The USD 1 billion question: Why Apple needs Google to fix Siri

Apple is finalising a deal to pay Alphabet's Google an estimated USD 1 billion annually for access to a powerful artificial intelligence model that will completely overhaul the core technology behind its voice assistant, Siri, reported by Bloomberg.

This financial commitment signals Apple's urgency to catch up in the fast-moving AI race and marks a pragmatic strategic alliance between two of the world's most competitive tech rivals.

Acknowledging the AI lag

The decision to license technology, rather than rely solely on its own development, somewhere addresses Siri's widely recognised failure to keep pace with competitors like Google Assistant and Amazon Alexa.

For years, Siri has struggled with complex, multi-step requests and contextual understanding. The new Siri, expected to debut in the spring of 2026, aims to solve these problems by tapping into Google’s advanced Large Language Model (LLM) technology.

The partnership involves a custom version of Google's Gemini AI model. This model, reportedly featuring 1.2 trillion parameters, represents a massive leap in complexity.

To put this in perspective, Apple's current cloud-based AI system has models estimated to be around 150 billion parameters, while its smaller, on-device models are around 3 billion parameters. The sheer scale of the Google model provides the power needed for sophisticated functions like summarisation and task planning.

A temporary, strategic partnership

This deal is not a long-term surrender. Apple views the Google Gemini model as an interim solution, a critical stopgap to refresh Siri's capabilities while its in-house AI research and development team catches up. Apple has reportedly tested models from other major players, including OpenAI's ChatGPT and Anthropic's Claude, before settling on Google as the best fit.

The use of Google’s AI is a behind-the-scenes arrangement. Apple plans to run the licensed Gemini model on its own Private Cloud Compute (PCC) servers. This setup is crucial for Apple's privacy-focused brand: it keeps user data walled off from Google’s main infrastructure, ensuring that customer queries are processed securely without being used to train Google's general models. Publicly, Apple will brand the resulting capabilities as its own, maintaining its image as a vertically integrated technology provider.

The business of collaboration

From a business standpoint, this deal is both a high-stakes investment for Apple and a strategic win for Google.

For Apple, the annual payment is necessary to ensure its flagship mobile experience remains relevant. The move is a rare public acknowledgment of a technological deficit and follows months of delays and internal reorganisation in its AI division, including executive changes made to steer the AI efforts back on track.

For Google, securing Apple as a paying client for a core AI model represents a major victory in the intensifying battle for AI dominance against competitors like Microsoft-backed OpenAI. The USD 1 billion annual fee is pure technology licensing revenue, independent of the existing, far larger deal where Google pays Apple an estimated USD 20 billion per year to remain the default search engine on Safari and other Apple devices. The new AI deal expands the intricate financial dependency between the two giants, securing Google's place at the core of Apple's intelligence layer, at least for the immediate future.