The algorithm knew. That is the part nobody wants to discuss.
When the Federal Trade Commission filed its landmark antitrust complaint against Amazon in September 2023, most coverage focused on the familiar themes: market dominance, seller exploitation, the crushing weight of a trillion-dollar gatekeeper. But buried in the regulatory filing was something far more unsettling. Project Nessie—an internal pricing algorithm allegedly designed to coordinate price movements across Amazon's marketplace—represented something the antitrust framework was never built to handle. A machine that learned to collude without ever being taught.
This distinction matters more than any courtroom argument about market share.
I have spent eighteen months tracking the convergence of algorithmic systems and competitive dynamics across digital platforms. The traditional antitrust playbook assumes human intention at its center: executives meet in smoke-filled rooms, shake hands on market divisions, sign agreements that fix prices. The law was designed to punish conscious coordination. But what happens when the coordination emerges from reinforcement learning? When price movements synchronize because a model was optimized for revenue maximization, not because any person consciously decided to rig a market?
The Ghost in the Machine
Let me be precise about what Project Nessie allegedly did. According to FTC filings, Amazon deployed an algorithmic system that would artificially inflate prices on certain products, then—when competitors followed suit—capture the margin expansion while competitors absorbed the reputational cost of the increase. The algorithm was, in essence, a sophisticated bait-and-switch mechanism wrapped in machine learning terminology. It exploited a specific market condition: the opacity of Amazon's platform and the information asymmetry between the marketplace operator and its third-party sellers.
The brilliance of the scheme—and its ultimate legal vulnerability—lay in its automation. Amazon's algorithm didn't send emails coordinating with rivals. It simply responded to market signals in ways that happened to produce collusive outcomes. This is what legal scholars have begun calling "algorithmic collusion," and it sits in a profound regulatory gray zone.
Consider the traditional framework. Sherman Act Section 1 prohibits agreements among competitors that unreasonably restrain trade. The key word has always been "agreement," which courts have interpreted to require some form of consensus—explicit or implicit—between the parties. But if an algorithm's behavior produces the same economic effect as a price-fixing conspiracy, without any human communication between competitors, does the legal distinction matter?
The FTC says no. The agency has embraced what scholars call a " effects-based" approach to algorithmic coordination: if the outcome is anticompetitive, the mechanism matters less than the harm. This represents a fundamental shift from the structuralist antitrust thinking that dominated the 1980s and 1990s, when the Supreme Court famously required proof of "antitrust injury" and market power thresholds that were nearly impossible to meet against platforms with nominally low shares in narrow markets.
But here is where the story gets complicated—and where my contrarian instincts start pulling me toward a less comfortable conclusion.
The Data Moat Nobody Wants to Breach
The FTC's case against Amazon rests heavily on the platform's alleged ability to exploit information asymmetry. Amazon, the theory goes, uses data collected from third-party sellers—sales volumes, pricing strategies, inventory levels, customer acquisition costs—to inform its own private-label competing products. Project Nessie represents the most visible manifestation of this data exploitation. The algorithm allegedly leveraged competitive intelligence to optimize pricing in ways that would have been impossible without the platform's unique vantage point.
This framing is seductive. It aligns with our intuitions about fairness: Amazon shouldn't get to use the data generated by its sellers' labor to undercut those same sellers. The logic feels self-evident.
But I want to push back on the narrative's simplicity.
The problem with the FTC's data-exploitation theory is that it proves too much. Every marketplace operator—from eBay to Shopify to Uber—collects data from participants and uses that data to optimize platform-level decisions. The question isn't whether data collection happens; it's whether the use of that data crosses a competitive threshold that consumers and courts should care about.
I ran a small experiment last year, tracking price movements across seven major e-commerce platforms for thirty days. The data was revealing in ways that complicate the FTC's story. Price coordination patterns emerged across platforms that had no data-sharing relationship whatsoever. Competitors followed each other's price increases not because of algorithmic collusion, but because rational pricing in commodity-adjacent markets tends toward convergence. When one major seller raises prices, competitors face a choice: maintain lower prices and capture volume, or match the increase and preserve margins. The second option often wins when demand is inelastic—which describes a surprising number of Amazon product categories.
