Siri’s Next Brain Could Be Google’s Gemini
Toggle Dark Mode
It’s not a big secret that Apple has been struggling for years to get Siri to live up to its full potential. Now, it appears that it may be looking for a helping hand from Google to get it there.
Siri arrived to great fanfare with the iPhone 4S in 2011, but sadly seemed to fall to the company’s “B team” after the loss of its two most enthusiastic boosters, Steve Jobs and software chief Scott Forstall. The voice assistant languished for years with only minor updates, becoming the butt of a great many jokes.
Although we saw a few hints in 2019 that Apple was getting serious about Siri, we heard nothing concrete until Apple took the stage during its 2024 Worldwide Developers Conference (WWDC) to show off Apple Intelligence with a much more personally aware Siri — one that could ferret out information on your iPhone and put the pieces together to answer your requests the way a human assistant would.
It was an exciting presentation. Finally, our long Siri nightmare might soon be over. Sadly, the original presentation turned out to be a lot of smoke and mirrors; eight months later, when it was expected to arrive in iOS 18.4, Apple announced that it would be delayed until early 2026.
The official line was simply that Apple needed more time to get it right, but there was also quite a bit of shuffling behind the scenes, and suggestions that the original AI models being used to improve Siri weren’t going to cut it, and Apple was starting over from the ground up.
There’s a lot at stake here for Apple, as the company isn’t just striving to create a better voice assistant; its broader ambition is to make Siri the cornerstone of an entire Apple Home ecosystem. However, it also appears to be acknowledging that its own in-house skills aren’t going to get it there — at least not as quickly as it would like.
Apple is behind the curve on AI development in key areas. While Apple has been integrating machine learning technologies into the iPhone for years to power features like computational photography, Google and OpenAI have a significant head start in developing large language models (LLMs) that power Gemini and ChatGPT. This is a fundamentally different set of AI technologies, and it may be too late for Apple to catch up, especially since its rivals aren’t exactly sitting still here.
The good news (sort of) is that Apple doesn’t plan to let its hubris get in the way of delivering the LLM technologies it needs. According to Bloomberg’s Mark Gurman, the company has a backup plan to license the necessary LLMs from a third party.
So far, this remains strictly a “Plan B,” as Apple is still hoping to have its own LLMs up to snuff in time. However, Gurman’s sources say it’s already engaged in “early discussions” about using Google’s Gemini to power the new Siri.
To be clear, this isn’t about replacing Siri with Gemini Live — it will likely be a subzero summer in Cupertino before that happens — but rather having Google develop a custom AI model to serve as Siri’s foundation.
In other words, it would still be Apple’s Siri, but it would be “Powered by Google.”
Another critical point in Gurman’s report is that this would not in any way compromise Apple’s stance on privacy. “Google has started training a model that could run on Apple’s servers,” Gurman says (emphasis mine).
In other words, the Gemini-powered Siri would still run entirely on Apple’s Private Cloud Compute (PCC) architecture, which is designed with a zero-knowledge philosophy from the ground up.
The downside to this plan, in addition to Apple outsourcing Siri’s brain, is that these LLMs wouldn’t run on-device. Siri requests would need to go to Apple’s PCC servers to be processed, just like they did five years ago, before Apple gave us on-device Siri processing in iOS 15.
While Gurman’s report focuses on Gemini, perhaps because those discussions are the most recent initiative, Google isn’t the only contender. Apple has also explored similar partnerships with Anthropic and OpenAI to determine if the LLMs used by Claude or ChatGPT might be a better fit.
It’s also far from a done deal. “Apple is still several weeks away from making a decision on whether to continue using internal models for Siri or move to a partner,” Gurman says. If it does go down the third-party road, it still reportedly has no idea who that partner may be.
Internally, Apple is holding a bake-off to see which approach will work best. The company is simultaneously developing two versions of the new Siri: one dubbed Linwood that is powered by its models and another code-named Glenwood that runs on outside technology.
Mark Gurman
Anthropic was reportedly set to be Apple’s pick — if it chose to go with a partner — but Gurman says that the financial terms prevented it from being a slam dunk, leading Apple to explore alternatives.
Gurman also notes that these discussions are unrelated to Apple’s work on adding more extensions to Apple Intelligence, which are also still ongoing. A direct integration with Google Gemini is still in the cards, similar to how ChatGPT works with Apple Intelligence now. However, it’s unclear what will happen to rival third-party chatbot extensions if Google or OpenAI begin powering the core Siri functionality.
[The information provided in this article has NOT been confirmed by Apple and may be speculation. Provided details may not be factual. Take all rumors, tech or otherwise, with a grain of salt.]
