Apple to pay Google $1 billion a year for Gemini-powered Siri overhaul

Apple, Google

Apple’s upcoming Siri overhaul will be powered in part by Google’s Gemini, according to fresh details from Bloomberg’s Mark Gurman. The two tech giants are reportedly close to sealing a deal worth about $1 billion per year, giving Apple access to Google’s 1.2 trillion parameter AI model — a significant leap from Apple’s own 150 billion parameter cloud-based model.

The partnership emerged after Apple’s internal “AI bake-off”, which saw Google outbid Anthropic for the contract to provide large-scale AI capability for Siri’s next-generation features. Sources told Bloomberg that the decision leaned less on raw performance and more on cost-effectiveness.

How will the deal work?

Under the agreement, Gemini will power Siri’s “summariser” and “planner” functions — essentially the intelligence behind understanding context and executing complex commands. However, Apple will retain some functions on its in-house models, ensuring that parts of Siri’s operations remain fully within Apple’s control. Importantly, Gemini will run on Apple’s own servers, inside its Private Cloud Compute system, meaning user data won’t leave Apple’s infrastructure or be shared with Google.

The collaboration, codenamed Project Glenwood, is being led by Mike Rockwell, who recently took charge of Apple’s Siri revamp. It marks one of Apple’s most significant external dependencies in years — a pragmatic move, given its lag in large-scale generative AI.

Still, Apple isn’t giving up on going solo. Bloomberg reports that Apple’s internal models team is already working on a 1 trillion parameter cloud-based model, aiming for consumer rollout as soon as next year. That effort, however, comes amid steady AI talent losses, including key members of the models division.

Apple’s partnership with Google reflects its short-term strategy: use Gemini to bridge the gap while building its own foundation models. But whether Cupertino can catch up with the rapid evolution of Gemini and other AI rivals will be one of the biggest questions of 2026 — and perhaps the true test of its long-term AI ambitions.