The concentration of AI capability is among the most dramatic economic transformations of the twenty-first century. A handful of corporations, primarily American and Chinese, now control the computational infrastructure, the data, the research talent, and the market access required to build and deploy large-scale AI systems.

This concentration is not a market failure. It is a market outcome, the product of network effects, data moats, and capital requirements that create structural barriers to competition. The more data a system has, the better it performs. The better it performs, the more users it attracts. The more users it attracts, the more data it generates. This virtuous cycle, for the incumbent, is a vicious cycle for potential competitors.

Data as means of production. In Marx’s framework, the means of production are the tools and resources required to create value. In the AI economy, data is perhaps the most important means of production. The organisations that control the largest, most diverse, and most relevant datasets have a structural advantage in building AI systems. And data is created, in the fullest sense, by users who receive no compensation for their contribution. Every search query, every social media post, every click, every purchase is data that is captured, processed, and used to create value, almost entirely for the platform, not the user.

This is what Marxist analysts call a form of appropriation, the taking of value created by one party for the benefit of another. The users of Google, Facebook, Amazon, and similar platforms are not the customers. They are the workers, creating the data that constitutes the company’s primary asset. They receive no wage. They receive the service, which is not the same thing.

Concentration and power. The concentration of AI capability in a small number of corporations translates directly into political power. The largest AI companies are among the most significant political actors in the world, lobbying governments, setting technical standards, defining the terms of public debate about AI regulation, and exercising de facto authority over the information environments of billions of people.

Marxist analysis suggests that this concentration is not temporary, it will not be resolved by market competition. It requires structural intervention: antitrust enforcement, data portability requirements, public investment in AI infrastructure, community ownership models. These are not primarily technical questions. They are political questions about who should control one of the most consequential technologies in human history.

Reflection question: What would it mean for AI infrastructure, the computational power, the data, the models, to be publicly owned? What would change? What would not?