This blog is the latest coverage in a series on Artificial Intelligence and the Google cases. You can read more of our coverage here, here, here, here, here, here, here, here, and here.
In January 2026, Apple and Google announced a landmark multiyear partnership. Under this agreement, Google’s Gemini models and cloud computing will serve as the foundation for Apple’s next generation of Apple Intelligence products, including a substantially rebuilt Siri.
Details have not been announced. But at first glance, this looks a lot like the default agreement in the Search case, which allowed Google to illegally maintain its monopoly. But there are important differences: whereas Google clearly had a monopoly position in Search, the AI market (at least for now) looks more competitive.
Nonetheless, given Google’s history of using distribution arrangements to build a moat around its market power, government and private enforcers are likely asking the question: Is this just another power play by Google to stop competition? And if so, is that illegal?
Recap: Google Search Remedies
The Hub and Spoke has previously covered how the Google Search case cast a spotlight on emerging AI markets. The core concern was familiar: durable advantages in one market (search) translating into advantage in a nascent market (generative AI) through access to data, defaults, and computing infrastructure.
Judge Mehta’s remedies order emphasized access to key inputs like data and barred formal exclusive contracts. But he also generally allowed Google to pay for certain default arrangements (with some caveats). Critics warned those constraints might prove insufficient to restore competition in Search, and also that default arrangements could allow Google to establish a dominant position in AI, too.
A New Default: Gemini at the Core of Apple AI
The Gemini-Apple deal appears to place Gemini at the center of Apple’s AI offerings, including features users will interact with via Siri and other system-level intelligence services. While Apple insists that Apple Intelligence will still operate on-device and in Private Cloud Compute, with privacy protections intact, the underlying Foundation Models will be built on Google’s technology. This will almost certainly give Google critical user interaction feedback to allow its foundational models to improve, mirroring how defaults allowed Google to prevent rivals from catching up. Judge Mehta called this a ‘flywheel effect’ – steering users to Google (and away from rivals) created a continuous feedback loop to allow Google (but not its rivals) to improve its product, thus attracting ever more users.
Importantly, the deal has been described as non-exclusive on paper, allowing Apple to work with other AI model providers. (Apple has announced a similar, but somewhat more limited, deal with OpenAI). But antitrust observers have been quick to note that formal non-exclusivity may still produce ‘exclusive in effect’ outcomes in digital markets, where defaults are enormously sticky. This is why antitrust tribunals in digital markets, including the landmark Microsoft and Google Search liability decisions, observed that a contract need not block all alternatives to harm competition; it is enough if it meaningfully forecloses rivals.
History Doesn’t Repeat, but It Does Rhyme
Three core antitrust themes that make the Gemini-Apple deal especially significant:
- Defaults as Distribution Power. Even absent formal exclusivity, being the default provider of the foundational AI model powering Apple’s ecosystem, including across billions of active devices, could achieve the same foreclosure effect that default Search agreements once did. In Search, evidence showed users rarely switch defaults even when offered significant incentives, a dynamic that could equally apply to AI models embedded at the OS level.
- Incentives and Innovation Dynamics. The deal also raises questions about incentives to innovate or compete independently. Historically, Google’s Search defaults effectively dissuaded Apple from developing its own search engine. Today, similar concerns arise over whether Apple’s reliance on Google’s foundation model technology will dampen, if not eliminate entirely, its incentive to enter the AI market.
- Market Structure, Tip Risks, and Early Enforcement Challenges. Section 2 of the Sherman Act requires either monopoly power, or a dangerous probability thereof. But in contrast to Search, in which there was no serious question that Google had monopoly power, the AI model market appears more competitive—at least for today. Major players include not only Google, but also Microsoft, OpenAI, Anthropic, and other emerging players like xAI. This makes it difficult to allege a theory of monopolization in AI related to Gemini’s integration agreements, as government enforcers did in Search. However, if AI is characterized by the same ‘flywheel effect’ that Judge Mehta found in Search, it is not unreasonable to expect the market to “tip,” i.e., for defaults to produce a dominant leader with a strong moat protecting its position. Once this kind of structure is in place, it can be very difficult to reverse, as the years-long litigation over Google Search shows.
Cloud Computing: An Often-Overlooked Bottleneck
Beyond models and defaults, the Gemini–Apple partnership also implicates cloud computing, an area where Google already holds a powerful and durable position. Training, deploying, and scaling frontier AI models requires enormous compute resources, specialized chips, and globally distributed infrastructure. Realistically only a handful of firms can provide them. Barriers to entry are exceptionally high, with billions in capital expenditures, long development timelines, and tight control over specialized hardware such as specialized chips at the very frontiers of current technology. In short, cloud infrastructure is not merely an adjacent market to AI, it is a critical input for competing in it.
By anchoring Apple’s AI ecosystem to Google’s cloud and model stack, the deal risks reinforcing Google’s role as both gatekeeper and supplier in a cloud compute market that some argue is already marked by high concentration. This echoes theoretical concerns raised by AI developers and enforcers alike: when a small number of firms control the compute layer, they may influence which models can scale, which applications can reach users, and on what terms. Even absent formal exclusivity, preferential access to Google Cloud for Apple Intelligence, one of the world’s largest consumer platforms, could prove an incredibly high barrier for rival AI developers who must compete for scarce and expensive compute on less favorable terms, or rely on competing cloud providers that lack comparable integration or scale.
Enforcers are Watching
Antitrust enforcers are almost certainly paying attention. The Department of Justice and a coalition of states have filed notices of cross-appeal of the remedies in Google Search, signaling interest in revisiting parts of the ruling. Commentators will closely watch to what extent the appeal focuses on the Remedies’ decision’s treatment of AI, as Google has already signaled it will do.
Judge Mehta has also clarified, at the request of government enforcers, that Google cannot force partners like Apple and Samsung to bundle Gemini with other services, a move explicitly aimed at preventing anticompetitive tying in AI distribution.
These actions suggest that enforcers are already mindful of how AI markets work, and are curious about how the industry is evolving.
The Future of AI Competition—Now
The Apple–Google Gemini deal already shows a broader challenge for antitrust law: how to identify and address competition problems in foundational technologies before markets have tipped, without chilling vigorous competition and beneficial innovation.
As the Hub and Spoke has noted, collaborations like this may accelerate product improvements, giving consumers powerful AI tools sooner and aligning with procompetitive outcomes. On the other hand, the durability of default advantages and a “flywheel effect” raises concerns mirrored in earlier antitrust cases. I.e., defaults and integrations, especially when coupled with cloud compute, may quietly erect barriers that entrench a small set of players, preventing rival models from attaining scale.
As the AI landscape continues to evolve, so too will the legal frameworks that govern competition within it. And the Gemini-Apple relationship may serve as one of the first major tests of those frameworks in action.
