Google–Apple Gemini Deal Underscores Tech’s Antitrust Catch-22

Apple Inc.’s newly announced collaboration with Google LLC to use Gemini as the foundation for Apple Intelligence, including Siri, should feel uncomfortably familiar. We have seen this movie before, and we know how it ends.

Google entered into a similar agreement with Apple in 2016: the Google–Apple Internet Services Agreement, or ISA, making Google the default search engine on Apple. Users were free, in theory, to use other search engines or download other browsers.

In practice, almost no one did. A federal judge concluded in 2024 that these distribution agreements were anticompetitive because the default placement produced foreclosure effects.

The Gemini deal looks similar. It positions Gemini at important access points in the Apple ecosystem, such as Siri. Initial reports suggest the deal is non-exclusive—that is, Apple could partner with other foundation models and consumers could download other chatbots or browsers that integrate with other foundation models.

But this was also true of the search deals. The core lesson of the Google search case was simple: In digital markets, defaults matter more than formal exclusivity.

Once entrenched, defaults are remarkably sticky. Evidence in the Google case showed that even when Microsoft Corp. offered 100% of Bing’s search revenue, Apple still concluded that no price would justify switching away from Google.

In his ruling, Judge Amit Mehta rejected the idea that a contract must foreclose all alternatives to be exclusive in effect. Drawing on the Microsoft precedent, he held that it is enough if a deal closes off a significant share of distribution opportunities. The Apple–Gemini deal risks recreating that dynamic for AI distribution.

The value of being the foundation model powering Apple’s ecosystem could be even greater than being the default search engine on Apple devices. Switching from a default search engine or downloading a different browser was at least possible.

By contrast, it’s unclear whether users will be able to change the underlying foundation model that powers their iPhone’s Intelligence features. And if so, that process would likely involve as much friction as changing a search engine, if not more.

There are some differences between the ISA and the Apple–Gemini deal. Safari was a critical gateway to the internet. Siri and other Apple Intelligence features aren’t yet the dominant gateway for AI-assisted queries, as most people still access AI tools through standalone apps or browsers.

However, as AI chatbots and agents become more capable and commonplace, voice assistants and OS-level AI features likely will grow in importance. As long as users remain anchored to iPhones and MacBooks as their primary hardware, whichever foundation model powers Apple’s ecosystem will enjoy a massive distribution advantage. Even as consumer behavior shifts and chatbots overtake search engines as the primary gateway to the internet, the dominant hardware ecosystem could still shape which chatbot or AI assistant emerges victorious.

Defenders of the deal will argue that this is simply product design—no different from Apple choosing its own camera lens supplier. Users don’t get to pick those, so why should AI models be different?

The answer depends on what we think foundation models are. If they are merely components, Apple should enjoy broad discretion. If they are general-purpose infrastructure akin to search, operating systems, or app stores, then default control takes on competitive significance. Given how deeply these models shape information access, creativity, and economic opportunity, the latter view is hard to dismiss.

The deal also exposes a deeper problem with how antitrust operates in fast-moving tech markets. Section 2 of the Sherman Act requires proof of anticompetitive effects. Today, the market for foundation models appears vibrant and competitive (with ChatGPT, Gemini, and Claude), so there are no clear anticompetitive effects attributable to the Apple–Gemini deal.

The moment one firm secures control over a critical distribution channel, that competition can dissipate quickly. Yet those effects tend to materialize only after defaults have done their work. We are thus stuck in an antitrust catch-22: Conduct is hard to challenge before markets tip, and hard to fix after they do. By the time illegal monopolization becomes evident and measurable, it is already difficult to reverse.

The Google search case itself illustrated this asymmetry. After years of litigation and a finding of liability, remedies were modest and anticlimactic, and Google’s dominance in search remains largely intact.

Meanwhile, Google’s innate advantages are increasingly translating into the AI domain. For the last few years, the rise of companies such as OpenAI and Anthropic PBC was held up as evidence that competition was thriving and that tech industry incumbents could be challenged.

But that narrative underestimated the old guard’s inherited advantages. Google brings to Gemini a vast body of publisher data scarped from its search index, easy integration into search results through AI overviews, Android as a launch platform, and now Apple’s distribution channels.

The Gemini deal also changes Apple’s incentives. If Apple has access to a foundation model through Google, will it want to invest in building or partnering with a competing one?

This dynamic isn’t new, either. In the search case, the court found that the distribution agreement with Google dampened Apple’s incentives to develop its own search engine. It functioned as a kind of non-aggression pact between giants, insulating each from competition by the other.

Apple, having missed the first wave of AI development, may now find it easier to integrate Google’s models than to build its own. That may be rational business strategy, but it reinforces a familiar pattern in which dominant firms deepen their respective moats and divide monopoly spoils rather than compete to erode them.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law and Bloomberg Tax, or its owners.

Author Information

Madhavi Singh is the deputy director of the Thurman Arnold Project and a resident fellow at the Information Society Project at Yale Law School.

Write for Us: Author Guidelines