The AI Deflation Dividend

Abstract

Will artificial intelligence compress wages? We build an analytically tractable general equilibrium model with an oligopolistic AI sector selling labor-augmenting services to competitive production sectors. A nested labor bundle distinguishes augmentable from unaugmentable tasks – those requiring physical presence, human judgment, or legal accountability – and delivers a natural wage floor proportional to each sector's unaugmentable share. Two channels act in opposition: a price-output channel that partially supports wages through the revenue effect, and a task-bundle channel through which cheap AI substitution compresses the marginal product of augmentable labor. The paper delivers two central results. First, the unaugmentable share drives a wedge between goods price deflation and nominal wage compression: because AI lowers output prices across all sectors while wages are anchored by the tasks AI cannot perform, consumer prices fall faster than paychecks shrink, raising real wages economy-wide. This price-deflation channel is the paper's central mechanism. Second, the model delivers a clear policy ranking: removing adoption barriers is approximately Pareto improving in the calibrated economy, while capping the oligopoly markup merely redistributes without growing the surplus. Calibrated to actual AI usage data (Anthropic Economic Index, 2026), barrier removal raises real wages by 25.2 percent relative to the no-AI baseline, 17.3 percentage points above the current-adoption equilibrium. A markup cap adds only 1.6 percentage points to real wages above the current equilibrium.