In January 2021, Twitter and Facebook suspended Donald Trump. Not a foreign government. Not a court. Not a constitutional body. Two California corporations decided that a sitting head of state could no longer address his own public. Whatever you think of Trump, that moment deserves more than a partisan reaction. It demands a structural question: what kind of entity has that power, and what are we calling it?

The answer is not a market actor in any traditional sense. It is not a government in any recognized legal sense. It is something else — something that the vocabulary of liberal democracy was not designed to name.

The concept of techno-feudalism is an attempt to name it. Not as provocation. Not as nostalgia for the pre-digital world. As a structural description of a power arrangement that is already operating, already extracting, and already governing — whether or not we have decided to call it governance.

The urgency is concrete. In the years since the smartphone became the primary interface of daily life, platforms have demonetized independent journalists overnight, frozen the financial accounts of protest movements, and deployed algorithmic systems to determine which ideas reach which audiences. These are acts of governance. They are being performed by entities incorporated as advertising technology companies.

The word "feudalism" needs grounding before it can do useful work. Medieval feudalism was not simply poverty with castles. It was a specific architecture of power, land, and obligation.

A lord granted a vassal the right to use land — the fundamental productive resource of an agrarian economy — in exchange for loyalty, military service, and tribute. The vassal might sub-grant portions downward, producing cascading layers of dependency. The peasant at the bottom worked the land but never owned it. Access was conditional. Exit was not meaningfully available. The alternative economy did not exist.

Several features define the structure. The productive resource — land — was never truly held by those who worked it. It was held conditionally, under terms set above. The lord's power was not limited to narrow economic extraction. It extended to justice, marriage, military service, and the regulation of daily conduct. There was no separation between private economic interest and public governance function. The lord was the local state. And the legal fiction that serfs were free people — technically at liberty to leave — coexisted with the structural reality that they had nowhere to go.

Historians debate whether feudalism was a coherent system or a retrospective analytical category applied unevenly across wildly different medieval arrangements. That debate matters for medieval history. For this one, what matters is the structural pattern: concentrated control over essential productive infrastructure, conditional access, cascading dependency, the fusion of economic and governance functions, and severely limited exit.

Hold that pattern in mind.

The digital economy runs on infrastructure controlled by a small number of private entities. Cloud computing — the invisible foundation beneath most modern software and commerce — is dominated by three companies: Amazon Web Services, Microsoft Azure, and Google Cloud. Together, they hold roughly two-thirds of the global market.

When Amazon Web Services terminated its contract with Parler in January 2021, the social network ceased to exist within hours. No government ordered it. A private infrastructure provider made a business decision. The lord withdrew the land.

Platform monopolies in consumer-facing markets follow the same logic. Network effects — the well-documented tendency of communication platforms to become more valuable as more people join — create winner-take-most dynamics that are extremely difficult to reverse through normal market competition. This is established economics, not speculation. The result is that platforms controlling access to audiences, customers, and communities have accumulated leverage qualitatively different from ordinary market power.

When a restaurant loses its Google listing or an Etsy seller loses their account, they don't merely lose revenue. They lose access to the marketplace itself. That is not a market outcome. That is eviction.

Shoshana Zuboff's analysis of surveillance capitalism makes the deeper claim. Major technology corporations have developed a new logic of accumulation that treats human behavioral data not as a byproduct of service provision but as the primary raw material of a new economic sector. Behavioral surplus — the excess data collected beyond what is needed to improve a service — is processed and sold as predictions about future human behavior. The buyers include advertisers, insurers, and political campaigns.

A corporation that can predict and modify human behavior at population scale is exercising something that looks very much like sovereignty over the conditions of human experience. The question is whether we are going to use that word.

Medieval feudalism was, at its economic core, a system of rent extraction — capturing value from the labor of others by controlling the land on which that labor depended. The digital parallel is structurally striking, even if the mechanisms differ.

Apple charges a 30 percent commission on all digital transactions conducted through apps on its iOS platform. For developers — including small independent creators for whom the App Store is the only practical channel to iPhone users — this is not a negotiated rate. It is a toll. Apple sets the terms, enforces them unilaterally, and can terminate a developer's access without any external review mechanism. The developer, by building on the platform, has agreed to operate under those conditions indefinitely.

