Apple has officially confirmed its selection of Google's Gemini AI models to power the next iteration of Siri. This significant collaboration, announced on Monday, sees two long-standing tech rivals joining forces in the rapidly evolving artificial intelligence arena.
Apple stated that after thorough evaluation, Google's technology presented the most capable foundation for its Apple Foundation Models, expressing enthusiasm for the innovative user experiences it will facilitate. This news immediately boosted shares for Alphabet, Google's parent company, propelling its market capitalisation past the $4 trillion mark for the first time. Alphabet's shares saw a 1.7% rise to $334.04 on Monday, allowing it to recently surpass Apple as the world's second-most valuable corporation, behind Nvidia.
A Multi-Year, Multi-Billion Pound Partnership
This multi-year partnership, reportedly under negotiation since 2024, will see Apple paying approximately $1 billion annually for access to Google's technology. Under the terms, Apple will deploy a customised 1.2 trillion-parameter version of Google's Gemini model. This represents a substantial upgrade from Apple's current 150 billion-parameter cloud-based system, being nearly eight times larger.
The integration of Gemini follows a year-long delay in Apple's promised Siri overhaul. Originally, Apple had aimed to launch the AI-enhanced assistant in 2025 but acknowledged the process was "taking longer than we anticipated." Reports indicate that Apple considered models from various providers, including OpenAI and Anthropic, before determining that Google offered superior capabilities alongside more favourable financial terms. For more on how other companies are developing AI, see Anthropic unveils healthcare AI tools days after OpenAI.
The revamped Siri is expected to debut with iOS 26.4, likely in March or April. Key enhancements will include improved contextual understanding, on-screen awareness, and the ability to execute complex, multi-step tasks across different applications.
Privacy and Strategic Implications
Despite partnering with a direct competitor, Apple has structured the agreement to uphold its stringent privacy commitments. Google's Gemini model will operate entirely on Apple's Private Cloud Compute servers. This ensures that no user data will be shared with Google or utilised for model training, a crucial point for Apple, which often emphasises its privacy-centric approach. Research from organisations like the Electronic Frontier Foundation regularly highlights the importance of such privacy safeguards in the age of AI^ the importance of such privacy safeguards in the age of AI.
Google's technology will specifically manage Siri's summariser and planner functions. These components are responsible for synthesising information and executing intricate tasks. Meanwhile, some features will continue to rely on Apple's in-house models.
This deal signifies both a strategic triumph for Google and an implicit acknowledgement from Apple that the rapid pace of AI development requires collaboration, even for the world's most valuable companies. While Apple intends to develop its own trillion-parameter model eventually, aiming to replace Gemini with an in-house solution, the timeline for this remains uncertain. This move echoes a broader trend of companies seeking external expertise, as detailed in our AI Vendor Vetting Checklist.
What are your thoughts on this significant partnership between Apple and Google? Share your predictions for the impact on Siri in the comments below.
Latest Comments (2)
The 1.2 trillion parameter figure for Gemini is quite large, but the performance gain from model size alone isn't always linear, especially without clear benchmark results for this customized version. I'd be interested to see how it performs on established multimodal benchmarks compared to the publicly available Gemini Ultra.
The jump from 150B to 1.2T parameters for Siri really highlights the arms race for model size. We're seeing similar pressures, albeit on a smaller scale, trying to balance student engagement with model cost in our LLM tutors. Bigger isn't always better for specific tasks, but general-purpose assistants like Siri definitely benefit from the scale.
Leave a Comment