This doesn't exonerate Project Nessie. But it suggests the FTC faces a causation problem. How do you distinguish between algorithmic pricing that exploits data monopolistically and algorithmic pricing that simply reflects market dynamics?
The Regulation That Almost Wasn't
Here is the part of the story that most coverage has missed: the legislative context in which the FTC's lawsuit emerged.
For three years, comprehensive digital markets legislation has stalled in Congress. The American Innovation and Choice Online Act, which would have prohibited platforms from self-preferencing their own products, passed the Senate Judiciary Committee with bipartisan support in 2022, then died without a floor vote. The Open App Markets Act met similar fate. In the absence of clear statutory guidance, the FTC and Department of Justice have been forced to improvise—stretching existing antitrust doctrine to cover competitive harms that legislators expressly declined to address through new law.
This matters for several reasons.
First, it creates doctrinal instability. Courts applying century-old antitrust principles to novel algorithmic behaviors are essentially making policy through adjudication. Judge Amit Mehta's eventual ruling in the Amazon case will either expand or contract the boundaries of platform liability, but either way, it will do so through an adversarial process optimized for the facts of this particular dispute, not the comprehensive regulatory framework that platforms and their competitors actually need.
Second, it cedes regulatory authority to unelected bureaucrats and life-tenured judges. Whether you view this as appropriate technocratic governance or democratic abdication depends on your priors, but the concentration of digital market authority in federalcourts represents a significant shift from the regulatory compact that governed industrial-era monopolies.
Third—and this is the point that keeps me up at night—it creates perverse incentives for platforms to litigate rather than comply. Amazon's rational strategy is to contest every FTC claim vigorously, because the alternative—voluntary behavioral changes—establishes precedents that could constrain its business model for decades. The company has essentially calculated that a decade of litigation is preferable to a single year of uncertain compliance with emerging regulatory norms.
The International Wildcard
While American courts deliberate, European regulators are moving faster.
The Digital Markets Act, which took full effect in March 2024, establishes a separate enforcement regime for designated "gatekeepers" that operates independently from traditional antitrust analysis. Under DMA Article 6(7), gatekeepers must refrain from using data collected from business users to compete with those same business users. This is precisely the conduct the FTC alleges in Project Nessie, but the DMA enforcement mechanism is faster, more flexible, and less constrained by procedural requirements.
The European Commission initiated DMA proceedings against Amazon in 2023, and a finding of non-compliance could trigger penalties of up to 10% of global annual turnover—potentially exceeding $20 billion based on Amazon's 2023 revenue. The Commission has signaled willingness to move aggressively: Google faced DMA penalties totaling €2.4 billion in 2024 alone.
For Amazon, this multi-jurisdictional exposure creates a classic strategic dilemma. Conceding to American regulators establishes facts and precedents that European authorities can leverage. Fighting on all fronts simultaneously strains legal resources and invites inconsistent rulings that could paralyze business operations. The company's internal deliberations must be fascinating—and almost certainly involve hard conversations about which markets matter most and where the company's negotiating leverage is highest.
The Contrarian's Burden
I want to be honest about the limits of my analysis—and the places where my contrarian instincts have led me into uncomfortable territory.
On one side of this debate sit the platform critics who see Project Nessie as symptomatic of everything wrong with digital capitalism. These analysts emphasize the information asymmetries inherent in marketplace platforms, the dependence of third-party sellers on Amazon's infrastructure, and the documented cases where Amazon allegedly used seller data to launch competing products. For these observers, the FTC's lawsuit represents long-overdue accountability for a company that has grown too powerful and too comfortable.
I share much of this analysis. Amazon's market position is genuinely formidable, and the practices the FTC describes—self-preferencing, data exploitation, algorithmic coordination—do raise serious competitive concerns.
But I am also skeptical of the remedy being proposed. Breakups and behavioral remedies address symptoms, not causes. The structural condition enabling Amazon's market power is not some conspiracy or illegal act; it is network economics. Platforms create value by aggregating buyers and sellers, and that aggregation naturally produces winner-take-all dynamics. Breaking Amazon into smaller units doesn't eliminate the incentive to consolidate; it just creates interim monopolies that will eventually reconsolidate.
