Apple to Pay $1 Billion for Google's AI Technology to Revamp Siri

According to sources familiar with the matter, Apple Inc. plans to pay approximately $1 billion annually to use an ultra-high-performance AI model developed by Alphabet Inc., the parent company of Google. This model, with 1.2 trillion parameters, will help Apple advance its long-awaited Siri voice assistant upgrade.

The sources mentioned that after extensive evaluation, the two companies are now finalizing the details of the agreement, which will give Apple access to Google's technology. As the negotiations have not been publicly disclosed, these individuals requested to remain anonymous.

The iPhone maker is hoping to rebuild the underlying architecture of Siri using Google's technology to lay the foundation for a series of new features to be launched next year. The 1.2 trillion parameters of Google's model, a measure of AI software complexity, will far surpass the parameters of Apple's current model.

Earlier, Apple had considered using other third-party models for this task. However, after testing Google's Gemini model, OpenAI's ChatGPT, and Anthropic's Claude earlier this year, Apple ultimately focused on Google's solution. Apple plans to use Google's technology as a transitional solution until its own in-house model reaches sufficient performance levels.

The new version of Siri is scheduled to be released in the spring of 2024. However, since the release is still months away, the plans and partnership details may still change. Both Apple and Google representatives declined to comment on the matter.

Following the announcement on Wednesday, the stock prices of both companies briefly rose to their daily highs. Apple's stock increased to $271.70, while Alphabet's stock once surged by up to 3.2% to $286.42.

The customized Gemini system represents a significant upgrade from the 150 billion-parameter model currently used in Apple's Intelligent Cloud version. This move will greatly enhance the system's performance, improving its ability to handle complex data and understand context.

Internally, Apple has named the project "Glenwood," led by Mike Rockwell, head of the Vision Pro headset development, and Craig Federighi, head of software engineering. The updated voice assistant, internally codenamed "Linwood," will be included in the iOS 26.4 system.

Under the agreement, Google's Gemini model will handle the core functions of summarization and planning for Siri — helping the voice assistant integrate information and determine the execution of complex tasks. Some other Siri features will still rely on Apple's in-house models.

The model will run on Apple's private cloud infrastructure to ensure user data is kept separate from Google's infrastructure. Apple has already allocated AI server hardware to support the computational power needed for the model.

Despite the scale of the partnership, it is unlikely that the two companies will publicly promote this collaboration. Instead, Apple will position Google as a "behind-the-scenes technology provider," a model that differs from their previous collaboration with Safari, where Google became the default search engine.

Additionally, this agreement is separate from earlier discussions about integrating Gemini directly into Siri as a chatbot. These discussions came close to fruition earlier this year but ultimately did not result in actual functionality. This partnership also does not involve integrating Google's AI search into Apple's operating systems.

In Apple's most recent earnings call, CEO Tim Cook mentioned that, in addition to existing ChatGPT options, Siri is expected to support more chatbots in the future.

Apple is not the only company using Gemini for AI capabilities: Snap and other major enterprises are developing on Google's Vertex AI platform. However, for Apple, this move represents an acknowledgment that it has fallen behind in the AI space and is now willing to rely on external technology to catch up.

That said, sources indicate that Apple does not intend to rely on Gemini as a long-term solution. Despite facing challenges with AI talent retention (including the departure of key model team leaders), management plans to continue developing its own AI technology, to eventually replace Gemini with an in-house solution.

To this end, Apple's model team is working on a cloud-native model with 1 trillion parameters, which is expected to be rolled out for consumer applications as early as next year.

Apple executives believe that this in-house model will match the performance of the customized Gemini model, but catching up with Google's continued upgrades to Gemini will not be easy.

In many of the performance rankings for large language models, the Gemini 2.5 Pro version consistently ranks among the top.

Furthermore, Apple is continuing its plan to introduce "Apple Intelligence" and the updated Siri to the Chinese market. The customized version of Siri for China is expected to use Apple's own models and will not include the Gemini model. Earlier reports indicated that Apple's customized version of Apple Intelligence for China will use an in-house model, paired with a filtering system developed by Alibaba. Apple has also been considering partnering with Baidu to advance its AI business in China.