The Platform Take-Rate: Who Owns the Toll Booth

A toll for controlling the pipe, not for making the product

Every age of trade has produced some version of the same fixture: a bridge, a mountain pass, a river ford, or a harbor that goods had to cross, held by whoever controlled the land or the water around it. The toll collector did not grow the wheat or weave the cloth passing underneath. The charge was for passage alone, and it scaled with whatever value was moving through, not with any labor the collector performed. Mobile software and digital advertising rebuilt the same fixture out of code and auction logic rather than stone and rope, and gave it a reach no medieval toll road, canal, or customs post ever had: a single company's rate now applies to a purchase made on the other side of the planet, the instant it happens.

Apple's App Store has charged a standard 30 percent commission on most in-app purchases and subscriptions since the store opened, a fee applied identically to a mobile game studio, a subscription news service, and a small merchant selling a digital product to iPhone owners. Programmatic advertising, the automated system that now clears most digital ad spend through instantaneous per-impression auctions, runs a parallel toll: publishers surveyed by industry researchers estimate that a fifth to a quarter of every advertising dollar, and by broader estimates considerably more, is skimmed away by a chain of intermediary vendors before it ever reaches the page the ad ran on.

Neither charge buys the platform anything specific to that transaction. Apple did not write the app being sold inside its store. The exchanges and data brokers in the ad-tech chain did not write the article carrying the ad, or make the product the ad is selling. What both are charging for is passage, the simple fact that a payment or a bid has to cross infrastructure someone else owns to reach its destination. Tracing how that charge was set, defended, and challenged in two very different corners of the internet shows what a toll costs an economy, and, increasingly, who gets to decide it.

The App Store's 30 percent, and a five year fight over three points of it

The App Store's commission is public and simple to state. Apple takes 30 percent of most digital transactions completed inside an iOS app, a headline rate that has held since the store's 2008 launch. For a company selling a subscription, an in-app upgrade, or any digital good to an iPhone owner, there was for over a decade no route around that fee. Apple controlled the only door onto its own devices and set the price of walking through it, collected automatically, with no negotiation available to the developer on the other side.

Epic Games challenged exactly this in August 2020, when it added its own direct payment option to Fortnite on iOS, bypassing Apple's in-app purchase system, and was promptly removed from the App Store. The resulting lawsuit argued that Apple's commission and the rules protecting it amounted to an illegal monopoly over app distribution and in-app payments. In September 2021, U.S. District Judge Yvonne Gonzalez Rogers ruled that Apple's anti-steering policy, which barred developers from even telling users a cheaper payment option existed outside the app, violated California's Unfair Competition Law. She stopped short of Epic's larger antitrust claim and upheld Apple's right to require its own payment system for purchases completed inside the app itself. Apple had to let developers point users elsewhere. It did not have to give up the toll on anything bought inside its own gate.

Apple's answer to the injunction was a new commission, 27 percent rather than zero, charged on purchases completed through an external link, alongside new friction on how that link could even be presented to a user. In 2025, the Ninth Circuit found this response still violated the spirit of the original order and held Apple in civil contempt. What had begun as a dispute over whether a fee could exist at all had narrowed, five years later, into a dispute over three percentage points and the conditions wrapped around them, litigated across a federal district court, an appeals court, and, as of 2026, a petition to the country's highest court.

In 2026, Apple asked the U.S. Supreme Court to review that contempt finding, keeping the matter open. Whatever the eventual ruling, the fact that a company built primarily on hardware sales was willing to fight a five year, multi court battle over a few points of commission on software transactions says plainly how much that toll is worth defending on its own terms, independent of anything Apple actually manufactures.

The ad-tech tax: a toll with no single collector and no agreed price

Programmatic advertising runs on real-time bidding, an auction mechanism that prices and sells a single ad impression the instant a web page loads, replacing the older practice of a salesperson negotiating a block of ad space weeks in advance. Between an advertiser's budget and a publisher's page now sits a chain of intermediaries: demand-side platforms that buy on the advertiser's behalf, ad exchanges that run the auction, supply-side platforms that sell on the publisher's behalf, and separate vendors for audience data and ad verification, each taking a cut of the same dollar as it moves through.

Digiday Research, surveying publishers directly about what they actually receive, found they estimate roughly 20 to 25 percent of ad spend disappears into this chain before it reaches them. That figure, self reported but consistent across the publishers surveyed, is treated here as the established baseline for the toll. It is close in shape, if not in mechanism, to Apple's headline App Store rate: a substantial minority of every dollar spent, taken before the party that made the content or the audience ever sees it.

Broader industry investigations push the number considerably higher. Analyses from Adalytics and ClickFortify, examining the full chain of fees rather than any single vendor's cut, put the total ad-tech tax at 30 to 50 percent or more of advertiser spend in a typical setup. Some investigations conducted in 2024 and 2025 found individual cases where intermediaries captured up to 98 percent of a bid's value before a publisher saw any return on the impression it sold. Those figures describe extremes rather than an average, and they come from advocacy-driven analyses rather than an audited industry census, which is why they are marked contested here rather than established.

