CONCEPT
Concentration of AI Means
The structural feature of the AI economy by which the means of knowledge production — training data, model weights, inference infrastructure — are controlled by a handful of corporations, inverting the distributed pattern that gave print culture its resilience.
The AI economy is distributed in ways that are less consequential than they appear and concentrated in ways that matter more than is commonly recognized. The internet infrastructure through which AI is accessed is genuinely global. The applications built on AI are geographically distributed. But the core of the AI stack — the training corpora that encode the knowledge base, the model weights that represent the trained capability, the inference infrastructure that makes real-time use possible — is controlled by a small number of corporations. A handful of organizations make decisions about what texts enter the training data, how to
weight different sources, what constraints to place on outputs, and what capabilities to expose through public APIs. These decisions shape the intellectual environment of every AI user, without requiring the users' consent or allowing their scrutiny. Concentration at the core of the
AI commons inverts the distributed pattern that gave print
culture its resilience, and the