Apple is moving away from building every part of its artificial‑intelligence stack in house and toward a hybrid approach: reports say the company has chosen a custom version of Google's Gemini model to power an upgraded Siri that could begin rolling out in spring 2026.
What reporters are saying
Bloomberg and other outlets, citing people familiar with the discussions, reported that Apple has been negotiating to license a custom Gemini model for use on its servers. Mark Gurman of Bloomberg, who has been reporting on Apple’s AI plans since the summer, reiterated on X that “the new Siri will lean on a custom Gemini model running on Apple’s cloud.” Reuters and CNET have also published accounts describing the same arrangement.
Multiple reports say Apple will run the custom Gemini instance on its own cloud compute rather than directing users to Google’s public servers. Apple’s on‑device models, the company says, will continue to handle personal and private data, while more compute‑intensive tasks would be handled by the server‑side Gemini instance.
Why Google — and what it might cost
According to reporting, Apple evaluated other AI providers, including Anthropic, but opted for Google in part because of price. Bloomberg’s coverage cited by other outlets suggests Anthropic’s proposal could have cost Apple roughly $1.5 billion a year, while the Google deal is said to be materially cheaper — with one report saying Apple may pay about $1 billion a year. Neither Apple nor Google has publicly confirmed the sums.
CNET and others note that the selection reflects both commercial and technical tradeoffs: Google’s Gemini is a leading large model and Google has the infrastructure to deliver custom variants. Apple also maintains a long, if sometimes tense, commercial relationship with Google; previous reporting has noted a multibillion‑dollar arrangement that keeps Google as the default search provider on Apple devices.
What this means for Siri and Apple Intelligence
Apple executives have acknowledged that the company needs to accelerate its AI efforts. CEO Tim Cook recently told CNBC that Apple intends to “integrate with more people over time,” and said the company was making progress on an AI‑upgraded Siri slated for release next year. Craig Federighi, Apple’s software chief, has previously said Apple may partner with other models like Gemini in the future.
For users, the change could deliver a more capable Siri: server‑side models can handle complex reasoning, long contexts and multimodal inputs that are difficult to run on phones alone. Sources familiar with Apple’s plans told reporters the integration will be positioned as part of Apple Intelligence — the company’s umbrella for on‑device and cloud enhancements to search, assistant and productivity features.
Apple is reportedly not likely to advertise Google’s role prominently. That decision would align with the company’s history of subsuming third‑party technology into its own branded services.
Privacy, security and competitive questions
Apple has long marketed Siri and much of its software as privacy‑forward, emphasizing on‑device processing for sensitive data. The reported architecture — on‑device models for private information and a server‑side Gemini instance for heavy lifting — attempts to preserve that distinction. But some privacy advocates and analysts say handing complex queries to a third‑party model, even one running on Apple’s cloud, raises questions about data handling, retention and auditability.
There are also antitrust and competition implications. Apple's commercial ties to Google, already scrutinized in previous legal battles over search arrangements, will draw fresh attention as the two companies deepen an AI relationship. Regulators and rivals may examine whether such tie‑ups disadvantage other model providers or entrench existing search and ad ecosystems.
Industry reactions and strategic context
Analysts describe the move as pragmatic: Apple has been slower than some peers to build large server‑side models and has relied on partnerships — for example, integrating OpenAI’s ChatGPT into some features earlier this year. Licensing a custom Gemini could accelerate Apple’s roadmap without the time and expense of developing an equivalent in house.
Critics say the approach underscores how the economics of modern generative AI push even the wealthiest tech companies toward cooperation. Supporters counter that the integration could deliver a better, safer assistant for billions of devices, provided Apple retains control over data flows and transparency about how the system works.
What to watch next
Key questions remain that will shape how the deal is perceived and deployed:
- Will Apple disclose that Gemini underpins the upgraded Siri, and how will it explain data flows to users?
- How will Apple reconcile server‑side processing with its privacy messaging?
- Will regulators scrutinize the financial and technical terms of the agreement given the companies’ broader commercial ties?
Apple, Google and Anthropic did not immediately respond to requests for comment in media reports. For consumers, the near‑term prospect is a more capable Siri and deeper Apple Intelligence features; for the industry, the reported deal is another sign that big tech players are increasingly trading specialized AI capabilities and infrastructure rather than relying solely on internal development.
As Siri’s long‑promised generative upgrades begin to appear, users and watchdogs alike will be watching both the capabilities delivered and the controls Apple builds around them.