us about the existence and use of natural resources?
Artificial Scarcity
First, the scarcity condition of a natural resource is one of the prerequisites of being able to make a profit off of them. Just as was the case with early forms of capitalism, platforms are constructed in a way that creates scarcity artificially. Prior to capitalism, peasants controlled their means of subsistence (in the form of farm lands) until an outside force violently enforced the private ownership of that land. Once the peasants were separated from their common control over the food they grew to survive, they were forced into a state of general market dependence in which they only had two choices: wage labor or starvation. In Capital Volume One, Marx described this process of primitive accumulation as such:
The process, therefore, that clears the way for the capitalist system, can be none other than the process which takes away from the laborer the possession of his means of production; a process that transforms, on the one hand, the social means of subsistence and of production into capital, on the other, the immediate producers into wage laborers. The so-called primitive accumulation, therefore, is nothing else than the historical process of divorcing the producer from the means of production. It appears as primitive, because it forms the prehistoric stage of capital and of the mode of production corresponding with it.
In other words, capitalism needed to separate people from their means of survival — forcing them into wage labor as the only alternative to starvation. The parallel to today is clear: just as peasants were pushed off their land, users are now pushed off their own data.
Digital platforms generate their value through the extraction and processing of user data. Clicks, posts, and messages are stored in their private databases where they can then profile their users and profit however they see fit, whether their customer is a data broker, an advertiser, or a government (e.g. the data ICE and CBP use to fuel their mass deportations). Users produce the data, yet they have no control or ownership over it. Mirroring the early capitalist mythologies of virtuous accumulators and the lazy masses, digital platforms increasingly encourage an ideology that worships capital owners over everyone else. Companies like Apple and Facebook Meta make big shows of their CEOs through keynote speeches and public stunts to convince the public that the owners are visionaries, geniuses, and trail-blazers. This worship can even be imposed on us algorithmically as Elon Musk showed us when he threatened his engineers at Twitter into boosting his own posts to rank 1000 times higher than anyone else’s.
Activity Generation
As platforms have evolved, we’ve seen that pre-existing behaviors aren’t just used as data. Rather, those behaviors are actively transformed into new forms that are easier to extract from. Every social event becomes an Instagram-able “moment” to be shared. Every bad social interaction becomes a viral video on Reddit or Facebook (the amount of death threats you get from strangers on the internet will vary heavily depending on your gender and socioeconomic status). Every night out at the bars (combined with a lack of public transportation options) becomes a necessary Uber trip or a drunk driving fine - your money is funneled to capital owners either way.
Domain Expansion
Just as is the case with oil, the limits of easily available surface-level data must be eventually be overcome to keep up with the capitalist demand for growth. As oil reserves in some areas have depleted, technology evolved in order to extract from more difficult areas - fracking to extract from deep rock formations and deep-sea drilling to move off-shore. The push for research and products to create the Internet of Things parallel those evolutions. Products like smart watches, smart TVs, health-trackers, and doorbell surveillance cameras reflect this tendency. This can also take the form of new capabilities for products you already use, like smartphones. Every new data point is a new inference about your personal profile, which means it’s profitable to track you using everything your device has to offer. Every user action, no matter how obscure or how private, is seen as open space to be captured by an algorithm. In this way, the ever-increasing capitalist demand for heavier surveillance of users echoes the “discovery” rhetoric of colonialism.
Some might wonder why we should even bother with trying to categorize data as a natural resource or otherwise. While just making these connections on it’s own won’t change anything, our conceptual understandings and framings of the things in our lives shape what’s considered to be politically possible and “realistic”. For example, if we were to conceptualize data as a mere product generated by companies, we would naturally have to center our focus on the distribution and effects of data after it’s created. If we instead think of data as something that must be extracted from our lives and then refined out of its natural state, we can better understand the ways in which data extraction itself can change our lives before it’s even processed. It’s an open question as to what solutions for data ownership, usage, and privacy will produce the best outcomes. If the current situation entails the sole ownership of data under one corporate entity that can charge rent for access, we should look to imagine and construct ways in which data can be accessed and harnessed for the common good.
Convergence and Fragmentation
One of the more puzzling aspects of platform growth are the ways in which platforms try to protect their control over data while expanding into new areas. As platforms have grown, they’ve shown a tendency to occupy very similar spaces - each company can have their own versions of social media, cloud storage, and advertising networks. One possible explanation is that they’ve all identified a limited subset of platform types and data sources that can remain profitable in the long term. Unlike the horizontal and vertical integrations and mergers of traditional business models, the ways platforms expand tend to be driven by their need to occupy key positions in data flows to remain profitable against competition. Rather than combining to create larger or more efficient unified platforms, this has mostly taken the form of every company trying to do a bit of everything.
In response to this tendency towards convergence, platforms have responded by building walled “ecosystems” through enclosure. Users get locked in through dependency and the inconvenience of switching away. Where there are alternatives, platforms make it as difficult as possible by making sure that none of your data is compatible with anything else. A lack of data portability means that any of the convenience you’ve gotten at the cost of your privacy will be lost when you try to leave the platform. Think of social media or email providers - it feels impossible to “migrate” your social connections, followers, group chats, or anywhere else. We’ve also seen this in the form of operating systems and app stores. Our digital lives increasingly rely on closed-source, proprietary software. Unlike open-source programs, where you’re free to look at all of the code that runs in the background and modify it however you like, our interactions with software are increasingly dominated by subscription models, surveillance, and incompatibility.
Looking Forward
Conceptually understanding platforms as profit-seeking firms that are built to extract, refine, and use data can help us understand what’s at stake politically and what responses might be required. While this framework can’t answer everything, it can certainly give us a more clear understanding of how platforms will respond to various interventions.
The first and most obvious strategy is through regulations and antitrust enforcement. The U.S. government could implement short-term fixes like cracking down on tax avoidance and holding companies accountable for privacy violations. Local regulations could help with lean platforms like Uber and AirB&B that exploit workers and make cities less affordable. While these are desirable and likely necessary, Srnicek argues they remain unimaginative — bandaid fixes that neglect structural conditions like network effects and tendencies toward monopoly. That said, they could raise consciousness and build political capacity. An FTC that aggressively fought platform power could bring specific wrongdoings into the public eye, and organizing around short-term demands could build coalitions for more radical demands.
Another approach involves publicly owned platforms — where the state invests in platforms that serve the public rather than capital. Run as public utilities, they could enhance our lives alongside infrastructure projects like rural internet access and sustainable electricity. This way, network effects could be harnessed for the common good instead of enriching shareholders.
The most ambitious approach is developing alternatives that overturn the current status quo. If platform mechanics could give people alternatives to market dependence and wage labor, the possibilities could be endless. Decentralized networking — like Monero, Peer-to-Peer torrents, and Tor — has shown ways to transfer information anonymously without a central authority. Newer technologies like Meshtastic and Meshcore offer off-grid communication that escapes privately owned cell towers. If scaled beyond niche hobbyist projects, we could overturn reliance on corporate-controlled communication. What if platforms could be harnessed not as a means to silo off and exploit data for oneself, but instead as a way to share art, knowledge, and the other great parts of human life that have been out of reach for the vast majority of people around the world?
I believe that Srnicek gives us the right framework. The harder question is what we do with it.