This is AI Frontiers. Today, who captures artificial intelligence margins? Adoption of artificial intelligence, or AI, is now near universal, but value capture remains rare. The winners will not be the firms with the most AI activity, but the ones that control data, contracts, and infrastructure to keep the margin rather than paying rent to suppliers. In McKinsey's latest global survey, 88% of respondents use AI in at least one business function. Yet only 7% have fully scaled it across the enterprise. Only about 6% attribute at least 5% of earnings before interest and taxes, or EBIT, to AI. Two years earlier, adoption had jumped from around 50% to 72%. The curve moved. Income statements mostly did not. Executives often treat AI as a technology adoption decision. That framing is wrong. AI strategy is a question of control over the operating model. If automation rests on a proprietary application programming interface, rented compute, and restrictive data terms, a firm swaps labor dependence for infrastructure dependence. The adopter gets faster without getting more defensible. If a firm cannot move its data, models, and workflows without operational disruption, it does not control its AI capability. Portfolio discipline separates value captures from broad adopters. Boston Consulting Group reports that 60% of companies reap hardly any material value from AI despite substantial spending. Firms that do capture value direct more than 80% of AI investment at reshaping core functions and building new offerings. They pursue 3.5 use cases on average against 6.1 for the rest, and expect 2.1 times the return on investment. Upstream concentration makes procurement a strategic layer. In organization for economic cooperation and development analysis, Nvidia's share of graphics processing units used for AI exceeds 80% per invoice, and the three largest cloud providers hold over 60 parts. Before deployment begins, buyers must define exit rights, data ownership, audit access, and portability for fine tuned model outputs. If a business case depends on permanent supplier access, unpriced compliance work, or unmeasured workflow change, the board is not approving a strategy. It is approving exposure. That's AI Frontiers.