Apple’s reported decision to rely on Google’s Gemini models to power a new generation of Siri signals a subtle but consequential shift in its approach to artificial intelligence. For two decades Apple has championed vertical integration, arguing that tight control over hardware, software and services delivers superior products. In AI—arguably the most strategic layer in modern computing—it appears increasingly willing to compromise.
According to people familiar with the discussions, Apple is exploring the use of Google’s Gemini models and cloud infrastructure to support its own “Apple Foundation Models”, which would continue to run on-device and within Apple’s private cloud. If finalised, the arrangement would allow Apple to retain control of the user experience while outsourcing the most expensive and uncertain element of modern AI: training and maintaining frontier-scale models. The move would reflect economic realism rather than technological retreat.
Building leading AI systems now demands extraordinary capital, data and computing power. Microsoft, Amazon and Meta Platforms have committed tens of billions of dollars to models and data centres, often with unclear payoffs. Apple has largely avoided this arms race. Its profits still rest on hardware margins and ecosystem loyalty, not on selling AI services to third parties. Turning to Google would allow it to modernise Siri without embarking on an open-ended spending contest.
For Google, the prospective deal would amount to a significant vote of confidence. After a hesitant start to the generative-AI boom triggered by OpenAI’s ChatGPT, Google has steadily rebuilt credibility. Gemini is no longer merely competitive; it is now viewed as robust enough to underpin the most visible interface on hundreds of millions of Apple devices. Such validation may matter more than benchmark results or marketing claims.
The talks also highlight a broader shift in AI’s economics. Cloud computing followed a similar trajectory, evolving from a source of differentiation into essential but largely interchangeable infrastructure. AI models appear to be moving in the same direction. As they become more modular, advantage flows less from owning the model and more from controlling distribution, integration and trust. Apple seems content to own the interface, even if the intelligence beneath it is supplied by a partner.
Still, the approach carries risks. Apple and Google are already bound by a lucrative search agreement that has drawn sustained regulatory scrutiny. Deepening their ties in AI may reinforce concerns that dominant firms entrench their power through private arrangements. There is also a strategic tension: Siri’s intelligence would partly depend on a rival whose business relies on scale and data aggregation, while Apple’s brand is built on privacy and control.
Whether or not the partnership is ultimately concluded, the logic behind it is revealing. In the emerging AI economy, controlling the gateway to users may matter more than owning the engine itself. Apple appears increasingly prepared to test that proposition.
-
Research Team at Shweta Labs
