When AI Costs Fall, Will Prices Follow? Information Delay and Algorithmic Pricing

A simulation-based market study explores whether falling AI model operating costs lead to lower commodity prices, and how delays in observing competitors' prices and limits on price adjustment frequency affect consumer benefits. The study sets up a framework with two differentiated sellers, logit demand, and tabular Q-learning, covering 16 conditions across 960 training runs. Results show that cost savings are not necessarily fully passed on to consumers in simulated scenarios: with the shortest information delay and price adjustment allowed each period, the average price reductions of the two firms are 55.6% and 44.0% of the cost reduction, respectively.