Some jurisdictions are beginning to push back through antitrust action. But the structural arrangement itself is not in dispute.

Workers who drive for Uber, deliver for DoorDash, or fulfill tasks on Amazon Mechanical Turk do not own the infrastructure through which they access their labor market. They access it conditionally, paying a share of their earnings as commission, operating under algorithmically enforced rules they cannot negotiate, subject to deactivation — the digital equivalent of eviction — at any time and for reasons that are often opaque.

Some economists contest the feudalism framing on the grounds that participation is voluntary. No one is forced to use an iPhone or drive for Uber. The response from those who find the analogy useful: voluntariness is compromised when the alternative is effective exclusion from the dominant economic infrastructure of your era.

A twelfth-century English peasant was also, technically, free to seek their fortune elsewhere.

What is the difference between a rule and an algorithm?

A rule, in the legal tradition of liberal democracy, is public. It is enacted through a defined process. It is applied consistently. It can be appealed. These features are not always honored in practice — legal systems can be corrupt, inaccessible, and inconsistent — but they constitute the normative standard against which actual systems are measured.

Algorithmic content moderation, credit scoring, predictive policing, and automated hiring systems operate on different principles. The rules are proprietary. The reasoning is opaque, sometimes even to the engineers who built it. Application is inconsistent in ways that are difficult to detect. Appeal mechanisms are typically inadequate or nonexistent.

These systems are now making consequential decisions about access to speech, credit, housing, employment, and liberty. The classic subjects of governance.

A person whose Facebook account is suspended loses the platform through which they maintain family relationships, run their small business, and participate in civic life. A person whose algorithmic credit score is miscalculated may be denied housing without ever knowing the reason. A person flagged by a predictive policing system may face increased police contact based on criteria they cannot examine or contest.

The legal scholar Frank Pasquale named this the rise of the "black box society" — a world in which the decisions that most affect our lives are made by systems whose workings we cannot inspect. Whether this constitutes feudal power or something genuinely new is debated. What is not seriously debated: it represents a significant departure from the legal norms of liberal democracy — norms developed precisely to constrain the arbitrary exercise of power by those who held it over those who didn't.

The lord's justice was personal and arbitrary. So is the algorithm's.

Medieval feudalism organized power spatially. The lord's authority was defined by the territory he controlled. The digital economy has introduced a new kind of territorial sovereignty — not geographic, but infrastructural. The relevant territory is not measured in acres. It is measured in data flows, network connections, and platform membership.

Yanis Varoufakis, the Greek economist and former finance minister, has developed the techno-feudalism thesis furthest. He argues that what we are witnessing is not an intensification of capitalism but a transition beyond it — to a qualitatively new system. In capitalism, profit is generated through the production and sale of goods and services in competitive markets. What the major platforms are capturing, he argues, is not profit in this classical sense but rent — value extracted not through production but through ownership of the infrastructure through which others must pass to participate in economic life.

This is a genuine theoretical dispute, not a settled question. Mainstream economists tend to argue that the platforms are profit-seeking firms operating in concentrated but still-addressable markets. Antitrust and regulation, they argue, remain adequate instruments. The techno-feudalist counter: the concentration has proceeded so far, and network effects are so powerful, that normal market mechanisms cannot restore competition. What is needed is something more like the political interventions that broke up earlier concentrations of infrastructural power — the regulation of railroads, the breakup of Standard Oil in 1911.

Both sides acknowledge that the geography of economic power has fundamentally changed. The dispute is whether existing frameworks — designed for territorial states and tangible property — can govern a world of cloud infrastructure, network effects, and behavioral data.

Locke through Rawls — the mainstream of liberal political philosophy — rested on the idea that legitimate authority requires consent from those subject to it. This was always an idealized account. It required setting aside the ways that class, race, and historical dispossession shaped what people could meaningfully consent to. But it provided a normative standard. A measuring stick.

When a user signs up for a social media platform, they are presented with terms of service. Legally binding contracts. Typically thousands of words. Updated frequently. Containing provisions most users will never read. Studies consistently show the vast majority of users do not read these agreements. In a formal legal sense, they have consented. In any meaningful philosophical sense, that consent deserves scrutiny.