The range itself is part of the story. Apple's 30 percent, and its contested 27 percent, are published numbers any developer can check against a settlement report. The ad-tech tax is diffuse by design, spread across vendors that each defend their own fee as fair payment for a discrete service, with no single company positioned, or incentivized, to audit the whole chain end to end. The toll is not smaller for being harder to see. It may simply be better hidden.

Why this counts as a new instrument in media economics

Media businesses have historically made money one of two ways. They sold the content itself, a subscription, a ticket, a printed copy, or they sold access to the audience that content assembled, advertising space the publisher itself controlled and priced. In both cases, revenue flowed to whoever made or gathered something a buyer wanted. The take-rate model breaks that link cleanly. Apple did not make the app. The ad-tech chain did not write the article or manufacture the product in the ad. The toll is charged for controlling the passage a transaction has to use, independent of who created what is moving through it.

This is also why the take-rate scales so cleanly with volume rather than with cost. Apple's marginal cost of processing one additional in-app purchase is close to nothing. The 30 percent commission does not track any expense Apple incurs on that transaction; it tracks the value of the transaction itself. The same logic runs through an ad exchange charging a percentage fee on every impression it clears rather than a flat rate per auction. A toll of this kind grows automatically as the businesses using the pipe succeed, with no added work and no added risk for the toll collector.

Economists have a word for income extracted this way, taken from controlling access rather than earned by producing something new: rent. A percentage-of-transaction toll on a chokepoint someone else has no practical way around is close to a textbook example of it, and it explains why a rate cut of three percentage points, from 30 to 27, was worth a five year court fight. That fight was never really about recovering a processing cost. It was about the size of a rent.

The clearest evidence that both tolls function as rent rather than as payment for a service is that neither has shrunk as the underlying cost of running the infrastructure fell. Cloud computing, payment processing, and bandwidth have all become cheaper over the same years these commissions stayed roughly flat. A cost-based fee would track that decline. A rent does not have to, because it was never priced against Apple's or an ad exchange's own expenses in the first place. It was priced against what the businesses on the other side of the toll could be made to pay, and what they could be made to pay changed very little once the rate itself was set, regardless of who was collecting it or what it cost them to run the pipe that year.

The geopolitical contest: who gets to tax the internet's commerce

A toll road only matters to the towns along that one road. Apple's App Store commission and the dominant ad-tech chain matter to nearly every market on earth at once, because the same handful of companies, chiefly Apple, Google, and the firms running the largest ad exchanges, operate identical infrastructure everywhere a smartphone or a browser is sold. A commission rate set in Cupertino, or a ruling issued by a court in San Francisco, is not really a local price. It functions, in practice, close to a global one, because there is no separate App Store to appeal to in another country.

That is what turned Epic Games v. Apple from an ordinary contract dispute into the kind of case that climbs to a country's federal appeals court and, potentially, its Supreme Court. It is also why the European Union built the Digital Markets Act around the same target: obligations on platforms designated as gatekeepers not to force their own payment systems on developers, not to block alternative ways of distributing apps, and not to punish developers for steering users toward a cheaper option elsewhere. Brussels and the U.S. federal courts are, in effect, two separate governments reaching the same question from different directions, over infrastructure neither of them owns but both of their citizens and businesses depend on.

The stakes are magnified by how concentrated the underlying infrastructure is. Apple's iOS is estimated to hold roughly 27 percent of global mobile operating system share to Android's roughly 72 percent, a split that leaves Apple and Google jointly responsible for the terms under which nearly every mobile transaction on earth is taxed. That concentration is what turns a single court's contempt finding, or a single regulation issued out of Brussels, into something closer to an early, uneven attempt at setting a floor or a ceiling on a global commerce tax, without any single body holding the formal authority to set one for the whole planet at once.

None of this makes the dispute a conventional trade war between nations. It is closer to a jurisdictional contest over a piece of infrastructure that sits, physically and legally, inside no single country's exclusive reach. Whichever court or regulator moves first and holds, in practice, tends to set the working standard everyone else measures their own rules against, which is its own quiet form of influence over how the rest of the world's digital commerce gets priced, long before any single treaty or trade body sits down to negotiate the question directly.

The next toll booth may not charge in cash

The App Store fight and the ad-tech tax share a structural feature worth naming plainly. Both tolls sit on infrastructure a transaction has no practical way around, which is the condition that makes a toll possible at all, whether the chokepoint is a mountain pass, a shipping lane, or a mobile operating system.

Discovery is now moving through a comparable kind of infrastructure. When a buyer asks an AI system which business to call or which product to buy, the system chooses, from everything it has access to, the small number of names it actually says out loud. No published commission line exists for this yet, and none of the sources behind this article document one, so it should be read as an open structural question rather than a documented toll: if being named by an answer engine becomes as necessary to reaching a customer as appearing in an app store or an ad exchange once was, the companies that control the naming would hold a comparable kind of power over commerce, whether or not they ever charge a visible fee for it.

The lesson from the App Store commission and the ad-tech tax is not that platforms are inherently unfair. It is that whoever controls the one route a transaction, or an answer, has to pass through gets to set a price for the crossing, and that price gets defended for years in court over a few percentage points, because a few percentage points of infrastructure everyone must use is worth more than most businesses could build for themselves on the other side of it.