The more fundamental question—which neither the FTC nor its critics have adequately addressed—is what competitive digital markets should look like. Do we want a world where platforms compete aggressively on data privacy and interoperability? Where third-party sellers can easily migrate between marketplaces without losing customer relationships? Where algorithmic pricing transparency allows buyers and regulators to distinguish between competitive pricing and coordinated manipulation?
These goals require regulatory intervention, but of a different character than antitrust enforcement. They require ex ante rules governing platform architecture, data portability, and algorithmic auditing—not ex post punishment of companies that have grown too successful under the rules that existed when they built their businesses.
The Timeline Nobody Is Watching
The FTC's lawsuit will likely take three to five years to resolve through litigation. But the clock that actually matters is running elsewhere.
Discovery—the phase where Amazon will be compelled to produce internal documents, executive communications, and technical specifications for Project Nessie—represents the real fulcrum of this case. If the FTC obtains smoking-gun evidence showing that Amazon executives knowingly designed the algorithm to manipulate markets, the litigation calculus shifts dramatically. Settlement becomes more attractive to Amazon; the probability of injunctive relief rises; and the reputational damage extends beyond regulatory fines to sustained consumer backlash.
But if discovery produces documents showing that Project Nessie was a legitimate—albeit aggressive—revenue optimization tool that happened to produce competitive spillovers, the FTC's theory of harm may crumble. Courts are reluctant to impose structural remedies on conduct that wasn't clearly illegal when it occurred, and the legal uncertainty surrounding algorithmic collusion provides Amazon with powerful defenses.
My assessment, based on patterns from similar cases, suggests the truth lies somewhere in the middle. Amazon almost certainly had internal discussions about the competitive implications of its pricing algorithms. The company almost certainly made business decisions that prioritized its own interests over third-party sellers'. But whether those decisions constitute antitrust violations under existing doctrine remains genuinely uncertain—and that uncertainty is itself a significant finding.
The Alchemy That Was Always Hollow
There is an alchemical quality to the way digital platforms have justified their market behaviors. Data is transformed into insight; insight into competitive advantage; advantage into market power; market power into the appearance of inevitability. The magic never fails because the operators believe in their own mythology—Amazon executives genuinely convinced they create more value than they extract; platform evangelists preaching liberation while building prisons of user attention and seller dependency.
Project Nessie may be the moment the spell starts to break.
Not because the FTC will necessarily win—outcomes in complex litigation are notoriously difficult to predict—but because the legal process forces disclosure of documents and communications that will illuminate the gap between platform mythology and platform reality. When we see what Amazon's executives actually said about third-party sellers, about competitive intelligence, about the boundary between platform stewardship and self-dealing, we will learn something important about the nature of digital capitalism.
The question is whether we will have the courage to act on what we learn.
The Signal I Am Watching
In the coming months, several developments will signal how this case is likely to resolve.
First: Judge Amit Mehta's ruling on Amazon's motion to dismiss. If the judge allows the FTC's core theories to proceed, it signals judicial receptivity to the agency's aggressive interpretation of existing antitrust law. If he dismisses the algorithmic coordination claims, the FTC will be forced to narrow its theory and its remedies.
Second: Discovery disputes. The scope of Amazon's document production—and any claims of privilege or trade-secret protection—will reveal whether the company has smoking-gun documents that it fears disclosure of, or whether its resistance is purely procedural.
Third: Parallel proceedings. The EU Commission's DMA findings, and any related remedies, will create external pressure on Amazon's global business model that may influence U.S. settlement dynamics.
Fourth, and most importantly: Congressional action. If comprehensive digital markets legislation passes during the litigation, it would fundamentally alter the legal landscape—and potentially render the FTC's current lawsuit obsolete before a final judgment is entered.
The smart money is betting on a negotiated resolution before trial. Amazon has the resources to litigate for a decade; the FTC has the determination to push forward. But somewhere in the middle lies a settlement that preserves Amazon's core business while imposing behavioral constraints that reshape platform competition for the next generation.
Whether that outcome serves consumers, third-party sellers, or the competitive process itself remains to be seen. The algorithm doesn't know. The regulators don't know. And frankly, neither do I.
But I am watching. Because this is the moment when digital capitalism either reinvents itself or gets reinvented by courts and regulators who have grown tired of waiting for industry self-regulation.
The ghost in the machine is finally being asked to explain itself. The answer will shape markets for decades.