Apple could pay Google $1 billion a year to power Siri’s next big upgrade with Gemini: Here’s what we know
Apple is nearing a deal to pay around $1 billion a year for an ‘ultrapowerful’ 1.2 trillion-parameter AI model from Google, according to a report by Bloomberg. The new model would be used by the Cupertino-based tech giant to deliver its revamped Siri update that it had promised at WWDC 2024 but has since been delayed multiple times. Reportedly, the two companies are now finalizing an agreement following an extensive evaluation process.
Apple is said to have previously thought about using other third-party AI models for powering the Siri overhaul and was also in favour of using Anthropic’s models, but the financial arrangements didn’t work out. Apple later chose to go with Google, with the two companies already having a deal in place where the iPhone maker gets paid for using Google as the default search engine on its devices.
The Siri revamp with the third-party model is being internally dubbed as Glenwood and is said to be led by Vision Pro headset creator Mike Rockell along with software engineering chief Craig Federighi. The updated Siri is expected to be rolled out with the iOS 26.4 update, which is code-named Linwood.
What will the Gemini partnership do for Siri?
The Gemini model is said to handle summarization and planner functions for Siri by helping the voice assistant synthesize information and decide how to execute complex tasks. Meanwhile, some of the Siri features could continue to rely on Apple’s in-house models. Apple is reportedly only relying on Google as an interim solution until its own AI models are powerful enough to power the Siri update.
The new Gemini model will run on Apple’s Private Cloud Compute servers in order to ensure that user data remains out of Google’s infrastructure.
The partnership is unlikely to be promoted publicly and Apple is said to treat Google as a behind-the-scenes technology supplier. The agreement will also not include using Gemini directly into Siri as a chatbot.
Notably, Siri currently routes the more complex queries to ChatGPT, but there is no option for getting them answered via any other chatbot. Reportedly, the discussions between Google and Apple about using Gemini as a chatbot in Siri came close to fruition in 2024 and earlier in the year, but it couldn’t materialize into a feature.
With Gemini being seen as a stopgap solution, Apple’s models team is said to be working on a 1 trillion-parameter cloud-based model which it believes could be ready for consumer application by next year.
Internally, Apple executives believe that the in-house model could reach a similar quality level as Gemini’s custom version.
Post Comment