According to a new report from Bloomberg, Apple is having discussions with Google about integrating its Gemini AI model into Apple's iOS ecosystem. Apparently, the Cupertino giant also previously held talks with OpenAI. This potential collaboration could significantly impact the user experience for millions of iPhone users worldwide.
Apple currently faces several challenges in both developing and deploying its own large language model at the scale required to serve its massive user base. It's interest in Gemini is likely not just about acquiring the technology, but rather gaining access to Google's extensive infrastructure. Even if Apple had its own Gemini-level model, serving it to its massive user base would be a challenge. Google, with its TPU-based infrastructure spread worldwide, is one of the few companies capable of handling such a task.
Currently, one of the primary obstacles for Apple is the limited availability of NVIDIA's high-performance GPUs. This is further complicated because of the strained relationship between the two companies over chip issues in the past.
The potential partnership between Apple and Google makes strategic sense for both companies. Google, with its extensive TPU-based infrastructure spread across the globe, is well-positioned to handle the enormous user base that iOS integration would unlock. While the short-term economics may be unclear, the long-term benefits for both companies are evident.
If the deal comes to fruition, it will likely result in a hybrid approach. Apple will focus on developing smaller models that can run on its devices, possibly with an API that allows developers to access certain functionalities without the need to download models independently. More complex queries, would be handled by Gemini in the cloud.
This arrangement aligns with how AI on Google's Pixel and Samsung's Galaxy devices currently functions—handling basic queries on-device while offloading more complex tasks to the cloud. While Apple may continue to develop its own cloud-based models, the company might ultimately find it impractical to build the necessary infrastructure to support such capabilities, opting instead to gradually shift more interactions to on-device processing.
With Apple also having talks with OpenAI, we could even see a model where users ultimately get to choose their AI provider and model similar to how they can choose their search engine provider today.
It is important to note that the details of this potential collaboration remain speculative. Any official announcements will likely be made at the company's Worldwide Developers Conference (WWDC) in June.