Apple’s ambitious plan to reinvent Siri involves a $1 billion annual pact with Google and the creation of a sophisticated hybrid AI system. The deal, which is reportedly being finalized, will give Apple access to a 1.2 trillion parameter Google Gemini model. However, this powerful AI will run exclusively on Apple’s own Private Cloud Compute servers, creating a unique system that leverages Google’s advanced model while adhering to Apple’s strict privacy policies by “walling off” all user data from Google.
This “behind-the-scenes” partnership is Apple’s interim solution to its generative AI deficit. After testing competitor models from OpenAI and Anthropic, Apple concluded that Google’s technology was necessary to power the “Glenwood” project, the internal effort to fix Siri. The new assistant, code-named “Linwood” and planned for a spring release, will represent a quantum leap in capability, moving from Apple’s 150-billion parameter models to Google’s 1.2 trillion parameter engine for its most complex tasks.
The new hybrid system will be selective in how it allocates tasks. Google’s Gemini model will be tasked with the “summariser” and “planner” functions, which require deep contextual understanding and the ability to plan multi-step actions. Simpler requests and other features of Apple Intelligence will continue to be processed by Apple’s smaller, proprietary models, both on-device and in its private cloud. This approach aims to optimize for both power and privacy.
The deal is a significant win for Google’s AI division, positioning it as a key technology supplier to its biggest competitor, similar to its lucrative Safari search deal but without the public branding. The partnership is separate from earlier talks about a direct Gemini chatbot integration, which failed to materialize. Apple has instead opted for a deeper, more fundamental integration into Siri’s core architecture, while still offering ChatGPT as an optional, separate chatbot.
Despite the $1 billion annual investment, Apple views this as a temporary measure. The company’s long-term strategy is to achieve AI independence. Its internal teams are reportedly working on a 1 trillion parameter cloud model intended to eventually replace Gemini. However, catching Google, whose Gemini 2.5 Pro is a top-performing model, remains a significant challenge, suggesting this “interim” solution could last for several years.
Apple’s $1B Siri Fix: A Hybrid System Running Google’s AI on Apple’s Servers