The deeper problem is structural. If declining to use Google means being unable to find reliable information, navigate an unfamiliar city, access your employer's systems, or be discoverable by potential employers — all of which is increasingly true for large portions of the global population — then consent to Google's data practices is not meaningfully free. This is the structural coercion argument.

It is genuinely contested. Some philosophers argue it proves too much: any sufficiently important service could be held to make consent to its terms non-voluntary. Others argue it is precisely correct, and points toward legal frameworks that impose baseline rights regardless of what users have technically signed.

The European Union's General Data Protection Regulation, enacted in 2018, and more recently the Digital Markets Act represent attempts to establish exactly this kind of baseline — to say that certain rights exist regardless of what users have technically agreed to. Whether these frameworks can be effectively enforced against globally operating corporations is a live and unresolved question.

The serf signed nothing. But the absence of a signature did not make the arrangement free.

What, precisely, is sovereignty, and who holds it now?

The classical Westphalian concept — supreme authority within defined territorial borders — was always a simplification. Corporations, churches, professional guilds, and international financial networks have always competed with formal state authority in particular domains. But the modern democratic state represented, at least aspirationally, the consolidation of governance into a single accountable entity. Its laws were public and appealable. Its monopoly on legitimate force was subject to constitutional constraints. Citizens knew, in principle, where to direct their grievances.

What is new about the present situation is not that private power exists. It has always existed. It is the specific combination of scale, scope, opacity, and indispensability that the major digital platforms now embody.

A medieval lord governed a village. Standard Oil monopolized oil distribution in one country. The major digital platforms govern the conditions of communication, commerce, and information access for billions of people across every national jurisdiction simultaneously. Their reach is global. Their accountability is fragmented and national at best. Their technical systems are complex enough that even sophisticated regulators struggle to understand what they are actually doing.

The political theorist Wendy Brown has described a broader phenomenon she calls the "undoing of the demos" — the erosion of conditions necessary for democratic self-governance. The techno-feudalism thesis is a specific instance of this diagnosis. Not just that powerful private actors are making governance decisions. The infrastructure on which democratic deliberation now depends — where citizens organize political action, access political information, communicate with representatives — is owned and operated by the same entities whose power democracy most urgently needs to address.

This is the recursive trap. The platforms through which you would organize to regulate the platforms are controlled by the platforms.

The techno-feudalism framework distorts as well as illuminates. All analogies do. Intellectual honesty requires naming what it gets wrong.

Medieval serfs were legally bound to the land. They lacked mobility, had no access to external legal systems, and lived under lords who held the power of life and death. Platform users in democratic countries can, in principle, choose different platforms, organize politically, access courts, and vote for legislators who can regulate technology companies. The asymmetry of power is real and growing. It is not the total, violent domination of medieval serfdom. Conflating the two risks trivializing genuine historical oppression and overstating the helplessness of contemporary digital subjects.

The platforms have also created genuine value. Google provides, at no direct monetary cost, the most powerful information retrieval system in human history to anyone with an internet connection. Amazon delivers goods to rural areas that previously had limited access to diverse markets. The value extraction and the value creation are not separable. That is part of what makes the situation genuinely difficult.

Serious reform efforts exist. The EU's regulatory agenda is the most ambitious. The GDPR, the Digital Services Act, the Digital Markets Act, and ongoing antitrust proceedings against multiple major platforms represent a sustained attempt to bring digital power under democratic governance. Some jurisdictions are experimenting with data trusts, platform cooperatives, and public digital infrastructure as structural alternatives to private platform monopolies. Academic work on algorithmic accountability, interoperability mandates, and data portability points toward concrete policy directions.

Whether any of this will prove adequate to the scale of the challenge is genuinely unknown.

What the techno-feudalism framework does, at its best, is force a clarifying question. When we submit to the governance of platforms, what are we giving up, and to whom? It asks us to look past the user interface — past the convenience, the connection, the entertainment — and examine the structural relationships underneath. It asks whether the fact that something is called a "service" rather than a "government" tells us what we actually need to know about the nature of the power being exercised.

The digital infrastructure of the twenty-first century will be governed by someone. The only serious question is whether that governance will be accountable to the people living inside it.