"Ownership at scale lowers the marginal cost of training advanced models, building a moat smaller competitors will struggle to match" [marginal cost]
The explicit linkage between compute ownership, marginal cost reduction, and competitive moat directly supports the claim that capital concentration is becoming the primary structural barrier in AI competition.
Reasoning from this article
The article's analyst quote articulates a structural dynamic: AI competition is shifting from who has the best algorithms to who owns the most efficient compute at scale. This mirrors historical patterns in industries like semiconductors and cloud, where capital intensity eventually consolidates markets around a small number of incumbents. The $800bn sector-wide capex figure for 2026 alone suggests the barrier to entry is rising faster than most challengers can match, making the current investment cycle potentially decisive for long-term market structure.