Apple is preparing to show off iOS 27 and a rebuilt Siri at its June 8 WWDC keynote, powered by a custom version of Google’s Gemini AI model with 1.2 trillion parameters. The deal is reported to cost Apple roughly $1 billion a year, making it one of the largest known AI licensing arrangements between two tech rivals. The partnership signals that Apple’s own in-house models have not yet reached the performance level needed to compete with Google, OpenAI, and other AI leaders, and it raises questions about how much of Apple’s core software stack will depend on outside infrastructure.
Why the Gemini deal changes the stakes for Apple’s AI strategy
Apple has spent years building its own machine learning capabilities, but the decision to license a massive external model from Google represents a sharp turn. Rather than relying solely on its own research, Apple is now tying its flagship voice assistant and broader Apple Intelligence suite to technology built by a direct competitor in hardware, search, and mobile operating systems. The arrangement carries operational risk: if Google adjusts pricing, restricts access, or changes the model’s capabilities, Apple’s most visible consumer-facing AI product could be directly affected.
The financial scale of the reported commitment, roughly $1 billion a year for a custom Gemini model, places this deal among Apple’s largest recurring technology expenses outside of its own chip and manufacturing supply chain. According to Bloomberg reporting, the agreement covers a 1.2 trillion parameter system tailored to Apple’s requirements, suggesting a long-term, infrastructure-level dependency rather than an experimental pilot. For investors, the question is whether Apple will disclose this relationship in its regulatory filings. A contract of this size would typically qualify as a material commitment or contingency, the kind of arrangement that companies are expected to flag in their quarterly risk-factor updates with the SEC.
Apple’s most recent quarterly filing, a Form 10-Q covering the fiscal quarter ended December 27, 2025 and filed with regulators, includes general language about AI-related risks and business contingencies. It discusses competitive pressures in emerging technologies, the costs of cloud infrastructure, and the possibility of relying on third parties for key services. But it contains no specific mention of the Gemini arrangement, no line-item disclosure tied to a billion-dollar AI licensing deal, and no commitment language that would alert shareholders to the scale of the reported partnership. If the contract is finalized at or near the reported value, Apple’s next two quarterly filings will be the first opportunity for the company to formally acknowledge the dependency in its risk factors.
What Bloomberg and the AP confirmed about the 1.2 trillion parameter model
The core reporting on this deal comes from two sources. Bloomberg reported that Apple plans to use a 1.2 trillion parameter Google Gemini model to power the new Siri, describing the system as a custom build tailored to Apple’s needs and placing the annual cost at approximately $1 billion. Separately, Associated Press coverage confirmed that Google’s AI will power the new Siri and other Apple Intelligence tools, framing the move as a significant shift in Apple’s AI strategy ahead of the WWDC unveiling.
A model with 1.2 trillion parameters would rank among the largest known AI systems deployed in a consumer product. Parameter count is an imperfect but useful proxy for a model’s capacity to process and generate language, handle complex queries, and perform reasoning tasks. Larger models tend to exhibit better performance on a wide range of benchmarks, though they also demand far more compute and careful optimization to run at scale. Apple’s choice to license a model of this size, rather than build one internally for the first wave of its revamped Siri, suggests the company concluded it could not match Google’s training infrastructure and data advantages within its product timeline.
The June 8 WWDC keynote is expected to be the public debut of this rebuilt Siri, giving Apple a narrow window to demonstrate that the investment translates into a noticeably better user experience. Users will be looking for faster responses, more natural conversation, and deeper integration with apps and services across the Apple ecosystem. If those improvements are obvious, the reliance on Google’s technology may be easier for Apple to justify. If the changes feel incremental, investors and developers may question why Apple chose to pay a rival at such a large scale rather than continue investing solely in its own models.
So far, no official Apple statement has confirmed the partnership’s terms, the model’s parameter count, or the annual price. Google has also not released public details about the custom Gemini variant. Both companies have declined to comment on the specifics of the arrangement in any on-the-record capacity, leaving the Bloomberg and AP reporting as the strongest available evidence. That information gap adds another layer of uncertainty for analysts trying to assess the long-term cost structure of Apple’s AI push.
Gaps in disclosure and what to watch in Apple’s next two filings
Several important questions remain open. First, Apple’s December 2025 10-Q does not reference the Gemini deal by name, by dollar amount, or by any identifiable proxy. That absence could mean the contract was not yet finalized as of the filing date, or it could reflect a judgment by Apple’s legal team that the arrangement did not yet meet the threshold for material disclosure. Either way, the gap between the scale of the reported deal and the silence in Apple’s public filings creates an information asymmetry for investors who do not closely track technology news.
Second, no primary document confirms the 1.2 trillion parameter figure or the $1 billion annual cost. These numbers originate from Bloomberg’s reporting and have not appeared in any SEC exhibit, Apple press release, or Google disclosure. Until one of the two companies acknowledges the terms, the figures carry the weight of strong institutional journalism but not official confirmation. That distinction matters because long-lived licensing deals can evolve: pricing can change with usage, performance targets can be renegotiated, and exclusivity provisions can be added or removed over time.
Third, the data-sharing and integration terms between Apple and Google remain entirely undisclosed. Siri processes sensitive user queries, from health questions to financial requests, and the routing of that data through a Google-built model raises privacy and competitive questions that go beyond raw performance. Investors and regulators will want to know whether Apple can constrain how user data is used to train or fine-tune Google’s systems, whether queries are anonymized or processed on-device before reaching Gemini, and whether Apple retains the ability to switch providers without disrupting core functionality.
These unknowns place more weight on Apple’s upcoming filings. Observers will be watching for any new language in risk factors that explicitly references dependence on third-party AI infrastructure, as well as any note disclosures that quantify long-term purchase commitments for cloud or model access. Even if the company does not name Google, more granular descriptions of AI-related obligations could signal that the Gemini contract has been fully executed and is now shaping Apple’s cost base.
At the same time, Apple faces a strategic balancing act. Over-disclosing the extent of its reliance on a competitor’s model could undercut the narrative that Apple Intelligence is a uniquely Apple-controlled experience. Under-disclosing, however, risks criticism from shareholders who expect clear visibility into billion-dollar recurring obligations. How the company threads that needle in its next two 10-Qs will be an early test of its governance approach to large-scale AI partnerships.
For now, the Gemini deal underscores how quickly the AI landscape has forced even the most vertically integrated companies to rethink their approach. Apple built its reputation on owning the full stack-from custom silicon to tightly controlled software-but in this cycle it is turning to Google for the core intelligence behind one of its most prominent features. Whether that choice proves to be a temporary bridge or the start of a longer-term dependency will become clearer as Apple discloses more, both on stage at WWDC and in the fine print of its SEC filings.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
