A medieval peasant depended on a lord because the lord controlled the land. Five hundred years later, we still see a familiar form of dependence: economic power can come from controlling what others depend upon. If a handful of companies control the AI systems that businesses increasingly need to compete, our capitalism could develop a relationship that looks feudal without actually becoming feudal. Policymakers should therefore look beyond AI market share and ask: can businesses still leave? If AI becomes infrastructure for production itself, switching between providers could become increasingly difficult as businesses build their data, workflows, and AI agents around particular environments to improve efficiency. In this unique time in history, policymakers have an opportunity to address this before dependence becomes irreversible by monitoring dependency alongside concentration, preserving interoperability and portability, and scrutinizing exclusive arrangements involving critical AI infrastructure. The economist Yanis Varoufakis describes a similar possibility in Technofeudalism: What Killed Capitalism. He argues that capitalism is being replaced by a system built around “cloud capital”: algorithms, software, computing infrastructure, data, and digital platforms through which businesses produce, distribute, and sell (Varoufakis). The important change is that economic power shifts toward controlling the digital environments through which others operate. The feudal analogy comes from dependence: those who control an essential resource gain power over those who depend on it. But isn't this simply capitalism? In many ways, yes. Capitalism is an economic system based on private ownership, markets, and profit, while monopoly capitalism occurs when productive capital becomes concentrated. Concentration alone, however, does not make an economy feudal. If businesses can freely switch providers or develop alternatives, AI remains productive capital within capitalism. The difference is not simply that businesses depend on a powerful company, but that dependence on AI could occur at a fundamental layer of production itself. AI could push this distinction into sharper focus. If AI-enabled companies can produce the same output with less labor and expense, firms that fail to adopt AI could weaken competitively. AI could shift from an optional tool to a requirement for competition. Meanwhile, advanced AI requires computing, energy, semiconductors, and financial resources at enormous scales. If those resources remain concentrated, AI capabilities could become concentrated as well. This creates a potential feedback loop: more capable AI → greater business dependence → more valuable infrastructure → more concentration → greater dependence For AI to create a feudal-like dependence within capitalism, three conditions would need to emerge: AI must become essential to meaningful competition, AI ownership must remain concentrated, and businesses must face significant barriers to leaving. A company may technically have another provider available, but if moving its data, workflows, and AI agents is prohibitively costly, its ability to switch is only theoretical. This is where I see a potential middle ground between monopoly capitalism and Varoufakis's full technofeudalism. I call it pseudo-technofeudalism, or PTF. PTF does not claim that capitalism disappears. Private property, markets, profit, investment, and capital accumulation can remain. PTF takes a narrower position: capitalism remains, but AI infrastructure could create a new form of economic dependence within it. Evidence suggests this dependence could emerge. Anthropic's Economic Index found that about 49% of jobs had seen at least a quarter of their tasks performed using Claude (Anthropic). The World Economic Forum found that 86% of employers expect AI to transform their businesses by 2030 (World Economic Forum). A 2026 IBM study found that 71% of executives said switching their primary AI vendor or model would be difficult (IBM Institute for Business Value). Together, these findings suggest that AI is becoming more important to production while businesses are already beginning to face switching barriers. For PTF, the important question is not simply whether AI eliminates jobs. It is whether businesses become dependent on AI to remain competitive. The internet offers a useful precedent: while it became infrastructure for information and communication, AI could become infrastructure for production itself. If businesses build their workflows around particular AI ecosystems, leaving them could become increasingly costly. A monopoly may control a market; PTF requires control over a fundamental layer of production itself. This is why policymakers should look beyond market share. First, regulators should monitor dependency alongside concentration, asking how difficult it would be for businesses to leave. The OECD should monitor Big Tech’s contracts, data-egress fees, limited interoperability, and proprietary ecosystems as potential switching barriers (Organisation for Economic Co-operation and Development). Second, policymakers should preserve portability and interoperability where technically feasible, so businesses can move data and systems between providers without barriers like time or fees becoming a threat to competition. Third, competition authorities should scrutinize exclusive contracts and vertical integration involving critical AI infrastructure. The Federal Trade Commission has already examined major AI-cloud partnerships over concerns involving computing access, switching costs, and exclusivity, and they should continue to do so by investigating mergers (Federal Trade Commission). None of this means concentration itself is necessarily harmful. AI requires enormous amounts of capital and infrastructure, and integration can produce genuine efficiencies. The goal should therefore not be to prevent successful AI companies from becoming large, but to prevent concentration from becoming irreversible dependence. The most important question surrounding AI power may not be who has the largest model or market share. It may be whether businesses can still leave. That is the point at which AI concentration stops being merely a question of who sells the technology and becomes a question of who controls the economy.
Economics
The AI Monopoly We Should Worry About May Not Look Like a Monopoly
Source: the supplied manuscript, reproduced without editorial rewriting. The PDF is typeset from this original text, not an author-supplied PDF. View count maintained by the publication; not live analytics.