Amazon's pricing AI and German fuel stations reveal hidden algorithmic collusion risks

The FTC alleges Amazon used a tool called Project Nessie to raise prices after rivals matched, earning over $1 billion, while a German study shows independent pricing algorithms can lift margins by about 38% when used by competing stations.

In its antitrust suit, the Federal Trade Commission alleges that Amazon deployed an internal pricing system named Project Nessie, which flagged items likely to be copied by competitors, raised their prices, and maintained the increase after rivals matched, producing over $1 billion in surplus profit. Amazon disputes the claim, stating the tool was retired long ago. The article references a 2024 Journal of Political Economy study showing that German gasoline stations using automated pricing software saw market margins climb about 38% when two stations in the same market adopted the tool, while margins were unchanged with only one adopter, indicating that independent algorithms can converge on higher prices without explicit agreements.

It categorizes algorithmic antitrust concerns into “ghost” (independent systems learning to avoid price wars), “mirror” (a firm predicting rivals, as with Amazon’s tool), and “hub” (shared platforms like RealPage that pool competitor data). Legal cases, including DOJ actions against RealPage and recent appellate rulings, highlight the difficulty of applying traditional antitrust frameworks to machine-learned coordination.

Municipal bans on algorithmic rent-setting tools and RealPage’s lawsuit against New York illustrate growing regulatory scrutiny. The piece advises corporate boards to map competitor-price visibility, impose constraints, audit algorithm behavior, conduct counterfactual competition tests, and document pricing mandates to mitigate hidden collusion risks.