Copyright MacRumors
Apple is planning to use an AI model designed by Google to power the smarter version of Siri that's coming out in 2026, and the technology is going to cost Apple approximately $1 billion per year. According to Bloomberg, Apple is close to establishing a deal for a 1.2 trillion parameter Google Gemini artificial intelligence model. The $1 billion estimated annual payment that Apple will make to Google for access to the Gemini model is just a fraction of what Google pays Apple to be the default search engine on Apple devices. Google paid Apple $20 billion for key search engine placement in 2022, and the payment may have gone up since then. Apple plans to use Gemini for Siri's summarizer and planner functions. Summarization features understand and condense information, while planner functions look at a task, determine the steps to complete the task, and then decide on the actions to take to get it done. A request like "read this PDF and tell me the important points" would use a summarizer function. Something more complicated like "schedule a lunch meeting with Bob" would use a planner function because it would require multiple steps like checking a calendar, picking a time, and sending a message. Gemini won't power all of Siri's features, and Apple will continue to use its own models for some capabilities. Apple plans to run the Gemini model on its own Private Cloud Compute servers, so it won't be paying Google for processing, but it will need to pay to power those servers. Apple is already expanding its Private Cloud Compute server setup in anticipation of the launch of the new version of Siri. Apple is expected to roll out the new version of Siri around March or April 2026.
