Amazon's Antitrust Defense Is Built on Lower Prices. Its Own Files Say It Raised Them.

In April 2018, on a single platform in a single month, prices rose on more than eight million items. No outside seller forced it. No cost shock explained it. It happened because an algorithm Amazon called Project Nessie had learned a quiet trick: raise a price, watch whether competitors copy the increase, and if they do, leave it there. The FTC says that one tool booked Amazon more than $1.4 billion in excess profit.³ Hold that number against the company's entire public defense, which is that it exists to make prices go down.

The official story Amazon tells about the FTC suit is that it is a case in search of a crime — a novel legal theory aimed at a company whose only sin is serving customers too well. That framing has a problem. The strongest evidence against Amazon isn't the FTC's theory at all. It's Amazon's own files, quoted in a primary court filing, in the company's own words.

The defense and the documents are pointed in opposite directions

Amazon's antitrust answer is the most appealing argument in modern competition law: lower prices and more selection are good for consumers, and a company that delivers both cannot be the villain. For decades that reasoning was nearly unbeatable in U.S. courts, because antitrust had narrowed to a single question — did the consumer pay more? Amazon's whole posture assumes the answer is obviously no. The FTC's reply is that Amazon itself built a machine whose entire purpose was to make the answer yes, then deployed it for years and counted the proceeds.

This is where the case stops being about theory and starts being about evidence. The unredacted complaint describes Project Nessie as a system that predicts whether rivals will follow an Amazon price hike, then keeps the higher price in place when they do.³ That is not a pricing tool that finds the lowest number. It is a tool that finds the highest number the market will tolerate without anyone undercutting it. Amazon told reporters the project 'didn't work as intended' and was scrapped. The court filing quotes its internal documents calling it 'an incredible success.'³ A senior executive — the CEO of Worldwide Amazon Stores — reportedly asked in early 2022 about switching on 'our old friend Nessie, perhaps with some new targeting logic.'⁷ You do not write a fond note about an old friend you buried.

Why the seller side is the harder problem to spin

The price-raising algorithm is the vivid part. The structural part is the seller economics, and it is bigger. Third-party sellers are not a sideshow at Amazon — they are the store. In FY2024 the Third-Party Seller Services segment brought in roughly $156 billion, and outside sellers accounted for about 61% of paid units sold.⁸ That is the leverage the FTC is pointing at. When a marketplace hosts the majority of its own sales through merchants who have nowhere comparable to go, the fees those merchants pay stop looking like the price of a service and start looking like a toll on captive traffic. The FTC alleges combined fees push many sellers toward paying close to half their revenue to Amazon — its allegation, which Amazon disputes — and even the narrower, documented logistics take rate is real money rising fast over time.¹

The fair objection: a thin algorithm and a hard market to define

The honest counter is that a billion-dollar profit number can be a rounding error at Amazon's scale, and that conduct facts are not the same as a winning case. Antitrust turns on market definition, and Amazon's strongest ground is precisely there: is there really an 'online superstore market' that Amazon monopolizes, or is it one competitor among Walmart, Target, Shopify-powered storefronts, and every other place a person can buy a thing? That fight is genuinely uncertain, and it is the one Amazon most wants to have, because it is abstract, expert-driven, and far from the embarrassing emails.

But two things cut against the comfortable read. First, a federal judge has already let the core claims survive — in September 2024 Judge Chun denied Amazon's motion to dismiss the Sherman Act and FTC Act claims, and notably held that price-copying conduct, even without an allegation of collusion, can plausibly violate the law.⁴ That is a doctrinal door opening, not a theory being laughed out. Second, the conduct evidence does double duty: it is not only proof of harm, it is proof that a powerful firm believed it could raise prices and keep them raised — which is, almost by definition, the behavior of something with market power. The documents don't just describe the crime. They describe the motive and the means.

It is worth being precise about what has and hasn't been resolved, because the headlines blur it. In September 2025 Amazon paid $2.5 billion to settle the FTC — but that money closed a different case, over how Prime was sold and cancelled, including a record civil penalty and redress for tens of millions of consumers.⁶ The marketplace monopoly case is untouched by that check and is grinding toward a bench trial that the parties keep moving; as of late 2025 they jointly proposed a date in early 2027.⁵ The settlement bought peace on one front. It bought nothing on the one that matters most.

Amazon will spend years arguing that markets are broad, competition is fierce, and the consumer always wins. It may even be right about the market definition — that part is a real fight. But the most dangerous witness in this case was never the FTC. It was a five-year-old algorithm with a friendly nickname and a memo describing it as an incredible success. A company can argue its way out of a theory. It is much harder to argue your way out of your own enthusiasm, written down, while the prices were going up.