Research

The Data Moat Dimension: Why Proprietary Training Data Is the Scarcest Resource in AI

Research Team · June 3, 2026

Among all nine AIPI dimensions, AI Data Moat shows the widest spread and the highest correlation with long-term positioning score stability. We explain why this dimension matters most.

Compute can be bought. Talent can be hired. Foundation models can be licensed by the hour. Proprietary training data can be none of these things. It has to be accumulated, almost always as the by-product of running a real business over many years. That is why, among the nine dimensions inside the AI Positioning Index, the AI Data Moat dimension is the one that most reliably separates durable advantage from temporary advantage.

Across the 502 companies scored, the data-moat dimension averages 46.1 with a standard deviation of 24.3 — the widest spread of any single dimension we track. That spread is the point. It means the market has not converged. Some companies sit on observations no competitor can replicate at any price. Others have nothing a well-funded rival could not assemble in a single quarter.