Apple will power Siri with Google’s Gemini AI in a billion-dollar deal

That, at least, is the story rolling out this week after a Bloomberg report saying Apple plans to run a customized version of Google’s Gemini model inside its Apple Intelligence effort to power the long-promised, AI-heavy revamp of Siri. The deal being negotiated would see Apple pay roughly $1 billion a year for access to the model and Google’s tooling.

Apple built its reputation on control: chips, software, services. So the idea that Apple would subcontract a core part of the next-generation Siri to Alphabet is striking. The reported deal is bigger than the usual enterprise cloud contract — both in cash and in symbolism. Bloomberg’s reporting says the customized Gemini that Apple would deploy uses about 1.2 trillion parameters, a scale that dwarfs the roughly 150-billion-parameter model Apple currently uses in some cloud features. Running something that big, and keeping user data private, requires serious infrastructure — which is why Gurman reports Apple will host the model on its own Private Cloud Compute servers rather than hand user traffic straight to Google’s datacenters..

A stopgap, or a strategic pivot?

If you squint, the arrangement makes sense as a practical fix. Apple’s internal AI efforts have lagged — and engineers have delayed the launch of the full Siri overhaul more than once. According to reporting going back months, Apple evaluated models from OpenAI and Anthropic before moving toward Google. The Gemini tie-up would let Apple ship a much more capable assistant while it keeps developing its own models behind the scenes. Executives have signaled that third-party models could be part of Apple Intelligence going forward.

Timing: “next spring,” according to the company

Apple CEO Tim Cook has told investors the reworked Siri is expected to arrive next spring — a timeline that’s echoed in multiple reports and in Bloomberg’s ongoing coverage of the project. That gives Apple roughly a few quarters to finish integration, test privacy controls, and iron out UX issues before a public rollout.

Related /

Privacy, optics and the technical tradeoffs

Apple will almost certainly stress the privacy architecture here: the claim is that the Gemini instance will run on Apple’s Private Cloud Compute and that Apple will continue using its own models for certain tasks that remain on-device or within narrowly defined workflows. But handing a critical part of a flagship feature to Google raises immediate questions. Will user prompts be sent to a Google-trained model even if the instance is hosted by Apple? How will telemetry and logging be handled? Will there be transparency for users about when Google’s tech is being used? Those are the kinds of product and legal questions that could define whether this looks like a pragmatic partnership — or a risky dependence on a competitor. Reuters’ coverage picks up many of the same points about scale and the tentative nature of negotiations.

Business reality: big checks between big techs

It’s worth noting the bigger business backdrop. Apple and Google have a long commercial relationship — Google pays Apple to be the default search engine on iPhones — and a billion-dollar-a-year AI license would be another major revenue line for Alphabet. For Apple, the cost is high, but the calculus is simple: ship a product that works and keeps customers happy, or keep burning cycles trying to catch up on models while losing ground in the market.

What it means for users and the market

For users, the promise is obvious: a smarter Siri that can summarize complex threads, make plans, synthesize info across apps and contexts, and generally do the “assistant” work people expect from modern LLM-backed tools. For competitors, Apple’s outsourcing part of the assistant to Google will be a fascinating twist; many will watch how Apple frames the integration and whether it can carve a privacy-forward narrative that distinguishes its approach from plain-vanilla cloud assistants.

For Apple’s AI roadmap, the deal reads like a conditional commitment: use the best available external tech now, but keep building your own systems until you don’t need to. Several reports emphasize Apple’s continued investment in homegrown models and infrastructure that might one day replace Gemini in whole or in part.

If true, this would be one of the more pragmatic, if awkward, alliances in recent tech history: an engineering-first, “ship now, build later” approach from Apple that acknowledges the massive resource cost of cutting-edge LLMs. It’s a reminder that, in the AI era, even the most vertically integrated companies sometimes have to borrow scale — and that “privacy-first” product promises are getting a lot harder to keep in isolation.