This is AI Frontiers. Today, AI monopolies extract value even when the price is zero. The textbook monopoly story starts with a higher price. That framing is no longer enough. In AI and platform markets, the user-facing price is often zero. But the extraction is real. The firm captures value by turning user behavior into data, data into prediction, and prediction into durable market power. Here is the mechanism. Consumer surplus, the gap between what you would have paid and what you actually pay, is not simply disappearing. It is being converted into an asset the firm owns. Every prompt, correction, click, and usage pattern has informational value. Aggregated across millions of users, those interactions become a production input, improving targeting, model refinement, and behavioral prediction. The user experiences the exchange as consumption. The firm records it as both consumption and capital formation. This is why free is often a misdescription. A service is not free in economic terms if participation supplies an input the provider can privatize. User activity functions as unpaid digital labor, not in the sense that users are employees, but in the sense that their actions produce something with exchange value, and that value is retained and monetized by the firm. The problem sharpens with AI. Use improves the product. Prompts, rankings, and workflow patterns all feedback into model refinement. A better model trained on user interaction is not just a better product, it is a stronger moat. And if users cannot carry their histories, preferences, and workflows to a rival, the surplus they helped create has already been appropriated. Data also enables price discrimination. Once a firm knows who is locked in, who is price sensitive, and who cannot afford interruption, it can organize the market so that more of the value ends up on its side. The most powerful monopolist is not the one with the highest public price. It is the one that knows enough to set the terms invisibly. The policy implication is direct. A regulator who looks only for higher prices will miss the core extraction mechanism of many AI monopolies. The right questions are about data portability, interoperability, contract terms, and merger analysis that treats data as a strategic asset. Where rivalry is real, productivity gains diffuse. Where control is concentrated, they accumulate. That's AI frontiers.