Ryan Haines / Android Authority
TL;DR
- Apple is reportedly ready to pay Google $1 billion per year to develop and maintain a specialized AI model.
- This special version of the Gemini model will be used to power the next generation of Siri.
- This represents Apple’s desperation to enter the AI race, especially with its inability to build a model comparable to Google’s Gemini.
Voice assistants from tech brands like Google and Amazon have recently undergone significant enhancements powered by AI. Both Alexa+ and Gemini, which has replaced Google Assistant, are now available to users for testing and critique. In the meantime, Apple’s ambitions to catch up to its competitors with Siri continue to drag on. Facing challenges in developing a conversational AI chatbot over the past several months, Apple reportedly considered partnering with other AI companies, such as OpenAI, Anthropic, or Google, and was later believed to have chosen a partner.
Over the weekend, Bloomberg‘s Mark Gurman reported that with the expected timeline for releasing a revamped Siri drawing closer, Apple has decided to use Gemini to power Siri’s next generation. In a more recent update, Gurman shares additional details about the deal, including Apple’s plans to pay $1 billion annually to Google for deploying an AI model to power up Siri.
According to the report, Apple plans to use a model with 1.2 trillion parameters (or variables used to train a particular model), which surpasses Apple’s existing 150-billion-parameter model that powers Apple Intelligence. The broad learning base should enable Siri to provide more elaborate responses and understand more language nuances than ever. While Google isn’t transparent about the size of models’ parameters, the Gemini Ultra model from 2023 was claimed to have 1.56 trillion parameters (though this cannot be verified), which might make the custom model for Apple one of the most comprehensive ones ever developed by Google. That could mean Siri may have more knowledge (to put it more crudely) than Gemini on Android phones.
Don’t want to miss the best from Android Authority?
The reports indicate that instead of using an existing Gemini model, Apple will utilize a custom version deployed on its own Private Cloud Compute architecture, rather than being hosted on Google’s servers. Apple has reportedly already allocated resources for this purpose.
According to earlier reports, Apple is scheduled to launch the revamped version of Siri in the spring of 2026, with the rollout expected to coincide with the iOS 26.4 update. In its previous announcements, Apple has also claimed a deeper integration within the system, where Siri might be able to accomplish tasks within apps, including third-party ones, likely bringing agentic utility to the iPhone.
Now, considering these changes were first proposed in 2024, some might say that Apple has already lost the plot, though we don’t think that. AI on phones is still evolving, and Apple can still catch up if it not only acts quickly but also ensures the upgraded Siri actually addresses the needs, instead of just bringing an upgrade for show — as it did with the ChatGPT integration in Siri.
Thank you for being part of our community. Read our Comment Policy before posting.
